News recommendation for anonymous readers is a useful but challenging task for many news portals, where interactions between readers and articles are limited within a temporary login session. Previous works tend to formulate session-based recommendation as a next item prediction task, while they neglect the implicit feedback from user behaviors, which indicates what users really like or dislike. Hence, we propose a comprehensive framework to model user behaviors through positive feedback (i.e., the articles they spend more time on) and negative feedback (i.e., the articles they choose to skip without clicking in). Moreover, the framework implicitly models the user using their session start time, and the article using its initial publishing time, in what we call neutral feedback. Empirical evaluation on three real-world news datasets shows the framework's promising performance of more accurate, diverse and even unexpectedness recommendations than other state-of-the-art session-based recommendation approaches.

Audio tagging aims at predicting sound events occurred in a recording. Traditional models require enormous laborious annotations, otherwise performance degeneration will be the norm. Therefore, we investigate robust audio tagging models in low-resource scenarios with the enhancement of knowledge graphs. Besides existing ontological knowledge, we further propose a semi-automatic approach that can construct temporal knowledge graphs on diverse domain-specific label sets. Moreover, we leverage a variant of relation-aware graph neural network, D-GCN, to combine the strength of the two knowledge types. Experiments on AudioSet and SONYC urban sound tagging datasets suggest the effectiveness of the introduced temporal knowledge, and the advantage of the combined KGs with D-GCN over single knowledge source.

Abstractive summarization is useful in providing a summary or a digest of news or other web texts and enhancing users reading experience, especially when they are reading on small displays such as mobile phones. However, existing encoder-decoder summarization models have difficulty learning the latent alignment between source documents and summaries because of their vast disparity in length. In this paper, we propose a extractor-abstractor framework in which the keyword-based extractor selects a few sets of salient sentences from the input document and then the abstractor paraphrases these sets of sentences in parallel, which are more aligned to the summary, to generate the final summary. The new extractor and abstractor are pretrained from a set of “pseudo summaries” extracted by specially designed heuristics, and then further trained together in a reinforcement learning framework. The results show that the proposed model generates high-quality summaries with faster training speed and less training memory footprint, and outperforms the state-of-the-art models on CNN/Daily Mail, Webis-TLDR-17, Webis-Snippet-20, WikiHow and DUC-2002 datasets.

Product descriptions on e-commerce websites often suffer from missing important aspects. Clarification question generation (CQGen) can be a promising approach to help alleviate the problem. Unlike traditional QGen assuming the existence of answers in the context and generating questions accordingly, CQGen mimics user behaviors of asking for unstated information. The generated CQs can serve as a sanity check or proofreading to help e-commerce merchant to identify potential missing information before advertising their product, and improve consumer experience consequently. Due to the variety of possible user backgrounds and use cases, the information need can be quite diverse but also specific to a detailed topic, while previous works assume generating one CQ per context and the results tend to be generic. We thus propose the task of Diverse CQGen and also tackle the challenge of specificity. We propose a new model named KPCNet, which generates CQs with Keyword Prediction and Conditioning, to deal with the tasks. Automatic and human evaluation on 2 datasets (Home & Kitchen, Office) showed that KPCNet can generate more specific questions and promote better group-level diversity than several competing baselines. 1

Few-shot relation classification seeks to classify incoming query instances after meeting only few support instances. This ability is gained by training with large amount of in-domain annotated data. In this paper, we tackle an even harder problem by further limiting the amount of data available at training time. We propose a few-shot learning framework for relation classification, which is particularly powerful when the training data is very small. In this framework, models not only strive to classify query instances, but also seek underlying knowledge about the support instances to obtain better instance representations. The framework also includes a method for aggregating cross-domain knowledge into models by open-source task enrichment. Additionally, we construct a brand new dataset: the TinyRel-CM dataset, a few-shot relation classification dataset in health domain with purposely small training data and challenging relation classes. Experimental results demonstrate that our framework brings performance gains for most underlying classification models, outperforms the state-of-the-art results given small training data, and achieves competitive results with sufficiently large training data.

Predicting endings for narrative stories is a grand challenge for machine commonsense reasoning. The task requires ac- curate representation of the story semantics and structured logic knowledge. Pre-trained language models, such as BERT, made progress recently in this task by exploiting spurious statistical patterns in the test dataset, instead of 'understanding' the stories per se. In this paper, we propose to improve the representation of stories by first simplifying the sentences to some key concepts and second modeling the latent relation- ship between the key ideas within the story. Such enhanced sentence representation, when used with pre-trained language models, makes substantial gains in prediction accuracy on the popular Story Cloze Test without utilizing the biased validation data.

One of the ultimate goals of e-commerce platforms is to satisfy various shopping needs for their customers. Much efforts are devoted to creating taxonomies or ontologies in e-commerce towards this goal. However, user needs in e-commerce are still not well defined, and none of the existing ontologies has the enough depth and breadth for universal user needs understanding. The semantic gap in-between prevents shopping experience from being more intelligent. In this paper, we propose to construct a large-scale E-commerce Cognitive Concept Net named "AliCoCo", which is practiced in Alibaba, the largest Chinese e-commerce platform in the world. We formally define user needs in e-commerce, then conceptualize them as nodes in the net. We present details on how AliCoCo is constructed semi-automatically and its successful, ongoing and potential applications in e-commerce.

Secure online transaction is an essential task for e-commerce platforms. Alipay, one of the world’s leading cashless payment platform, provides the payment service to both merchants and individual customers. The fraud detection models are built to protect the customers, but stronger demands are raised by the new scenes, which are lacking in training data and labels. The proposed model makes a difference by utilizing the data under similar old scenes and the data under a new scene is treated as the target domain to be promoted. Inspired by this real case in Alipay, we view the problem as a transfer learning problem and design a set of revise strategies to transfer the source domain models to the target domain under the framework of gradient boosting tree models. This work provides an option for the cold-start and data-sharing problems.

Understanding latent user needs beneath shopping behaviors is critical to e-commercial applications. Without a proper definition of user needs in e-commerce, most industry solutions are not driven directly by user needs at current stage, which prevents them from further improving user satisfaction. Representing implicit user needs explicitly as nodes like "outdoor barbecue" or "keep warm for kids" in a knowledge graph, provides new imagination for various e- commerce applications. Backed by such an e-commerce knowledge graph, we propose a supervised learning algorithm to conceptualize user needs from their transaction history as "concept" nodes in the graph and infer those concepts for each user through a deep attentive model. Online experiments demonstrate the effectiveness and stability of our model, and online industry strength tests show substantial advantages of such user needs understanding.

This paper targets to a novel but practical recommendation problem named exact-K recommendation. It is different from traditional top-K recommendation, as it focuses more on (constrained) combinatorial optimization which will optimize to recommend a whole set of K items called card, rather than ranking optimization which assumes that "better" items should be put into top positions. Thus we take the first step to give a formal problem definition, and innovatively reduce it to Maximum Clique Optimization based on graph. To tackle this specific combinatorial optimization problem which is NP-hard, we propose Graph Attention Networks (GAttN) with a Multi-head Self-attention encoder and a decoder with attention mechanism. It can end-to-end learn the joint distribution of the K items and generate an optimal card rather than rank individual items by prediction scores. Then we propose Reinforcement Learning from Demonstrations (RLfD) which combines the advantages in behavior cloning and reinforcement learning, making it sufficient-and-efficient to train the model. Extensive experiments on three datasets demonstrate the effectiveness of our proposed GAttN with RLfD method, it outperforms several strong baselines with a relative improvement of 7.7% and 4.7% on average in Precision and Hit Ratio respectively, and achieves state-of-the-art (SOTA) performance for the exact-K recommendation problem.

Many emerging applications are based on finding interesting subsequences from sequence data. Finding “prominent streaks,” a set of the longest contiguous subsequences with values all above (or below) a certain threshold, from sequence data is one of that kind that receives much attention. Motivated from real applications, we observe that prominent streaks alone are not insightful enough but require the discovery of something we coined as “historic moments” as companions. In this article, we present an algorithm to efficiently compute historic moments from sequence data. The algorithm is incremental and space optimal , meaning that when facing new data arrival, it is able to efficiently refresh the results by keeping minimal information. Case studies show that historic moments can significantly improve the insights offered by prominent streaks alone. Furthermore, experiments show that our algorithm can outperform the baseline in both time and space.

Knowledge acquisition is an iterative process. Most prior work used syntactic bootstrapping approaches, while semantic bootstrapping was proposed recently. Unlike syntactic bootstrapping, semantic bootstrapping bootstraps directly on knowledge rather than on syntactic patterns, that is, it uses existing knowledge to understand the text and acquire more knowledge. It has been shown that semantic bootstrapping can achieve superb precision while retaining good recall on extracting isA relation. Nonetheless, the working mechanism of semantc bootstrapping remains elusive. In this extended abstract, we present a theoretical analysis as well as an experimental study to provide deeper insights into semantic bootstrapping.

Knowledge acquisition is an iterative process. Most previous work has focused on bootstrapping techniques based on syntactic patterns, that is, each iteration finds more syntactic patterns for subsequent extraction. However, syntactic bootstrapping is incapable of resolving the inherent ambiguities in the syntactic patterns. The precision of the extracted results is thus often poor. On the other hand, semantic bootstrapping bootstraps directly on knowledge rather than on syntactic patterns, that is, it uses existing knowledge to understand the text and acquire more knowledge. It has been shown that semantic bootstrapping can achieve superb precision while retaining good recall. Nonetheless, the working mechanism of semantic bootstrapping remains elusive. In this paper, we present a detailed analysis of semantic bootstrapping from a theoretical perspective. We show that the efficiency and effectiveness of semantic bootstrapping can be theoretically guaranteed. Our experimental evaluation results substantiate the theoretical analysis.

This paper studies the problem of automatic detection of false rumors on Sina Weibo, the popular Chinese microblogging social network. Traditional feature-based approaches extract features from the false rumor message, its author, as well as the statistics of its responses to form a flat feature vector. This ignores the propagation structure of the messages and has not achieved very good results. We propose a graph-kernel based hybrid SVM classifier which captures the high-order propagation patterns in addition to semantic features such as topics and sentiments. The new model achieves a classification accuracy of 91.3% on randomly selected Weibo dataset, significantly higher than state-of-the-art approaches. Moreover, our approach can be applied at the early stage of rumor propagation and is 88% confident in detecting an average false rumor just 24 hours after the initial broadcast.

Many location based services, such as FourSquare, Yelp, TripAdvisor, Google Places, etc., allow users to compose reviews or tips on points of interest (POIs), each having a geographical coordinates. These services have accumulated a large amount of such geo-tagged review data, which allows deep analysis of user preferences in POIs. This paper studies two types of user preferences to POIs: topical-region preference and category aware topical-aspect preference. We propose a unified probabilistic model to capture these two preferences simultaneously. In addition, our model is capable of capturing the interaction of different factors, including topical aspect, sentiment, and spatial information. The model can be used in a number of applications, such as POI recommendation and user recommendation, among others. In addition, the model enables us to investigate whether people like an aspect of a POI or whether people like a topical aspect of some type of POIs (e.g., bars) in a region, which offer explanation for recommendations. Experiments on real world datasets show that the model achieves significant improvement in POI recommendation and user recommendation in comparison to the state-of-the-art methods. We also propose an efficient online recommendation algorithm based on our model, which saves up to 90% computation time.

Measuring semantic similarity between two terms is essential for a variety of text analytics and understanding applications. Currently, there are two main approaches for this task, namely the knowledge based and the corpus based approaches. However, existing approaches are more suitable for semantic similarity between words rather than the more general multi-word expressions (MWEs), and they do not scale very well. Contrary to these existing techniques, we propose an efficient and effective approach for semantic similarity using a large scale semantic network. This semantic network is automatically acquired from billions of web documents. It consists of millions of concepts, which explicitly model the context of semantic relationships. In this paper, we first show how to map two terms into the concept space, and compare their similarity there. Then, we introduce a clustering approach to orthogonalize the concept space in order to improve the accuracy of the similarity measure. Finally, we conduct extensive studies to demonstrate that our approach can accurately compute the semantic similarity between terms of MWEs and with ambiguity, and significantly outperforms 12 competing methods under Pearson Correlation Coefficient. Meanwhile, our approach is much more efficient than all competing algorithms, and can be used to compute semantic similarity in a large scale.

Wikification, which stands for the process of linking terms in a plain text document to Wikipedia articles which represent the correct meanings of the terms, can be thought of as a generalized Word Sense Disambiguation problem. It disambiguates multi-word expressions (MWEs) in addition to single words. Existing Wikification techniques either models the context of a given term as well as the Wikipedia article as bags of words, or compute global constraints among Wikipedia concepts by the link graph or link distributions. The first method doesn't achieve good results because the MWEs can have very different meanings than its constituent words which themselves are ambiguous. The second method doesn't produce high accuracy because the link structure or link distribution is often biased or incomplete by themselves due to the fact that Wikipedia pages are often sparsely linked. In this paper, we present a simple but powerful framework of sense disambiguation using co-occurrences of Wikipedia links in the Wikipedia corpus. We propose an iterative method to enrich the sparsely-linked articles by adding more links and then use the resulting link co-occurrence matrix to disambiguate an input document by a sliding window algorithm. Our prototype system achieves 89.97% precision and 76.43% recall on average for three benchmark data and compares favorably against four state-of-the-art wikification techniques.

Computing semantic similarity between two terms is essential for a variety of text analytics and understanding applications. However, existing approaches are more suitable for semantic similarity between words rather than the more general multi-word expressions (MWEs), and they do not scale very well. Therefore, we propose a lightweight and effective approach for semantic similarity using a large scale semantic network automatically acquired from billions of web documents. Given two terms, we map them into the concept space, and compare their similarity there. Furthermore, we introduce a clustering approach to orthogonalize the concept space in order to improve the accuracy of the similarity measure. Extensive studies demonstrate that our approach can accurately compute the semantic similarity between terms with MWEs and ambiguity, and significantly outperforms 12 competing methods.

Clustering of images from search results can improve the user experience of image search. Most of the existing systems use both visual features and surrounding texts as signals for clustering while this paper demonstrates the use of an external knowledge base to make better sense out of the text signals in a prototype system called CISC. Once we understand the semantics of the text better, the result of the clustering is significantly improved. In addition to clustering the images by their semantic entities, our system can also conceptualize each image cluster into a set of concepts to represent the meaning of the cluster.

This paper is concerned with information extraction from top-k web pages, which are web pages that describe top k instances of a topic which is of general interest. Examples include “the 10 tallest buildings in the world”, “the 50 hits of 2010 you don't want to miss”, etc. Compared to other structured information on the web (including web tables), information in top-k lists is larger and richer, of higher quality, and generally more interesting. Therefore top-k lists are highly valuable. For example, it can help enrich open-domain knowledge bases (to support applications such as search or fact answering). In this paper, we present an efficient method that extracts top-k lists from web pages with high performance. Specifically, we extract more than 1.7 million top-k lists from a web corpus of 1.6 billion pages with 92.0% precision and 72.3% recall.

List data is an important source of structured data on the web. This paper is concerned with "top-k" pages, which are web pages that describe a list of k instances of a particular topic or concept. Examples include "the 10 tallest persons in the world" and "the 50 hits of 2010 you don't want to miss". Compared to normal web list data, "top-k" lists contain richer information and are easier to understand. Therefore the extraction of such lists can help enrich existing knowledge bases about general concepts, or act as a preprocessing step to produce facts for a fact answering engine. We present an efficient system that extracts the target lists from web pages with high accuracy. We have used the system to process up to 160 million, or 1/10 of a high-frequency web snapshot from Bing, and obtained over 140,000 lists with 90.4% precision.

Knowledge is indispensable to understanding. The ongoing information explosion highlights the need to enable machines to better understand electronic text in human language. Much work has been devoted to creating universal ontologies or taxonomies for this purpose. However, none of the existing ontologies has the needed depth and breadth for universal understanding. In this paper, we present a universal, probabilistic taxonomy that is more comprehensive than any existing ones. It contains 2.7 million concepts harnessed automatically from a corpus of 1.68 billion web pages. Unlike traditional taxonomies that treat knowledge as black and white, it uses probabilities to model inconsistent, ambiguous and uncertain information it contains. We present details of how the taxonomy is constructed, its probabilistic modeling, and its potential applications in text understanding.

In this demonstration, we will present LEARNPADS, a fully automatic system for generating ad hoc data processing tools. When presented with a collection of ad hoc data, the system (1) analyzes the data, (2) infers a PADS [4, 5] description, (3) generates parser, printer, validation and traversal libraries and (4) links these libraries with format-independent tool suites to form stand-alone applications. These applications provide statistical analysis, XML conversion, CSV conversion, the ability to query with the Galax XQuery engine [3], and the ability to graph selected data elements, all directly from ASCII ad hoc data without human intervention. SIGMOD attendees will see both the user experience with LEARNPADS and the internals of the multi-phase inference algorithm which lies at the heart of the system.

We propose a new main memory index structure for abstract regions (objects) which may heavily overlap, the RCtree. These objects are "dynamic" and may have short life spans. The novelty is that rather than representing an object by its minimum bounding rectangle (MBR), possibly with pre-processed segmentation into many small MBRs, we use the actual shape of the object to maintain the index. This saves significant space for objects with large spatial extents since pre-segmentation is not needed. We show that the query performance of RC-tree is much better than many indexing schemes on synthetic overlapping data sets. The performance is also competitive on real-life GIS nonoverlapping data sets.