Multimodal Deep Learning enhances decision-making by integrating diverse information sources, such as texts, images, audio, and videos. To develop trustworthy multimodal approaches, it is essential to understand how uncertainty impacts these models. We propose LUMA, a unique multimodal dataset, featuring audio, image, and textual data from 50 classes, specifically designed for learning from uncertain data. It extends the well-known CIFAR 10/100 dataset with audio samples extracted from three audio corpora, and text data generated using the Gemma-7B Large Language Model (LLM). The LUMA dataset enables the controlled injection of varying types and degrees of uncertainty to achieve and tailor specific experiments and benchmarking initiatives. LUMA is also available as a Python package including the functions for generating multiple variants of the dataset with controlling the diversity of the data, the amount of noise for each modality, and adding out-of-distribution samples. A baseline pre-trained model is also provided alongside three uncertainty quantification methods: Monte-Carlo Dropout, Deep Ensemble, and Reliable Conflictive Multi-View Learning. This comprehensive dataset and its tools are intended to promote and support the development, evaluation, and benchmarking of trustworthy and robust multimodal deep learning approaches. We anticipate that the LUMA dataset will help the research community to design more trustworthy and robust machine learning approaches for safety critical applications. The code and instructions for downloading and processing the dataset can be found at: https://github.com/bezirganyan/LUMA.

Time series anomaly detection is a vital task in many domains, including patient monitoring in healthcare, forecasting in finance, and predictive maintenance in energy industries. This has led to a proliferation of anomaly detection methods, including deep learning-based methods. Benchmarks are essential for comparing the performances of these models as they emerge, in a fair, rigorous, and reproducible approach. Although several benchmarks for comparing models have been proposed, these usually rely on a one-time execution over a limited set of datasets, with comparisons restricted to a few models. We propose OrionBench– an end-user centric, continuously maintained benchmarking framework for unsupervised time series anomaly detection models. Our framework provides universal abstractions to represent models, extensibility to add new pipelines and datasets, hyperparameter standardization, pipeline verification, and frequent releases with published updates of the benchmark. We demonstrate how to use OrionBench, and the performance of pipelines across 17 releases published over the course of four years. We also walk through two real scenarios we experienced with OrionBench that highlight the importance of continuous benchmarking for unsupervised time series anomaly detection.

Choosing a suitable deep learning architecture for multimodal data fusion is a challenging task, as it requires the effective integration and processing of diverse data types, each with distinct structures and characteristics. In this paper, we introduce MixMAS, a novel framework for sampling-based mixer architecture search tailored to multimodal learning. Our approach automatically selects the optimal MLP-based architecture for a given multimodal machine learning (MML) task. Specifically, MixMAS utilizes a sampling-based micro-benchmarking strategy to explore various combinations of modality-specific encoders, fusion functions, and fusion networks, systematically identifying the architecture that best meets the task’s performance metrics.

Explanations of machine learning (ML) model predictions generated by Explainable AI (XAI) techniques such as SHAP are essential for people using ML outputs for decision-making. We explore the potential of Large Language Models (LLMs) to transform these explanations into human-readable, narrative formats that align with natural communication. We address two key research questions: (1) Can LLMs reliably transform traditional explanations into high-quality narratives? and (2) How can we effectively evaluate the quality of narrative explanations? To answer these questions, we introduce Explingo, which consists of two LLM-based subsystems, a Narrator and Grader. The Narrator takes in ML explanations and transforms them into natural-language descriptions. The Grader scores these narratives on a set of metrics including accuracy, completeness, fluency, and conciseness.Our experiments demonstrate that LLMs can generate high-quality narratives that achieve high scores across all metrics, particularly when guided by a small number of human-labeled and bootstrapped examples. We also identified areas that remain challenging, in particular for effectively scoring narratives in complex domains. The findings from this work have been integrated into an open-source tool that makes narrative explanations available for further applications.

The flexible nature of large language models allows them to be used for diverse applications. Recent studies have showcased numerous abilities of these models, including performing time series forecasting. In this paper, we present a novel study of large language models used for the challenging task of time series anomaly detection. This problem entails two novel aspects for LLMs specifically: first, the model needs to be able to identify part of an input sequence (or multiple parts) as anomalous; and second, the model needs to work with time series data rather than with text input. We introduce SIGLLM, a framework for time series anomaly detection using large language models. Our framework includes a time-series-to-text conversion module, as well as end-to-end pipelines that prompt language models to perform time series anomaly detection. We investigate two paradigms for testing the abilities of large language models to perform the detection task. First, we present a prompt-based detection method that directly asks a language model to indicate which elements of the input are anomalies. Second, we leverage the forecasting capability of a large language model to guide the anomaly detection process. We evaluated our framework on 11 datasets spanning various sources and 10 pipelines. We show that the forecasting method significantly outperformed the prompting method in all 11 datasets with respect to the F1 score. Moreover, while large language models are capable of finding anomalies, state-of-the-art deep learning models are still superior in performance, achieving 30% improvement.

In this paper, we propose M2-Mixer, an MLP-Mixer based architecture with multi-head loss for multimodal classification. It achieves better performances than the convolutional, recurrent, or neural architecture search based baseline models with the main advantage of conceptual and computational simplicity. The proposed multi-head loss function addresses the problem of modality predominance (i.e., when one of the modalities is favored over the others by the training algorithm). Our experiments demonstrate that our multimodal mixer architecture, combined with the multi-head loss function, outperforms the baseline models on two benchmark multimodal datasets: AVMNIST and MIMIC-III with respectively, on average, + 0.43% in accuracy and 6. 4 times reduction in training time and + 0.33% in accuracy and 13. 3 times reduction in training time, compared with previous best performing models.

Anomaly detection on time series data is increasingly common across various industrial domains that monitor metrics in order to prevent potential accidents and economic losses. However, a scarcity of labeled data and ambiguous definitions of anomalies can complicate these efforts. Recent unsupervised machine learning methods have made remarkable progress in tackling this problem using either single-timestamp predictions or time series reconstructions. While traditionally considered separately, these methods are not mutually exclusive and can offer complementary perspectives on anomaly detection. This paper first highlights the successes and limitations of prediction-based and reconstruction-based methods with visualized time series signals and anomaly scores. We then propose AER (Auto-encoder with Regression), a joint model that combines a vanilla auto-encoder and an LSTM regressor to incorporate the successes and address the limitations of each method. Our model can produce bi-directional predictions while simultaneously reconstructing the original time series by optimizing a joint objective function. Furthermore, we propose several ways of combining the prediction and reconstruction errors through a series of ablation studies. Finally, we compare the performance of the AER architecture against two prediction-based methods and three reconstruction-based methods on 12 well-known univariate time series datasets from NASA, Yahoo, Numenta, and UCR. The results show that AER has the highest averaged F1 score across all datasets (a 23.5% improvement compared to ARIMA) while retaining a runtime similar to its vanilla auto-encoder and regressor components. Our model is available in Orion1, an opensource benchmarking tool for time series anomaly detection.

The automatic discovery of functional dependencies (FDs) has been widely studied as one of the hardest problems in data profiling. Existing approaches have focused on making the FD computation efficient while inspecting single relations at a time. In this paper, for the first time we address the problem of inferring FDs for multiple relations as they occur in integrated views by solely using the functional dependencies of the base relations of the view itself. To this purpose, we leverage logical inference and selective mining and show that we can discover most of the exact FDs from the base relations and avoid the full computation of the FDs for the integrated view itself, while at the same time preserving the lineage of FDs of base relations. We propose algorithms to speedup the inferred FD discovery process and mine FDs on-the-fly only from necessary data partitions. We present InFine (INferred FunctIoNal dEpendency), an end-to-end solution to discover inferred FDs on integrated views by leveraging provenance information of base relations. Our experiments on a range of real-world and synthetic datasets demonstrate the benefits of our method over existing FD discovery methods that need to rerun the discovery process on the view from scratch and cannot exploit lineage information on the FDs. We show that InFine outperforms traditional methods necessitating the full integrated view computation by one to two order of magnitude in terms of runtime. It is also the most memory efficient method while preserving FD provenance information using mainly inference from base table with negligible execution time.

It is widely accepted that data preparation is one of the most time-consuming steps of the machine learning (ML) lifecycle. It is also one of the most important steps, as the quality of data directly influences the quality of a model. In this tutorial, we will discuss the importance and the role of exploratory data analysis (EDA) and data visualisation techniques to find data quality issues and for data preparation, relevant to building ML pipelines. We will also discuss the latest advances in these fields and bring out areas that need innovation. To make the tutorial actionable for practitioners, we will also discuss the most popular open-source packages that one can get started with along with their strengths and weaknesses. Finally, we will discuss on the challenges posed by industry workloads and the gaps to be addressed to make data-centric AI real in industry settings.

The detection of anomalies in time series data is a critical task with many monitoring applications. Existing systems often fail to encompass an end-to-end detection process, to facilitate comparative analysis of various anomaly detection methods, or to incorporate human knowledge to refine output. This precludes current methods from being used in real-world settings by practitioners who are not ML experts. In this paper, we introduce Sintel, a machine learning framework for end-to-end time series tasks such as anomaly detection. The framework uses state-of-the-art approaches to support all steps of the anomaly detection process. Sintel logs the entire anomaly detection journey, providing detailed documentation of anomalies over time. It enables users to analyze signals, compare methods, and investigate anomalies through an interactive visualization tool, where they can annotate, modify, create, and remove events. Using these annotations, the framework leverages human knowledge to improve the anomaly detection pipeline. We demonstrate the usability, efficiency, and effectiveness of Sintel through a series of experiments on three public time series datasets, and through a real-world use case with spacecraft experts. Sintel's framework, code, and datasets are open-sourced at https://github.com/sintel-dev/

We demonstrate DORA THE EXPLORER, a system that guides users in finding items of interest in a very large data set. DORA THE EXPLORER provides users with the full spectrum of exploration modes and is driven by Data Familiarity or Curiosity, as well as User Interventions. DORA THE EXPLORER is able to handle data and search scenario complexity, i.e., the difficulty to find scattered/clustered individual records in the data set, and user ability to express what s/he needs. DORA THE EXPLORER relies on Deep Reinforcement Learning that combines intrinsic (curiosity) and extrinsic (familiarity) rewards. DORA's main goal is to support scientific discovery from data. We describe the system architecture and illustrate it with three demonstration scenarios on a 2.6 mil-lion galaxies SDSS, a large sky survey data set1. A video of DORA THE EXPLORER is available at https://bit.ly/dora-demo, the codehttps://github.com/apersonnaz/rl-guided-galaxy-exploration, and the application at https://bit.ly/dora-application

We develop S4, a Statistically-Sound Segment Search framework that combines principled data partitioning and sound statistical testing to verify common hypotheses in retail data and return interpretable customer data segments. Our framework accommodates one-sample, two-sample, and multiple-sample testing, to provide various aggregations and comparisons of customer transactions. To control the proportion of false discoveries in multiple hypothesis testing, we enforce an FDR-controlling procedure and formulate a unified optimization problem that returns customer data segments that satisfy the test for a given significance level, maximize coverage of the input data, and are within a risk capital. We develop a greedy algorithm to explore different data partitions and test multiple hypotheses in a sound manner. Our extensive experiments on four retail data sets examine the interaction between significance, risk and coverage, and demonstrate the expressivity, usefulness, and scalability of S4 in practice.

We demonstrate QeNoBi, a system for mining and querying customer behavioral patterns. QeNoBi combines an interactive visual interface, on-demand mining, and efficient topk processing, to provide the exploration of customer behavior over time. QeNoBi relies on two distinct data models: a customercentric graph that represents customers with similar purchasing behaviors and is annotated with a change algebra to reflect their behavior evolution, and product-centric time series that reflect the evolution of customer purchases over time. Users can query both representations along three dimensions: shape (the sketched trend of the behavior), scope (the set of customers/products of interest), and time granularity. QeNoBi provides a holistic behavior exploration capability by allowing users to seamlessly switch between customer-centric and product-centric views in a coordinated manner, thereby catering to various needs. A demonstration of QeNoBi is available at https://bit.ly/2HlcO3S.

Artificial Intelligence and machine learning techniques can offer powerful tools for addressing the greatest challenges facing humanity and helping society adapt to a rapidly changing climate, respond to disasters and pandemic crisis, and reach the United Nations (UN) Sustainable Development Goals (SDGs) by 2030. In recent approaches for mitigation and adaptation, data analytics and ML are only one part of the solution that requires interdisciplinary and methodological research and innovations. For example, challenges include multi-modal and multi-source data fusion to combine satellite imagery with other relevant data, handling noisy and missing ground data at various spatio-temporal scales, and ensembling multiple physical and ML models to improve prediction accuracy. Despite recognized successes, there are many areas where ML is not applicable, performs poorly or gives insights that are not actionable. This tutorial will survey the recent and significant contributions in KDD and ML for sustainable development and will highlight current challenges that need to be addressed to transform and equip engaged sustainability science with robust ML-based tools to support actionable decision-making for a more sustainable future.

The 2nd International Workshop on Data Quality Assessment for Machine Learning (DQAML'21) is organized in conjunction with the Special Interest Group on Knowledge Discovery and Data Mining (SIGKDD). This workshop aims to serve as a forum for the presentation of research related to data quality assessment and remediation in AI/ML pipeline. Data quality is a critical issue in the data preparation phase and involves numerous challenging problems related to detection, remediation, visualization and evaluation of data issues. The workshop aims to provide a platform to researchers and practitioners to discuss such challenges across different modalities of data like structured, time series, text and graphical. The aim is to attract perspectives from both industrial and academic circles.

International audience

Data cleaning and preparation has been a long-standing challenge in data science to avoid incorrect results and misleading conclusions obtained from dirty data. For a given dataset and a given machine learning-based task, a plethora of data preprocessing techniques and alternative data curation strategies may lead to dramatically different outputs with unequal quality performance. Most current work on data cleaning and automated machine learning, however, focus on developing either cleaning algorithms or user-guided systems or

With the emergence of machine learning (ML) techniques in database research, ML has already proved a tremendous potential to dramatically impact the foundations, algorithms, and models of several data management tasks, such as error detection, data cleaning, data integration, and query inference. Part of the data preparation, standardization, and cleaning processes, such as data matching and deduplication for instance, could be automated by making a ML model "learn" and predict the matches routinely. Data integration can also benefit from ML as the data to be integrated can be sampled and used to design the data integration algorithms. After the initial manual work to setup the labels, ML models can start learning from the new incoming data that are being submitted for standardization, integration, and cleaning. The more data supplied to the model, the better the ML algorithm can perform and deliver accurate results. Therefore, ML is more scalable compared to traditional and time-consuming approaches. Nevertheless, many ML algorithms require an out-of-the-box tuning and their parameters and scope are often not adapted to the problem at hand. To make an example, in cleaning and integration processes, the window sizes of values used for the ML models cannot be arbitrarily chosen and require an adaptation of the learning parameters. This tutorial will survey the recent trend of applying machine learning solutions to improve data management tasks and establish new paradigms to sharpen data error detection, cleaning, and integration at the data instance level, as well as at schema, system, and user levels.

Functional dependencies (FDs) play an important role in maintaining data quality. They can be used to enforce data consistency and to guide repairs over a database. In this work, we investigate the problem of missing values and its impact on FD discovery. When using existing FD discovery algorithms, some genuine FDs could not be detected precisely due to missing values or some non-genuine FDs can be discovered even though they are caused by missing values with a certain NULL semantics. We define a notion of genuineness and propose algorithms to compute the genuineness score of a discovered FD. This can be used to identify the genuine FDs among the set of all valid dependencies that hold on the data. We evaluate the quality of our method over various real-world and semi-synthetic datasets with extensive experiments. The results show that our method performs well for relatively large FD sets and is able to accurately capture genuine FDs.

A large amount of Distributed Reflective Denial-of-Service (DRDoS) attacks are launched every day, and our understanding of the modus operandi of their perpetrators is yet very limited as we are submerged with so Big Data to analyze and do not have reliable and complete ways to validate our findings. In this paper, we propose a first analytic pipeline that enables us to cluster and characterize attack campaigns into several main profiles that exhibit similarities. These similarities are due to common technical properties of the underlying infrastructures used to launch these attacks. Although we do not have access to the ground truth and we do not know how many perpetrators are acting behind the scene, we can group their attacks based on relevant commonalities with cluster ensembling to estimate their number and capture their profiles over time. Specifically, our results show that we can repeatably identify and group together common profiles of attacks while considering domain expert's constraint in the cluster ensembles. From the obtained consensus clusters, we can generate comprehensive rules that characterize past campaigns and that can be used for classifying the next ones despite the evolving nature of the attacks. Such rules can be further used to filter out garbage traffic in Internet Service Provider networks.

Error detection is the process of identifying problematic data cells that are different from their ground truth. Functional dependencies (FDs) have been widely studied in support of this process. Oftentimes, it is assumed that FDs are given by experts. Unfortunately, it is usually hard and expensive for the experts to define such FDs. In addition, automatic data profiling over dirty data in order to find correct FDs is known to be a hard problem. In this paper, we propose an end-to-end solution to detect FD-detectable errors from dirty data. The broad intuition is that given a dirty dataset, it is feasible to automatically find approximate FDs, as well as data that is possibly erroneous. Arguably, at this point, only experts can confirm true FDs or true errors. However, in practice, experts never have enough budget to find all errors. Hence, our problem is, given a limited budget of expert's time, which questions we should ask, either FDs, cells, or tuples, such that we can find as many data errors as possible. We present efficient algorithms to interact with the user. Extensive experiments demonstrate that our proposed framework is effective in detecting errors from dirty data.

Recent studies have shown that the use of a priori knowledge can significantly improve the results of unsupervised classification. However, capturing and formatting such knowledge as constraints is not only very expensive requiring the sustained involvement of an expert but it is also very difficult because some valuable information can be lost when it cannot be encoded as constraints. In this paper, we propose a new constraint-based clustering approach based on ontology reasoning for automatically generating constraints and bridging the semantic gap in satellite image labeling. The use of ontology as a priori knowledge has many advantages that we leverage in the context of satellite image interpretation. The experiments we conduct have shown that our proposed approach can deal with incomplete knowledge while completely exploiting the available one.

The evolution of the Web from a technology platform to a social ecosystem has resulted in unprecedented data volumes being continuously generated, exchanged, and consumed. User-generated content on the Web is massive, highly dynamic, and characterized by a combination of factual data and opinion data. False information, rumors, and fake contents can be easily spread across multiple sources, making it hard to distinguish between what is true and what is not. Truth discovery (also known as fact-checking) has recently gained lot of interest from Data Science communities. This tutorial will attempt to cover recent work on truth-finding and how it can scale Big Data. We will provide a broad overview with new insights, highlighting the progress made on truth discovery from information extraction, data and knowledge fusion, as well as modeling of misinformation dynamics in social networks. We will review in details current models, algorithms, and techniques proposed by various research communities whose contributions converge towards the same goal of estimating the veracity of data in a dynamic world. Our aim is to bridge theory and practice and introduce recent work from diverse disciplines to database people to be better equipped for addressing the challenges of truth discovery in Big Data.

Many emerging applications, from domains such as healthcare and oil & gas, require several data processing systems for complex analytics. This demo paper showcases system, a framework that provides multi-platform task execution for such applications. It features a three-layer data processing abstraction and a new query optimization approach for multi-platform settings. We will demonstrate the strengths of system by using real-world scenarios from three different applications, namely, machine learning, data cleaning, and data fusion.

Estimation of data veracity is recognized as one of the grand challenges of big data. Typically, the goal of truth discovery is to determine the veracity of multi-source, conflicting data and return, as outputs, a veracity label and a confidence score for each data value, along with the trustworthiness score of each source claiming it. Although a plethora of methods has been proposed, it is unlikely a technique dominates all others across all data sets. Furthermore, the performance evaluation of the methods entirely depends on the availability of labeled ground truth data (i.e., data whose veracity has been manually checked). In the context of Big Data, acquiring the complete ground truth data is out-of-reach. In this paper, we propose an ensembling method that mitigates the two problems of method selection and ground truth data sparsity. Our approach combines the results of a set of truth discovery methods and preliminary experiments suggest that it improves the quality performance over the single methods when samples of ground truth data are used.

In the Web, a massive amount of user-generated contents is available through various channels, e.g., texts, tweets, Web tables, databases, multimedia-sharing platforms, etc. Conflicting information, rumors, erroneous and fake contents can be easily spread across multiple sources, making it hard to distinguish between what is true and what is not. How do you figure out that a lie has been told often enough that it is now considered to be true? How many lying sources are required to introduce confusion in what you knew before to be the truth? To answer these questions, we present AllegatorTrack, a system that discovers true claims among conflicting data from multiple sources.

The standardization and adoption of Semantic Web technologies has resulted in an unprecedented volume of data being published as Linked Data (LD). However, the “publish first, refine later” philosophy leads to various quality problems arising in the underlying data such as incompleteness, inconsistency and semantic ambiguities. In this article, we describe the current state of Data Quality in the Web of Data along with details of the three papers accepted for the International Journal on Semantic Web and Information Systems' (IJSWIS) Special Issue on Web Data Quality. Additionally, we identify new challenges that are specific to the Web of Data and provide insights into the current progress and future directions for each of those challenges.

Data glitches are errors in a data set, they are complex entities that often span multiple attributes and records. When they co-occur in data, the presence of one type of glitch can hinder the detection of another type of glitch. This phenomenon is called masking. In this paper, we define two important types of masking, and we propose a novel, statistically rigorous indicator called masking index for quantifying the hidden glitches in four cases of masking: outliers masked by missing values, outliers masked by duplicates, duplicates masked by missing values, and duplicates masked by outliers. The masking index is critical for data quality profiling and data exploration, it enables a user to measure the extent of masking and hence the confidence in the data. In this sense, it is a valuable data quality index for measuring the true cleanliness of the data. It is also an objective and quantitative basis for choosing an anomaly detection method that is best suited for the glitches that are present in any given data set. We demonstrate the utility and effectiveness of the masking index by intensive experiments on synthetic and real-world datasets.

Various computational procedures or constraint-based methods for data repairing have been proposed over the last decades to identify errors and, when possible, correct them. However, these approaches have several limitations including the scalability and quality of the values to be used in replacement of the errors. In this paper, we propose a new data repairing approach that is based on maximizing the likelihood of replacement data given the data distribution, which can be modeled using statistical machine learning techniques. This is a novel approach combining machine learning and likelihood methods for cleaning dirty databases by value modification. We develop a quality measure of the repairing updates based on the likelihood benefit and the amount of changes applied to the database. We propose SCARE (SCalable Automatic REpairing), a systematic scalable framework that follows our approach. SCARE relies on a robust mechanism for horizontal data partitioning and a combination of machine learning techniques to predict the set of possible updates. Due to data partitioning, several updates can be predicted for a single record based on local views on each data partition. Therefore, we propose a mechanism to combine the local predictions and obtain accurate final predictions. Finally, we experimentally demonstrate the effectiveness, efficiency, and scalability of our approach on real-world datasets in comparison to recent data cleaning approaches.

Quantitative Data Cleaning (QDC) is the use of statistical and other analytical techniques to detect, quantify, and correct data quality problems (or glitches). Current QDC approaches focus on addressing each category of data glitch individually. However, in real-world data, different types of data glitches co-occur in complex patterns. These patterns and interactions between glitches offer valuable clues for developing effective domain-specific quantitative cleaning strategies. In this paper, we address the shortcomings of the extant QDC methods by proposing a novel framework, the DEC (Detect-Explore-Clean) framework. It is a comprehensive approach for the definition, detection and cleaning of complex, multi-type data glitches. We exploit the distributions and interactions of different types of glitches to develop data-driven cleaning strategies that may offer significant advantages over blind strategies. The DEC framework is a statistically rigorous methodology for evaluating and scoring glitches and selecting the quantitative cleaning strategies that result in cleaned data sets that are statistically proximal to user specifications. We demonstrate the efficacy and scalability of the DEC framework on very large real-world and synthetic data sets.

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.

We consider the problem of efficiently finding the top- k answers for join queries over web-accessible databases. Classical algorithms for finding top- k answers use branch-and-bound techniques to avoid computing scores of all candidates in identifying the top- k answers. To be able to apply such techniques, it is critical to efficiently compute (lower and upper) bounds and expected scores of candidate answers in an incremental fashion during the evaluation. In this paper, we describe novel techniques for these problems. The first contribution of this paper is a method to efficiently compute bounds for the score of a query result when tuples in tables from the "FROM" clause are discovered incrementally, through either sorted or random access. Our second contribution is an algorithm that, given a set of partially evaluated candidate answers, determines a good order in which to access the tables to minimize wasted efforts in the computation of top- k answers. We evaluate our algorithms on a variety of queries and data sets and demonstrate the significant benefits they provide.

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.