An intelligent assistant shall be an agent that knows you and the world, can receive your requests or predict your needs, and provide you the right services at the right time with your permission. As smart devices such as Amazon Alexa, Google Home, Ray-ban Meta get popular, Intelligent Assistants are gradually playing an important role in people's lives. The Emergence of wearable devices brings more opportunities and calls for the next generation of Intelligent Assistants. In this talk, we discuss the many challenges and opportunities we face to grow intelligent assistants from voice-only to multi-modal, from context-agnostic to context-aware, from listening to the users' requests to predicting the user's needs, and from server-side to on-device. We describe our LLM-based solutions toward multi-modality, contextualization, personalization, and retrieval-augmentation, and discuss how they are enabled on devices. We expect these new challenges to open doors to new research areas and start a new chapter for providing personal assistance services

We introduce Lumos, the first end-to-end multimodal question-answering system with text understanding capabilities. At the core of Lumos is a Scene Text Recognition (STR) component that extracts text from first person point-of-view images, the output of which is used to augment input to a Multimodal Large Language Model (MM-LLM). While building Lumos, we encountered numerous challenges related to STR quality, overall latency, and model inference. In this paper, we delve into those challenges, and discuss the system architecture, design choices, and modeling techniques employed to overcome these obstacles. We also provide a comprehensive evaluation for each component, showcasing high quality and efficiency.

Large Language Models (LLMs) have demonstrated strong capabilities in comprehending and generating human language, as well as emerging abilities like reasoning and using tools. These advancements have been revolutionizing techniques in every front, including the development of personal assistants. However, their inherent limitations such as lack of factuality and hallucinations make LLMs less suitable for creating knowledgeable and trustworthy assistants.

Large language models (LLMs) demonstrate an impressive ability to internalize knowledge and answer natural language questions. Although previous studies validate that LLMs perform well on general knowledge while presenting poor performance on long-tail nuanced knowledge, the community is still doubtful about whether the traditional knowledge graphs should be replaced by LLMs. In this paper, we askif the schema of knowledge graph (i.e., taxonomy) is made obsolete by LLMs.Intuitively, LLMs should perform well on common taxonomies and at taxonomy levels that are common to people. Unfortunately, there lacks a comprehensive benchmark that evaluates the LLMs over a wide range of taxonomies from common to specialized domains and at levels from root to leaf so that we can draw a confident conclusion. To narrow the research gap, we constructed a novel taxonomy hierarchical structure discovery benchmark named TaxoGlimpse to evaluate the performance of LLMs over taxonomies. TaxoGlimpse covers ten representative taxonomies from common to specialized domains with in-depth experiments of different levels of entities in this taxonomy from root to leaf. Our comprehensive experiments of eighteen LLMs under three prompting settings validate that LLMs perform miserably poorly in handling specialized taxonomies and leaf-level entities. Specifically, the QA accuracy of the best LLM drops by up to 30% as we go from common to specialized domains and from root to leaf levels of taxonomies.

Virtual Intelligent Assistants take user requests in the voice form, perform actions such as setting an alarm, turning on a light, and answering a question, and provide answers or confirmations in the voice form or through other channels such as a screen. Assistants have become prevalent in the past decade, and users have been taking services from assistants like Amazon Alexa, Apple Siri, Google Assistant, and Microsoft Cortana.

Knowledge Graphs (KGs) have been used to support a wide range of applications, from web search to personal assistant. In this paper, we describe three generations of knowledge graphs: entity-based KGs , which have been supporting general search and question answering ( e.g. , at Google and Bing); text-rich KGs , which have been supporting search and recommendations for products, bio-informatics, etc. ( e.g. , at Amazon and Alibaba); and the emerging integration of KGs and LLMs, which we call dual neural KGs. We describe the characteristics of each generation of KGs, the crazy ideas behind the scenes in constructing such KGs, and the techniques developed over time to enable industry impact. In addition, we use KGs as examples to demonstrate a recipe to evolve research ideas from innovations to production practice, and then to the next level of innovations, to advance both science and business.

The Data Management community has explored many options in recent years to improve our publication culture and review processes, ranging from innovative journal-conference hybrids that decouple publication from presentation, incorporating journal-style reviewing for conference-style papers, requesting code reproducibility and code/data availability, multiple submission deadlines in a year, new categories of papers, informal shepherding processes, guidelines for diversity and inclusion, automated COI check, and so on. This panel seeks to examine our many experiments, comparing them with other CS disciplines, and help determine (i) have our experiments worked? (ii) what has their impact been? and (iii) can we do better?

Automatic extraction of product attributes from their textual descriptions is essential for online shopper experience. One inherent challenge of this task is the emerging nature of e-commerce products — we see new types of products with their unique set of new attributes constantly. Most prior works on this matter mine new values for a set of known attributes but cannot handle new attributes that arose from constantly changing data. In this work, we study the attribute mining problem in an open-world setting to extract novel attributes and their values. Instead of providing comprehensive training data, the user only needs to provide a few examples for a few known attribute types as weak supervision. We propose a principled framework that first generates attribute value candidates and then groups them into clusters of attributes. The candidate generation step probes a pre-trained language model to extract phrases from product titles. Then, an attribute-aware fine-tuning method optimizes a multitask objective and shapes the language model representation to be attribute-discriminative. Finally, we discover new attributes and values through the self-ensemble of our framework, which handles the open-world challenge. We run extensive experiments on a large distantly annotated development set and a gold standard human-annotated test set that we collected. Our model significantly outperforms strong baselines and can generalize to unseen attributes and product types.

Although product graphs (PGs) have gained increasing attentions in recent years for their successful applications in product search and recommendations, the extensive power of PGs can be limited by the inevitable involvement of various kinds of errors. Thus, it is critical to validate the correctness of triples in PGs to improve their reliability. Knowledge graph (KG) embedding methods have strong error detection abilities. Yet, existing KG embedding methods may not be directly applicable to a PG due to its distinct characteristics: (1) PG contains rich textual signals, which necessitates a joint exploration of both text information and graph structure; (2) PG contains a large number of attribute triples, in which attribute values are represented by free texts. Since free texts are too flexible to define entities in KGs, traditional way to map entities to their embeddings using ids is no longer appropriate for attribute value representation; (3) Noisy triples in a PG mislead the embedding learning and significantly hurt the performance of error detection. To address the aforementioned challenges, we propose an end-to-end noise-tolerant embedding learning framework, PGE, to jointly leverage both text information and graph structure in PG to learn embeddings for error detection. Experimental results on real-world product graph demonstrate the effectiveness of the proposed framework comparing with the state-of-the-art approaches.

Understanding product attributes plays an important role in improving online shopping experience for customers and serves asan integral part for constructing a product knowledge graph. Most existing methods focus on attribute extraction from text description or utilize visual information from product images such as shape and color. Compared to the inputs considered in prior works, a product image in fact contains more information, represented by a rich mixture of words and visual clues with a layout carefully designed to impress customers. This work proposes a more inclusive framework that fully utilizes these different modalities for attribute extraction.Inspired by recent works in visual question answering, we use a transformer based sequence to sequence model to fuse representations of product text, Optical Character Recognition (OCR) tokens and visual objects detected in the product image. The framework is further extended with the capability to extract attribute value across multiple product categories with a single model, by training the decoder to predict both product category and attribute value and conditioning its output on product category. The model provides a unified attribute extraction solution desirable at an e-commerce platform that offers numerous product categories with a diverse body of product attributes. We evaluated the model on two product attributes, one with many possible values and one with a small set of possible values, over 14 product categories and found the model could achieve 15% gain on the Recall and 10% gain on the F1 score compared to existing methods using text-only features.

Knowledge graphs have been pivotal in supporting downstream applications like search, recommendation, and question answering, among others. Therefore, knowledge graphs have naturally become key enabling technologies in e-Commerce platforms. Developing a high coverage product knowledge graph is more challenging than generic knowledge graphs. The highly specific and complex domain, the sparsity of training data, along with the dynamic taxonomies and product types, can constrain the resulting knowledge graphs. In this tutorial we present best practices and ML innovations in industry towards building a scalable product knowledge graph. Contributions in this domain benefit from the general literature in areas including information extraction and data mining, tailored to address the specific characteristics of e-Commerce platforms.

Existing recommender systems in the e-commerce domain primarily focus on generating a set of relevant items as recommendations; however, few existing systems utilize underlying item attributes as a key organizing principle in presenting recommendations to users. Mining important attributes of items from customer perspectives and presenting them along with item sets as recommendations can provide users more explainability and help them make better purchase decision. In this work, we generalize the attribute-aware item-set recommendation problem, and develop a new approach to generate sets of items (recommendations) with corresponding important attributes (explanations) that can best justify why the items are recommended to users. In particular, we propose a system that learns important attributes from historical user behavior to derive item set recommendations, so that an organized view of recommendations and their attribute-driven explanations can help users more easily understand how the recommendations relate to their preferences. Our approach is geared towards real world scenarios: we expect a solution to be scalable to billions of items, and be able to learn item and attribute relevance automatically from user behavior without human annotations. To this end, we propose a multi-step learning-based framework called Extract-Expect-Explain (EX3), which is able to adaptively select recommended items and important attributes for users. We experiment on a large-scale real-world benchmark and the results show that our model outperforms state-of-the-art baselines by an 11.35% increase on NDCG with adaptive explainability for item set recommendation.

Recent years have witnessed the success of machine learning and especially deep learning in many research areas such as Vision and Language Processing, Information Retrieval and Recommender Systems, Social Networks and Conversational Agents. Though various learning approaches have demonstrated satisfying performance in perceptual tasks such as associative learning and matching by extracting useful similarity patterns from data, the area still sees a large amount of research needed to advance the ability of reasoning towards cognitive intelligence in the coming years. This includes but is not limited to neural logical reasoning, neural-symbolic reasoning, causal reasoning, knowledge reasoning and commonsense reasoning. The workshop focuses on the research of machine reasoning techniques and their application in various intelligent tasks. It will gather researchers as well as practitioners in the field for discussions, idea communications, and research promotions. It will also generate insightful debates about the recent progress in machine intelligence to a broader community, including but not limited to CV, IR, NLP, ML, DM, AI and beyond.

Information extraction from semi-structured webpages provides valuable long-tailed facts for augmenting knowledge graph. Relational Web tables are a critical component containing additional entities and attributes of rich and diverse knowledge. However, extracting knowledge from relational tables is challenging because of sparse contextual information. Existing work linearize table cells and heavily rely on modifying deep language models such as BERT which only captures related cells information in the same table. In this work, we propose a novel relational table representation learning approach considering both the intra- and inter-table contextual information. On one hand, the proposed Table Convolutional Network model employs the attention mechanism to adaptively focus on the most informative intra-table cells of the same row or column; and, on the other hand, it aggregates inter-table contextual information from various types of implicit connections between cells across different tables. Specifically, we propose three novel aggregation modules for (i) cells of the same value, (ii) cells of the same schema position, and (iii) cells linked to the same page topic. We further devise a supervised multi-task training objective for jointly predicting column type and pairwise column relation, as well as a table cell recovery objective for pre-training. Experiments on real Web table datasets demonstrate our method can outperform competitive baselines by of F1 for column type prediction and by of F1 for pairwise column relation prediction.

Text categorization is an essential task in Web content analysis. Considering the ever-evolving Web data and new emerging categories, instead of the laborious supervised setting, in this paper, we focus on the minimally-supervised setting that aims to categorize documents effectively, with a couple of seed documents annotated per category. We recognize that texts collected from the Web are often structure-rich, i.e., accompanied by various metadata. One can easily organize the corpus into a text-rich network, joining raw text documents with document attributes, high-quality phrases, label surface names as nodes, and their associations as edges. Such a network provides a holistic view of the corpus’ heterogeneous data sources and enables a joint optimization for network-based analysis and deep textual model training. We therefore propose a novel framework for minimally supervised categorization by learning from the text-rich network. Specifically, we jointly train two modules with different inductive biases – a text analysis module for text understanding and a network learning module for class-discriminative, scalable network learning. Each module generates pseudo training labels from the unlabeled document set, and both modules mutually enhance each other by co-training using pooled pseudo labels. We test our model on two real-world datasets. On the challenging e-commerce product categorization dataset with 683 categories, our experiments show that given only three seed documents per category, our framework can achieve an accuracy of about 92%, significantly outperforming all compared methods; our accuracy is only less than 2% away from the supervised BERT model trained on about 50K labeled documents.

Multi-source entity linkage focuses on integrating knowledge from multiple sources by linking the records that represent the same real world entity. This is critical in high-impact applications such as data cleaning and user stitching. The state-of-the-art entity linkage pipelines mainly depend on supervised learning that requires abundant amounts of training data. However, collecting well-labeled training data becomes expensive when the data from many sources arrives incrementally over time. Moreover, the trained models can easily overfit to specific data sources, and thus fail to generalize to new sources due to significant differences in data and label distributions. To address these challenges, we present AdaMEL, a deep transfer learning framework that learns generic high-level knowledge to perform multi-source entity linkage. AdaMEL models the attribute importance that is used to match entities through an attribute-level self-attention mechanism, and leverages the massive unlabeled data from new data sources through domain adaptation to make it generic and data-source agnostic. In addition, AdaMEL is capable of incorporating an additional set of labeled data to more accurately integrate data sources with different attribute importance. Extensive experiments show that our framework achieves state-of-the-art results with 8.21% improvement on average over methods based on supervised learning. Besides, it is more stable in handling different sets of data sources in less runtime.

Knowledge graphs have been used to support a wide range of applications and enhance search and QA for Google, Bing, Amazon Alexa, etc. However, we often miss long-tail knowledge, including unpopular entities, unpopular relations, and unpopular verticals. In this talk we describe our efforts in harvesting knowledge from semi-structured websites, which are often populated according to some templates using vast volume of data stored in underlying databases.

If one customer buys a tennis racket, what are the best 3 complementary products to purchase together? 3 tennis ball packs, 3 headbands, 3 overgrips, or 1 of each respectively? Complementary product recommendation (CPR), aiming at providing product suggestions that are often bought together to serve a joint demand, forms a pivotal component of e-commerce service, however, existing methods are far from optimal. Given one product, how to recommend its complementary products of different types is the key problem we tackle in this work. We first conduct an extensive analysis to correct the inaccurate assumptions adopted by existing work to show that co-purchased products are not always complementary and further propose a new strategy to generate clean distant supervision labels for CPR modeling. Moreover, to bridge in the gap from existing work that CPR does not only need relevance modeling but also requires diversity to fulfill the whole purchase demand, we develop a deep learning framework, P-Companion to explicitly model both relevance and diversity. More specifically, given one product with its product type, P-Companion first uses an encoder-decoder network to predict multiple complementary product types, then a transfer metric learning network is developed to project the embedding of query product to each predicted complementary product type subspace and further learn the complementary relationship based on the distant supervision labels within each subspace. The whole framework can be trained from end-to-end and robust to cold-start products attributed to a novel pretrained product embedding module named Product2Vec, based on graph attention networks. Extensive offline experiments show that P-Companion outperforms state-of-the-art baselines by 7.1% increase on the [email protected] score with well-controlled diversity. Production-wise, we deploy P-Companion to provide online recommendations for over 200M products at Amazon and observe significant gains on product sales and profit.

Online recommendation is an essential functionality across a variety of services, including e-commerce and video streaming, where items to buy, watch, or read are suggested to users. Justifying recommendations, i.e., explaining why a user might like the recommended item, has been shown to improve user satisfaction and persuasiveness of the recommendation. In this paper, we develop a method for generating post-hoc justifications that can be applied to the output of any recommendation algorithm. Existing post-hoc methods are often limited in providing diverse justifications, as they either use only one of many available types of input data, or rely on the predefined templates. We address these limitations of earlier approaches by developing J-Recs, a method for producing concise and diverse justifications. J-Recs is a recommendation model-agnostic method that generates diverse justifications based on various types of product and user data (e.g., purchase history and product attributes). The challenge of jointly processing multiple types of data is addressed by designing a principled graph-based approach for justification generation. In addition to theoretical analysis, we present an extensive evaluation on synthetic and real-world data. Our results show that J-Recs satisfies desirable properties of justifications, and efficiently produces effective justifications, matching user preferences up to 20% more accurately than baselines.

Entity linkage (EL) is a critical problem in data cleaning and integration. In the past several decades, EL has typically been done by rule-based systems or traditional machine learning models with hand-curated features, both of which heavily depend on manual human inputs. With the ever-increasing growth of new data, deep learning (DL) based approaches have been proposed to alleviate the high cost of EL associated with the traditional models. Existing exploration of DL models for EL strictly follows the well-known twin-network architecture. However, we argue that the twin-network architecture is sub-optimal to EL, leading to inherent drawbacks of existing models. In order to address the drawbacks, we propose a novel and generic contrastive DL framework for EL. The proposed framework is able to capture both syntactic and semantic matching signals and pays attention to subtle but critical differences. Based on the framework, we develop a contrastive DL approach for EL, CorDEL, with a simple yet powerful variant called CorDEL-Sum. We evaluate CorDEL with extensive experiments conducted on both public benchmark datasets and a real-world dataset. CorDEL outperforms previous state-of-the-art models by 5.2% on public benchmark datasets. Moreover, CorDEL yields a 29.4% improvement over the current best DL model on the real-world dataset, while reducing the number of training parameters by 96.8%.

Can one build a knowledge graph (KG) for all products in the world? Knowledge graphs have firmly established themselves as valuable sources of information for search and question answering, and it is natural to wonder if a KG can contain information about products offered at online retail sites. There have been several successful examples of generic KGs, but organizing information about products poses many additional challenges, including sparsity and noise of structured data for products, complexity of the domain with millions of product types and thousands of attributes, heterogeneity across large number of categories, as well as large and constantly growing number of products.

How do we surface the large amount of information present in HTML documents on the Web, from news articles to Rotten Tomatoes pages to tables of sports scores? Such information can enable a variety of applications including knowledge base construction, question answering, recommendation, and more. In this tutorial, we present approaches for information extraction (IE) from Web data that can be differentiated along two key dimensions: 1) the diversity in data modality that is leveraged, e.g. text, visual, XML/HTML, and 2) the thrust to develop scalable approaches with zero to limited human supervision.

Taxonomies have found wide applications in various domains, especially online for item categorization, browsing, and search. Despite the prevalent use of online catalog taxonomies, most of them in practice are maintained by humans, which is labor-intensive and difficult to scale. While taxonomy construction from scratch is considerably studied in the literature, how to effectively enrich existing incomplete taxonomies remains an open yet important research question. Taxonomy enrichment not only requires the robustness to deal with emerging terms but also the consistency between existing taxonomy structure and new term attachment. In this paper, we present a self-supervised end-to-end framework, Octet, for Online Catalog Taxonomy EnrichmenT. Octet leverages heterogeneous information unique to online catalog taxonomies such as user queries, items, and their relations to the taxonomy nodes while requiring no other supervision than the existing taxonomies. We propose to distantly train a sequence labeling model for term extraction and employ graph neural networks (GNNs) to capture the taxonomy structure as well as the query-item-taxonomy interactions for term attachment. Extensive experiments in different online domains demonstrate the superiority of Octet over state-of-the-art methods via both automatic and human evaluations. Notably, Octet enriches an online catalog taxonomy in production to 2 times larger in the open-world evaluation.

Given multiple input signals, how can we infer node importance in a knowledge graph (KG)? Node importance estimation is a crucial and challenging task that can benefit a lot of applications including recommendation, search, and query disambiguation. A key challenge towards this goal is how to effectively use input from different sources. On the one hand, a KG is a rich source of information, with multiple types of nodes and edges. On the other hand, there are external input signals, such as the number of votes or pageviews, which can directly tell us about the importance of entities in a KG. While several methods have been developed to tackle this problem, their use of these external signals has been limited as they are not designed to consider multiple signals simultaneously. In this paper, we develop an end-to-end model MultiImport, which infers latent node importance from multiple, potentially overlapping, input signals. MultiImport is a latent variable model that captures the relation between node importance and input signals, and effectively learns from multiple signals with potential conflicts. Also, MultiImport provides an effective estimator based on attentive graph neural networks. We ran experiments on real-world KGs to show that MultiImport handles several challenges involved with inferring node importance from multiple input signals, and consistently outperforms existing methods, achieving up to 23.7% higher [email protected] than the state-of-the-art method.

Product catalogs are valuable resources for eCommerce website. In the catalog, a product is associated with multiple attributes whose values are short texts, such as product name, brand, functionality and flavor. Usually individual retailers self-report these key values, and thus the catalog information unavoidably contains noisy facts. It is very important to validate the correctness of these values in order to improve shopper experiences and enable more effective product recommendation. Due to the huge volume of products, an effective automatic validation approach is needed. In this paper, we propose to develop an automatic validation approach that verifies the correctness of textual attribute values for products. This can be formulated as a task as cross-checking a textual attribute value against product profile, which is a short textual description of the product on eCommerce website. Although existing deep neural network models have shown success in conducting cross-checking between two pieces of texts, their success has to be dependent upon a large set of quality labeled data, which are hard to obtain in this validation task: products span a variety of categories. Due to the category difference, annotation has to be done on all the categories, which is impossible to achieve in real practice.

How do we surface the large amount of information present in HTML documents on the Web, from news articles to scientific papers to Rotten Tomatoes pages to tables of sports scores? Such information can enable a variety of applications including knowledge base construction, question answering, recommendation, and more. In this tutorial, we present approaches for Information Extraction (IE) from Web data that can be differentiated along two key dimensions: 1) the diversity in data modality that is leveraged, e.g. text, visual, XML/HTML, and 2) the thrust to develop scalable approaches with zero to limited human supervision. We cover the key ideas and intuition behind existing approaches to emphasize their applicability and potential in various settings.

Entity matching seeks to identify data records over one or multiple data sources that refer to the same real-world entity. Virtually every entity matching task on large datasets requires blocking, a step that reduces the number of record pairs to be matched. However, most of the traditional blocking methods are learning-free and key-based, and their successes are largely built on laborious human effort in cleaning data and designing blocking keys.

Knowledge graph (e.g. Freebase, YAGO) is a multi-relational graph representing rich factual information among entities of various types. Entity alignment is the key step towards knowledge graph integration from multiple sources. It aims to identify entities across different knowledge graphs that refer to the same real world entity. However, current entity alignment systems overlook the sparsity of different knowledge graphs and can not align multi-type entities by one single model. In this paper, we present a Collective Graph neural network for Multi-type entity Alignment, called CG-MuAlign. Different from previous work, CG-MuAlign jointly aligns multiple types of entities, collectively leverages the neighborhood information and generalizes to unlabeled entity types. Specifically, we propose novel collective aggregation function tailored for this task, that (1) relieves the incompleteness of knowledge graphs via both cross-graph and self attentions, (2) scales up efficiently with mini-batch training paradigm and effective neighborhood sampling strategy. We conduct experiments on real world knowledge graphs with millions of entities and observe the superior performance beyond existing methods. In addition, the running time of our approach is much less than the current state-of-the-art deep learning methods.

The database research community has been notably successful in impacting the industry and academia since the invention of the relational model. Examples of innovation in the last decade include columnar storage for data analytic platforms, cloud data services, HTAP systems, and a new generation of data wrangling systems. Despite this success, critical self-assessment by the community and identifying key opportunities for the future is essential if we are to continue the tradition of impactful research. In the Fall of 2018, following a long tradition that dates back to 1988 [4], and five years after the last such meeting [1], a group of approximately thirty database researchers gathered at the University of Washington, Seattle for two days to discuss the opportunities we have as a community for impactful research. A report from that meeting is now available [2]. The discussions in the Seattle meeting focused not just on technical challenges and opportunities but also on topics related to how we organize ourselves as a community. This VLDB panel follows on from a previous discussion at SIGMOD 2020 [3], to provide a forum for the broader database community to review and debate the findings from the Seattle Report on Database Research [2] as well as to identify other challenges, and opportunities that need to be taken into account.

A knowledge graph (KG) describes entities and relations between them; for example, between entities Amazon and Seattle, there can be a headquarter_located_at relation. Recent years have witnessed broad applications of KG in search (e.g., by Google and Bing) and question answering (e.g., by Amazon Alexa or Google Home). In this talk, we ask the question: Can one build a knowledge graph (KG) for all products in the world?

Knowledge bases, massive collections of facts (RDF triples) on diverse topics, support vital modern applications. However, existing knowledge bases contain very little data compared to the wealth of information on the Web. This is because the industry standard in knowledge base creation and augmentation suffers from a serious bottleneck: they rely on domain experts to identify appropriate web sources to extract data from. Efforts to fully automate knowledge extraction have failed to improve this standard: these automated systems are able to retrieve much more data and from a broader range of sources, but they suffer from very low precision and recall. As a result, these large-scale extractions remain unexploited. In this paper, we present MIDAS, a system that harnesses the results of automated knowledge extraction pipelines to repair the bottleneck in industrial knowledge creation and augmentation processes. MIDAS automates the suggestion of good-quality web sources and describes what to extract with respect to augmenting an existing knowledge base. We make three major contributions. First, we introduce a novel concept, web source slices, to describe the contents of a web source. Second, we define a profit function to quantify the value of a web source slice with respect to augmenting an existing knowledge base. Third, we develop effective and highly-scalable algorithms to derive high-profit web source slices. We demonstrate that MIDAS produces high-profit results and outperforms the baselines significantly on both real-world and synthetic datasets.

As data volume and variety have increased, so have the ties between machine learning and data integration become stronger. For machine learning to be effective, one must utilize data from the greatest possible variety of sources; and this is why data integration plays a key role. At the same time machine learning is driving automation in data integration, resulting in overall reduction of integration costs and improved accuracy. This tutorial focuses on three aspects of the synergistic relationship between data integration and machine learning: (1) we survey how state-of-the-art data integration solutions rely on machine learning-based approaches for accurate results and effective human-in-the-loop pipelines, (2) we review how end-to-end machine learning applications rely on data integration to identify accurate, clean, and relevant data for their analytics exercises, and (3) we discuss open research challenges and opportunities that span across data integration and machine learning.

How can we estimate the importance of nodes in a knowledge graph (KG)? A KG is a multi-relational graph that has proven valuable for many tasks including question answering and semantic search. In this paper, we present GENI, a method for tackling the problem of estimating node importance in KGs, which enables several downstream applications such as item recommendation and resource allocation. While a number of approaches have been developed to address this problem for general graphs, they do not fully utilize information available in KGs, or lack flexibility needed to model complex relationship between entities and their importance. To address these limitations, we explore supervised machine learning algorithms. In particular, building upon recent advancement of graph neural networks (GNNs), we develop GENI, a GNN-based method designed to deal with distinctive challenges involved with predicting node importance in KGs. Our method performs an aggregation of importance scores instead of aggregating node embeddings via predicate-aware attention mechanism and flexible centrality adjustment. In our evaluation of GENI and existing methods on predicting node importance in real-world KGs with different characteristics, GENI achieves 5-17% higher [email protected] than the state of the art.

Estimation of the accuracy of a large-scale knowledge graph (KG) often requires humans to annotate samples from the graph. How to obtain statistically meaningful estimates for accuracy evaluation while keeping human annotation costs low is a problem critical to the development cycle of a KG and its practical applications. Surprisingly, this challenging problem has largely been ignored in prior research. To address the problem, this paper proposes an efficient sampling and evaluation framework, which aims to provide quality accuracy evaluation with strong statistical guarantee while minimizing human efforts. Motivated by the properties of the annotation cost function observed in practice, we propose the use of cluster sampling to reduce the overall cost. We further apply weighted and two-stage sampling as well as stratification for better sampling designs. We also extend our framework to enable efficient incremental evaluation on evolving KG, introducing two solutions based on stratified sampling and a weighted variant of reservoir sampling. Extensive experiments on real-world datasets demonstrate the effectiveness and efficiency of our proposed solution. Compared to baseline approaches, our best solutions can provide up to 60% cost reduction on static KG evaluation and up to 80% cost reduction on evolving KG evaluation, without loss of evaluation quality.

Knowledge graphs have been used to support a wide range of applications and enhance search results for multiple major search engines, such as Google and Bing. At Amazon we are building a Product Graph, an authoritative knowledge graph for all products in the world. The thousands of product verticals we need to model, the vast number of data sources we need to extract knowledge from, the huge volume of new products we need to handle every day, and the various applications in Search, Discovery, Personalization, Voice, that we wish to support, all present big challenges in constructing such a graph. In this talk we describe four scientific directions we are investigating in building and using such a knowledge graph. First, we have been developing advanced extraction technologies to harvest product knowledge from semi-structured sources on the web and from text product profiles. Our annotation-based extraction tool selects a few webpages (typically below 10 pages) from a website for annotations, and can derive XPaths to extract from the whole website with average precision and recall of 97% [1]. Our distantly supervised extraction tool, CERES, uses an existing knowledge graph to automatically generate (noisy) training labels, and can obtain a precision over 90% when extracting from long-tail websites in various languages [1]. Our OpenTag technique extends state-of-the-art techniques such as Recursive Neural Network (RNN) and Conditional Random Field with attention and active learning, to achieve over 90% precision and recall in extracting attribute values (including values unseen in training data) from product titles, descriptions, and bullets [3].

Extraction of missing attribute values is to find values describing an attribute of interest from a free text input. Most past related work on extraction of missing attribute values work with a closed world assumption with the possible set of values known beforehand, or use dictionaries of values and hand-crafted features. How can we discover new attribute values that we have never seen before? Can we do this with limited human annotation or supervision? We study this problem in the context of product catalogs that often have missing values for many attributes of interest.

There is now more data to analyze than ever before. As data volume and variety have increased, so have the ties between machine learning and data integration become stronger. For machine learning to be effective, one must utilize data from the greatest possible variety of sources; and this is why data integration plays a key role. At the same time machine learning is driving automation in data integration, resulting in overall reduction of integration costs and improved accuracy. This tutorial focuses on three aspects of the synergistic relationship between data integration and machine learning: (1) we survey how state-of-the-art data integration solutions rely on machine learning-based approaches for accurate results and effective human-in-the-loop pipelines, (2) we review how end-to-end machine learning applications rely on data integration to identify accurate, clean, and relevant data for their analytics exercises, and (3) we discuss open research challenges and opportunities that span across data integration and machine learning.

As data volume and variety have increased, so have the ties between machine learning and data integration become stronger. For machine learning to be effective, one must utilize data from the greatest possible variety of sources; and this is why data integration plays a key role. At the same time machine learning is driving automation in data integration, resulting in overall reduction of integration costs and improved accuracy. This tutorial focuses on three aspects of the synergistic relationship between data integration and machine learning: (1) we survey how state-of-the-art data integration solutions rely on machine learning-based approaches for accurate results and effective human-in-the-loop pipelines, (2) we review how end-to-end machine learning applications rely on data integration to identify accurate, clean, and relevant data for their analytics exercises, and (3) we discuss open research challenges and opportunities that span across data integration and machine learning.

The web contains countless semi-structured websites, which can be a rich source of information for populating knowledge bases. Existing methods for extracting relations from the DOM trees of semi-structured webpages can achieve high precision and recall only when manual annotations for each website are available. Although there have been efforts to learn extractors from automatically generated labels, these methods are not sufficiently robust to succeed in settings with complex schemas and information-rich websites. In this paper we present a new method for automatic extraction from semi-structured websites based on distant supervision. We automatically generate training labels by aligning an existing knowledge base with a website and leveraging the unique structural characteristics of semi-structured websites. We then train a classifier based on the potentially noisy and incomplete labels to predict new relation instances. Our method can compete with annotation-based techniques in the literature in terms of extraction quality. A large-scale experiment on over 400,000 pages from dozens of multi-lingual long-tail websites harvested 1.25 million facts at a precision of 90%.

Querying heterogeneous and large-scale knowledge graphs is expensive. This paper studies a graph summarization framework to facilitate knowledge graph search. (1) We introduce a class of reduced summaries. Characterized by approximate graph pattern matching, these summaries are capable of summarizing entities in terms of their neighborhood similarity up to a certain hop, using small and informative graph patterns. (2) We study a diversified graph summarization problem. Given a knowledge graph, it is to discover top-k summaries that maximize a bi-criteria function, characterized by both informativeness and diversity. We show that diversified summarization is feasible for large graphs, by developing both sequential and parallel summarization algorithms. (a) We show that there exists a 2-approximation algorithm to discover diversified summaries. We further develop an anytime sequential algorithm which discovers summaries under resource constraints. (b) We present a new parallel algorithm with quality guarantees. The algorithm is parallel scalable, which ensures its feasibility in distributed graphs. (3) We also develop a summary-based query evaluation scheme, which only refers to a small number of summaries. Using real-world knowledge graphs, we experimentally verify the effectiveness and efficiency of our summarization algorithms, and query processing using summaries.

Collecting structured knowledge for real-world entities has become a critical task for many applications. A big gap between the knowledge in existing knowledge repositories and the knowledge in the real world is the knowledge on tail verticals (i.e., less popular domains). Such knowledge, though not necessarily globally popular, can be personal hobbies to many people and thus collectively impactful. This paper studies the problem of knowledge verification for tail verticals ; that is, deciding the correctness of a given triple. Through comprehensive experimental study we answer the following questions. 1) Can we find evidence for tail knowledge from an extensive set of sources, including knowledge bases, the web, and query logs? 2) Can we judge correctness of the triples based on the collected evidence? 3) How can we further improve knowledge verification on tail verticals? Our empirical study suggests a new knowledge-verification framework, which we call F acty , that applies various kinds of evidence collection techniques followed by knowledge fusion. F acty can verify 50% of the (correct) tail knowledge with a precision of 84%, and it significantly outperforms state-of-the-art methods. Detailed error analysis on the obtained results suggests future research directions.

Mining and searching heterogeneous and large knowledge graphs is challenging under real-world resource constraints such as response time. This paper studies a framework that discover to facilitate knowledge graph search. 1) We introduce a class of summaries characterized by graph patterns. In contrast to conventional summaries defined by frequent subgraphs, the summaries are capable of adaptively summarize entities with similar neighbors up to a bounded hop. 2) We formulate the computation of graph summarization as a bi-criteria pattern mining problem. Given a knowledge graph G, the problem is to discover k diversified summaries that maximizes the informativeness measure. Although this problem is NP-hard, we show that it is 2-approximable. We also introduce an online mining algorithm that trade-off speed and accuracy, under given resource constraints. 3) We develop query evaluation algorithms that make use of the summaries as views. These algorithms efficiently compute (approximate) answers with high accuracy, and only refer to a small number of summaries. Our experimental study verifies that online mining over large knowledge graphs is feasible, and can suggest bounded search in knowledge graphs.

Recently there has been a rapid increase in the number of data sources and data services, such as cloud-based data markets and data portals, that facilitate the collection, publishing and trading of data. Data sources typically exhibit large heterogeneity in the type and quality of data they provide. Unfortunately, when the number of data sources is large, it is difficult for users to reason about the actual usefulness of sources for their applications and the trade-offs between the benefits and costs of acquiring and integrating sources. In this demonstration we present \textsc{SourceSight}, a system that allows users to interactively explore a large number of heterogeneous data sources, and discover valuable sets of sources for diverse integration tasks. \textsc{SourceSight}~uses a novel multi-level source quality index that enables effective source selection at different granularity levels, and introduces a collection of new techniques to discover and evaluate relevant sources for integration.

With the mission "leave no valuable data behind", we developed techniques for knowledge fusion to guarantee the correctness of the knowledge. This talk starts with describing a few crazy ideas we have tested. The first, known as "Knowledge Vault", used 15 extractors to automatically extract knowledge from 1B+ Webpages, obtaining 3B+ distinct (subject, predicate, object) knowledge triples and predicting well-calibrated probabilities for extracted triples. The second, known as "Knowledge-Based Trust", estimated the trustworthiness of 119M webpages and 5.6M websites based on the correctness of their factual information. We then present how we bring the ideas to business in filling the gap between the knowledge at Google Knowledge Graph and the knowledge in the world.

Recent research shows that copying is prevalent for Deep-Web data and considering copying can significantly improve truth finding from conflicting values. However, existing copy detection techniques do not scale for large sizes and numbers of data sources, so truth finding can be slowed down by one to two orders of magnitude compared with the corresponding techniques that do not consider copying. In this paper, we study how to improve scalability of copy detection on structured data. Our algorithm builds an inverted index for each shared value and processes the index entries in decreasing order of how much the shared value can contribute to the conclusion of copying. We show how we use the index to prune the data items we consider for each pair of sources, and to incrementally refine our results in iterative copy detection. We also apply a sampling strategy with which we are able to further reduce copy-detection time while still obtaining very similar results as on the whole data set. Experiments on various real data sets show that our algorithm can reduce the time for copy detection by two to three orders of magnitude; in other words, truth finding can benefit from copy detection with very little overhead.

We present a method called TIMEMACHINE to generate a timeline of events and relations for entities in a knowledge base. For example for an actor, such a timeline should show the most important professional and personal milestones and relationships such as works, awards, collaborations, and family relationships. We develop three orthogonal timeline quality criteria that an ideal timeline should satisfy: (1) it shows events that are relevant to the entity; (2) it shows events that are temporally diverse, so they distribute along the time axis, avoiding visual crowding and allowing for easy user interaction, such as zooming in and out; and (3) it shows events that are content diverse, so they contain many different types of events (e.g., for an actor, it should show movies and marriages and awards, not just movies). We present an algorithm to generate such timelines for a given time period and screen size, based on submodular optimization and web-co-occurrence statistics with provable performance guarantees. A series of user studies using Mechanical Turk shows that all three quality criteria are crucial to produce quality timelines and that our algorithm significantly outperforms various baseline and state-of-the-art methods.

Large-scale knowledge repositories are becoming increasingly important as a foundation for enabling a wide variety of complex applications. In turn, building high-quality knowledge repositories critically depends on the technologies of knowledge curation and knowledge fusion, which share many similar goals with data integration, while facing even more challenges in extracting knowledge from both structured and unstructured data, across a large variety of domains, and in multiple languages.

A lot of systems and applications are data-driven, and the correctness of their operation relies heavily on the correctness of their data. While existing data cleaning techniques can be quite effective at purging datasets of errors, they disregard the fact that a lot of errors are systematic, inherent to the process that produces the data, and thus will keep occurring unless the problem is corrected at its source. In contrast to traditional data cleaning, in this paper we focus on data diagnosis: explaining where and how the errors happen in a data generative process.

International audience

We study the problem of group linkage: linking records that refer to multiple entities in the same group. Applications for group linkage include finding businesses in the same chain, finding social network users from the same organization, and so on. Group linkage faces new challenges compared to traditional entity resolution. First, although different members in the same group can share some similar global values of an attribute, they represent different entities so can also have distinct local values for the same or different attributes, requiring a high tolerance for value diversity. Second, we need to be able to distinguish local values from erroneous values.

The quality of web sources has been traditionally evaluated using exogenous signals such as the hyperlink structure of the graph. We propose a new approach that relies on endogenous signals, namely, the correctness of factual information provided by the source. A source that has few false facts is considered to be trustworthy. The facts are automatically extracted from each source by information extraction methods commonly used to construct knowledge bases. We propose a way to distinguish errors made in the extraction process from factual errors in the web source per se, by using joint inference in a novel multi-layer probabilistic model. We call the trustworthiness score we computed Knowledge-Based Trust (KBT) . On synthetic data, we show that our method can reliably compute the true trustworthiness levels of the sources. We then apply it to a database of 2.8B facts extracted from the web, and thereby estimate the trustworthiness of 119M webpages. Manual evaluation of a subset of the results confirms the effectiveness of the method.

With the abundant availability of information one can mine from the Web today, there is increasing interest to develop a complete understanding of the history of an entity (i.e., a person, a company, a music genre, a country, etc.) (see, for example, [7, 9, 10, 11]) and to depict trends over time [5, 12, 13]. This, however, remains a largely difficult and manual task despite more than a couple of decades of research in the areas of temporal databases and data integration.

Keys for graphs aim to uniquely identify entities represented by vertices in a graph. We propose a class of keys that are recursively defined in terms of graph patterns, and are interpreted with subgraph isomorphism. Extending conventional keys for relations and XML, these keys find applications in object identification, knowledge fusion and social network reconciliation. As an application, we study the entity matching problem that, given a graph G and a set Σ of keys, is to find all pairs of entities (vertices) in G that are identified by keys in Σ. We show that the problem is intractable, and cannot be parallelized in logarithmic rounds. Nonetheless, we provide two parallel scalable algorithms for entity matching, in MapReduce and a vertex-centric asynchronous model. Using real-life and synthetic data, we experimentally verify the effectiveness and scalability of the algorithms.

The web is a rich resource of structured data. There has been an increasing interest in using web structured data for many applications such as data integration, web search and question answering. In this paper, we present Dexter, a system to find product sites on the web, and detect and extract product specifications from them. Since product specifications exist in multiple product sites, our focused crawler relies on search queries and backlinks to discover product sites. To perform the detection, and handle the high diversity of specifications in terms of content, size and format, our system uses supervised learning to classify HTML fragments (e.g., tables and lists) present in web pages as specifications or not. To perform large-scale extraction of the attribute-value pairs from the HTML fragments identified by the specification detector, D exter adopts two lightweight strategies: a domain-independent and unsupervised wrapper method, which relies on the observation that these HTML fragments have very similar structure; and a combination of this strategy with a previous approach, which infers extraction patterns by annotations generated by automatic but noisy annotators. The results show that our crawler strategy to locate product specification pages is effective: (1) it discovered 1:46A M product specification pages from 3; 005 sites and 9 different categories; (2) the specification detector obtains high values of F-measure (close to 0:9) over a heterogeneous set of product specifications; and (3) our efficient wrapper methods for attribute-value extraction get very high values of precision (0.92) and recall (0.95) and obtain better results than a state-of-the-art, supervised rule-based wrapper.

The problem of identifying and repairing data errors has been an area of persistent focus in data management research. However, while traditional data cleaning techniques can be effective at identifying several data discrepancies, they disregard the fact that many errors aresystematic, inherent to the process that produces the data, and thus will keep occurring unless the root cause is identified and corrected. In this demonstration, we will present a large-scale diagnostic framework called DataXRay. Like a medical X-ray that aids the diagnosis of medical conditions by revealing problems underneath the surface, DataXRayreveals hidden connections and common properties among data errors. Thus, in contrast to traditional cleaning methods, which treat the symptoms, our system investigates the underlying conditions that cause the errors. The core of DataXRaycombines an intuitive and principled cost model derived by Bayesian analysis, and an efficient, highly-parallelizable diagnostic algorithm that discovers common properties among erroneous data elements in a top-down fashion. Our system has a simple interface that allows users to load different datasets, to interactively adjust key diagnostic parameters, to explore the derived diagnoses, and to compare with solutions produced by alternative algorithms. Through this demonstration, participants will understand (1) the characteristics of good diagnoses, (2) how and why errors occur in real-world datasets, and (3) the distinctions with other related problems and approaches.

Recent years have witnessed a proliferation of large-scale knowledge bases, including Wikipedia, Freebase, YAGO, Microsoft's Satori, and Google's Knowledge Graph. To increase the scale even further, we need to explore automatic methods for constructing knowledge bases. Previous approaches have primarily focused on text-based extraction, which can be very noisy. Here we introduce Knowledge Vault, a Web-scale probabilistic knowledge base that combines extractions from Web content (obtained via analysis of text, tabular data, page structure, and human annotations) with prior knowledge derived from existing knowledge repositories. We employ supervised machine learning methods for fusing these distinct information sources. The Knowledge Vault is substantially bigger than any previously published structured knowledge repository, and features a probabilistic inference system that computes calibrated probabilities of fact correctness. We report the results of multiple studies that explore the relative utility of the different information sources and extraction methods.

Many applications rely on Web data and extraction systems to accomplish knowledge-driven tasks. Web information is not curated, so many sources provide inaccurate, or conflicting information. Moreover, extraction systems introduce additional noise to the data. We wish to automatically distinguish correct data and erroneous data for creating a cleaner set of integrated data. Previous work has shown that a naive voting strategy that trusts data provided by the majority or at least a certain number of sources may not work well in the presence of copying between the sources. However, correlation between sources can be much broader than copying: sources may provide data from complementary domains (negative correlation), extractors may focus on different types of information (negative correlation), and extractors may apply common rules in extraction (positive correlation, without copying). In this paper we present novel techniques modeling correlations between sources and applying it in truth finding. We provide a comprehensive evaluation of our approach on three real-world datasets with different characteristics, as well as on synthetic data, showing that our algorithms outperform the existing state-of-the-art techniques.

Data integration is a challenging task due to the large numbers of autonomous data sources. This necessitates the development of techniques to reason about the benefits and costs of acquiring and integrating data. Recently the problem of source selection (i.e., identifying the subset of sources that maximizes the profit from integration) was introduced as a preprocessing step before the actual integration. The problem was studied for static sources and used the accuracy of data fusion to quantify the integration profit.

The task of data fusion is to identify the true values of data items ( e.g. , the true date of birth for Tom Cruise ) among multiple observed values drawn from different sources ( e.g. , Web sites) of varying (and unknown) reliability. A recent survey [20] has provided a detailed comparison of various fusion methods on Deep Web data. In this paper, we study the applicability and limitations of different fusion techniques on a more challenging problem: knowledge fusion . Knowledge fusion identifies true subject-predicate-object triples extracted by multiple information extractors from multiple information sources. These extractors perform the tasks of entity linkage and schema alignment, thus introducing an additional source of noise that is quite different from that traditionally considered in the data fusion literature, which only focuses on factual errors in the original sources. We adapt state-of-the-art data fusion techniques and apply them to a knowledge base with 1.6B unique knowledge triples extracted by 12 extractors from over 1B Web pages, which is three orders of magnitude larger than the data sets used in previous data fusion papers. We show great promise of the data fusion approaches in solving the knowledge fusion problem, and suggest interesting research directions through a detailed error analysis of the methods.

Record linkage clusters records such that each cluster corresponds to a single distinct real-world entity. It is a crucial step in data cleaning and data integration. In the big data era, the velocity of data updates is often high, quickly making previous linkage results obsolete. This paper presents an end-to-end framework that can incrementally and efficiently update linkage results when data updates arrive. Our algorithms not only allow merging records in the updates with existing clusters, but also allow leveraging new evidence from the updates to fix previous linkage errors. Experimental results on three real and synthetic data sets show that our algorithms can significantly reduce linkage time without sacrificing linkage quality.

No abstract available.

Despite the abundance of useful information on the Web, different Web sources often provide conflicting data, some being out-of-date, inaccurate, or erroneous. Data fusion aims at resolving conflicts and finding the truth. Advanced fusion techniques apply iterative MAP (Maximum A Posteriori) analysis that reasons about trustworthiness of sources and copying relationships between them. Providing explanations for such decisions is important for a better understanding, but can be extremely challenging because of the complexity of the analysis during decision making.

Data integration systems offer a uniform interface for querying a large number of autonomous and heterogeneous data sources. Ideally, answers are returned as sources are queried and the answer list is updated as more answers arrive. Choosing a good ordering in which the sources are queried is critical for increasing the rate at which answers are returned. However, this problem is challenging since we often do not have complete or precise statistics of the sources, such as their coverage and overlap. It is further exacerbated in the Big Data era, which is witnessing two trends in Deep-Web data: first, obtaining a full coverage of data in a particular domain often requires extracting data from thousands of sources; second, there is often a big variation in overlap between different data sources. In this paper we present OASIS , an O nline query A nswering S ystem for overlapp I ng Sources. OASIS has three key components for source ordering. First, the Overlap Estimation component estimates overlaps between sources according to available statistics under the Maximum Entropy principle. Second, the Source Ordering component orders the sources according to the new contribution they are expected to provide, and adjusts the ordering based on statistics collected during query answering. Third, the Statistics Enrichment component selects critical missing statistics to enrich at runtime. Experimental results on both real and synthetic data show high efficiency and scalability of our algorithm.

The Web has enabled the availability of a vast amount of useful information in recent years. However, the web technologies that have enabled sources to share their information have also made it easy for sources to copy from each other and often publish without proper attribution. Understanding the copying relationships between sources has many benefits, including helping data providers protect their own rights, improving various aspects of data integration, and facilitating in-depth analysis of information flow. The importance of copy detection has led to a substantial amount of research in many disciplines of Computer Science, based on the type of information considered, such as text, images, videos, software code, and structured data. This seminar explores the similarities and differences between the techniques proposed for copy detection across the different types of information. We also examine the computational challenges associated with large-scale copy detection, indicating how they could be detected efficiently, and identify a range of open problems for the community.

We are often thrilled by the abundance of information surrounding us and wish to integrate data from as many sources as possible. However, understanding, analyzing, and using these data are often hard. Too much data can introduce a huge integration cost, such as expenses for purchasing data and resources for integration and cleaning. Furthermore, including low-quality data can even deteriorate the quality of integration results instead of bringing the desired quality gain. Thus, "the more the better" does not always hold for data integration and often "less is more". In this paper, we study how to select a subset of sources before integration such that we can balance the quality of integrated data and integration cost. Inspired by the Marginalism principle in economic theory, we wish to integrate a new source only if its marginal gain, often a function of improved integration quality, is higher than the marginal cost, associated with data-purchase expense and integration resources. As a first step towards this goal, we focus on data fusion tasks, where the goal is to resolve conflicts from different sources. We propose a randomized solution for selecting sources for fusion and show empirically its effectiveness and scalability on both real-world data and synthetic data.

The amount of useful information available on the Web has been growing at a dramatic pace in recent years and people rely more and more on the Web to fulfill their information needs. In this paper, we study truthfulness of Deep Web data in two domains where we believed data are fairly clean and data quality is important to people's lives: Stock and Flight. To our surprise, we observed a large amount of inconsistency on data from different sources and also some sources with quite low accuracy. We further applied on these two data sets state-of-the-art data fusion methods that aim at resolving conflicts and finding the truth, analyzed their strengths and limitations, and suggested promising research directions. We wish our study can increase awareness of the seriousness of conflicting data on the Web and in turn inspire more research in our community to tackle this problem.

Many data sets containtemporal recordsover a long period of time; each record is associated with a time stamp and describes some aspects of a real-world entity at that particular time. From such data, users often wish to search for entities in a particular period and understand the history of one entity or all entities in the data set. A major challenge for enabling such search and exploration is to identify records that describe the same real-world entity over a long period of time; however, linking temporal records is hard given that the values that describe an entity can evolve over time (e.g., a person can move from one affiliation to another). We demonstrate the Chronos system which offers users the useful tool for finding real-world entities over time and understanding history of entities in the bibliography domain. The core of Chronos is a temporal record-linkage algorithm, which is tolerant to value evolution over time. Our algorithm can obtain an F-measure of over 0.9 in linking author records and fix errors made byDBLP. We show how Chronos allows users to explore the history of authors, and how it helps users understand our linkage results by comparing our results with those of existing systems, highlighting differences in the results, explaining our decisions to users, and answering "what-if" questions.

Data integration systems offer users a uniform interface to a set of data sources. Previous work has typically assumed that the data sources are independent of each other; however, in scenarios involving large numbers of sources, such as the Web or large enterprises, there is an eco-system of dependent sources, where some sources copy parts of their data from others.

Correctness of data residing in a database is vital. While integrity constraint enforcement can often ensure data consistency, it is inadequate to protect against updates that involve careless, unintentional errors, e.g., whether a specified update to an employee's record was for the intended employee. We propose a novel approach that is complementary to existing integrity enforcement techniques, to guard against such erroneous updates.

The Web has enabled the availability of a vast amount of useful information in recent years. However, the web technologies that have enabled sources to share their information have also made it easy for sources to copy from each other and often publish without proper attribution. Understanding the copying relationships between sources has many benefits, including helping data providers protect their own rights, improving various aspects of data integration, and facilitating in-depth analysis of information flow.

Web technologies have enabled data sharing between sources but also simplified copying (and often publishing without proper attribution). The copying relationships can be complex: some sources copy from multiple sources on different subsets of data; some co-copy from the same source, and some transitively copy from another. Understanding such copying relationships is desirable both for business purposes and for improving many key components in data integration, such as resolving conflicts across various sources, reconciling distinct references to the same real-world entity, and efficiently answering queries over multiple sources. Recent works have studied how to detect copying between a pair of sources, but the techniques can fall short in the presence of complex copying relationships. In this paper we describe techniques that discover global copying relationships between a set of structured sources. Towards this goal we make two contributions. First, we propose a global detection algorithm that identifies co-copying and transitive copying, returning only source pairs with direct copying. Second, global detection requires accurate decisions on copying direction; we significantly improve over previous techniques on this by considering various types of evidence for copying and correlation of copying on different data items. Experimental results on real-world data and synthetic data show high effectiveness and efficiency of our techniques.

We live in the Information Era, with access to a huge amount of information from a variety of data sources. However, data sources are of different qualities, often providing conflicting, out-of-date and incomplete data. Data sources can also easily copy, reformat and modify data from other sources, propagating erroneous data. These issues make the identification of high quality information and sources non-trivial. We demonstrate the Solomon system, whose core is a module that detects copying between sources. We demonstrate that we can effectively detect copying relationship between data sources, leverage the results in truth discovery, and provide a user-friendly interface to facilitate users in identifying sources that best suit their information needs.

Many data-management applications require integrating data from a variety of sources, where different sources may refer to the same real-world entity in different ways and some may even provide erroneous data. An important task in this process is to recognize and merge the various references that refer to the same entity. In practice, some attributes satisfy a uniqueness constraint---each real-world entity (or most entities) has a unique value for the attribute ( e.g. , business contact phone, address, and email). Traditional techniques tackle this case by first linking records that are likely to refer to the same real-world entity, and then fusing the linked records and resolving conflicts if any. Such methods can fall short for three reasons: first, erroneous values from sources may prevent correct linking; second, the real world may contain exceptions to the uniqueness constraints and always enforcing uniqueness can miss correct values; third, locally resolving conflicts for linked records may overlook important global evidence. This paper proposes a novel technique to solve this problem. The key component of our solution is to reduce the problem into a k -partite graph clustering problem and consider in clustering both similarity of attribute values and the sources that associate a pair of values in the same record. Thus, we perform global linkage and fusion simultaneously, and can identify incorrect values and differentiate them from alternative representations of the correct value from the beginning. In addition, we extend our algorithm to be tolerant to a few violations of the uniqueness constraints. Experimental results show accuracy and scalability of our technique.

Many data management applications, such as setting up Web portals, managing enterprise data, managing community data, and sharing scientific data, require integrating data from multiple sources. Each of these sources provides a set of values and different sources can often provide conflicting values. To present quality data to users, it is critical that data integration systems can resolve conflicts and discover true values. Typically, we expect a true value to be provided by more sources than any particular false one, so we can take the value provided by the majority of the sources as the truth. Unfortunately, a false value can be spread through copying and that makes truth discovery extremely tricky. In this paper, we consider how to find true values from conflicting information when there are a large number of sources, among which some may copy from others. We present a novel approach that considers dependence between data sources in truth discovery. Intuitively, if two data sources provide a large number of common values and many of these values are rarely provided by other sources ( e.g. , particular false values), it is very likely that one copies from the other. We apply Bayesian analysis to decide dependence between sources and design an algorithm that iteratively detects dependence and discovers truth from conflicting information. We also extend our model by considering accuracy of data sources and similarity between values. Our experiments on synthetic data as well as real-world data show that our algorithm can significantly improve accuracy of truth discovery and is scalable when there are a large number of data sources.

Modern information management applications often require integrating data from a variety of data sources, some of which may copy or buy data from other sources. When these data sources model a dynamically changing world ( e.g. , people's contact information changes over time, restaurants open and go out of business), sources often provide out-of-date data. Errors can also creep into data when sources are updated often. Given out-of-date and erroneous data provided by different, possibly dependent, sources, it is challenging for data integration systems to provide the true values. Straightforward ways to resolve such inconsistencies ( e.g. , voting) may lead to noisy results, often with detrimental consequences. In this paper, we study the problem of finding true values and determining the copying relationship between sources, when the update history of the sources is known. We model the quality of sources over time by their coverage, exactness and freshness . Based on these measures, we conduct a probabilistic analysis. First, we develop a Hidden Markov Model that decides whether a source is a copier of another source and identifies the specific moments at which it copies. Second, we develop a Bayesian model that aggregates information from the sources to decide the true value for a data item, and the evolution of the true values over time. Experimental results on both real-world and synthetic data show high accuracy and scalability of our techniques.

The amount of information produced in the world increases by 30% every year and this rate will only go up. With advanced network technology, more and more sources are available either over the Internet or in enterprise intranets. Modern data management applications, such as setting up Web portals, managing enterprise data, managing community data, and sharing scientific data, often require integrating available data sources and providing a uniform interface for users to access data from different sources; such requirements have been driving fruitful research on data integration over the last two decades [11, 13].

Data integration systems offer a uniform interface to a set of data sources. Despite recent progress, setting up and maintaining a data integration application still requires significant upfront effort of creating a mediated schema and semantic mappings from the data sources to the mediated schema. Many application contexts involving multiple data sources (e.g., the web, personal information management, enterprise intranets) do not require full integration in order to provide useful services, motivating a pay-as-you-go approach to integration. With that approach, a system starts with very few (or inaccurate) semantic mappings and these mappings are improved over time as deemed necessary.

Dataspaces are collections of heterogeneous and partially unstructured data. Unlike data-integration systems that also offer uniform access to heterogeneous data sources, dataspaces do not assume that all the semantic relationships between sources are known and specified. Much of the user interaction with dataspaces involves exploring the data, and users do not have a single schema to which they can pose queries. Consequently, it is important that queries are allowed to specify varying degrees of structure, spanning keyword queries to more structure-aware queries.

The explosion of information available in digital form has made search a hot research topic for the Information Management Community. While most of the research on search is focused on the WWW, individual computer users have developed their own vast collections of data on their desktops, and these collections are in critical need for good search and query tools. The problem is exacerbated by the proliferation of varied electronic devices (laptops, PDAs, cellphones) that are at our disposal, which often hold subsets or variations of our data. In fact, several recent venues have noted Personal Information Management (PIM) as an area of growing interest to the data management community [1, 8, 6]

Reference reconciliation is the problem of identifying when different references (i.e., sets of attribute values) in a dataset correspond to the same real-world entity. Most previous literature assumed references to a single class that had a fair number of attributes (e.g., research publications). We consider complex information spaces: our references belong to multiple related classes and each reference may have very few attribute values. A prime example of such a space is Personal Information Management, where the goal is to provide a coherent view of all the information on one's desktop.Our reconciliation algorithm has three principal features. First, we exploit the associations between references to design new methods for reference comparison. Second, we propagate information between reconciliation decisions to accumulate positive and negative evidences. Third, we gradually enrich references by merging attribute values. Our experiments show that (1) we considerably improve precision and recall over standard methods on a diverse set of personal information datasets, and (2) there are advantages to using our algorithm even on a standard citation dataset benchmark.

No abstract available.