Web Tables Research Articles

NPEL: Neural Paired Entity Linking in Web Tables

This paper studies entity linking (EL) in Web tables, which aims to link the string mentions in table cells to their referent entities in a knowledge base. Two main problems exist in previous studies: 1) contextual information is not well utilized in mention-entity similarity computation; 2) the assumption on entity coherence that all entities in the same row or column are highly related to each other is not always correct. In this paper, we propose NPEL , a new N eural P aired E ntity L inking framework, to overcome the above problems. In NPEL, we design a deep learning model with different neural networks and an attention mechanism, to model different kinds of contextual information of mentions and entities, for mention-entity similarity computation in Web tables. NPEL also relaxes the above assumption on entity coherence by a new paired entity linking algorithm, which iteratively selects two mentions with the highest confidence for EL. Experiments on real-world datasets exhibit that NPEL has the best performance compared with state-of-the-art baselines in different evaluation metrics.

Watchog: A Light-weight Contrastive Learning based Framework for Column Annotation

Relational Web tables provide valuable resources for numerous downstream applications, making table understanding, especially column annotation that identifies semantic types and relations of columns, a hot topic in the field of data management. Despite recent efforts to improve different tasks in table understanding by using the power of large pre-trained language models, existing methods heavily rely on large-scale and high-quality labeled instances, while they still suffer from the data sparsity problem due to the imbalanced data distribution among different classes. In this paper, we propose the Watchog framework, which employs contrastive learning techniques to learn robust representations for tables by leveraging a large-scale unlabeled table corpus with minimal overhead. Our approach enables the learned table representations to enhance fine tuning with much fewer additional labeled instances than in prior studies for downstream column annotation tasks. Besides, we further proposed optimization techniques for semi-supervised settings. Experimental results on popular benchmarking datasets illustrate the superiority of our proposed techniques in two column annotation tasks under different settings. In particular, our Watchog framework effectively alleviates the class imbalance issue caused by a long-tailed label distribution. In the semi-supervised setting, Watchog outperforms the best-known method by up to 26% and 41% in Micro and Macro F1 scores, respectively, on the task of semantic type detection.

CrowdDA: Difficulty-aware crowdsourcing task optimization for cleaning web tables

Web tables are rich sources of structured data for collecting and analyzing data, but there exist various data quality problems such as missing or inconsistent values. Crowdsourcing provides a new solution that leverages the human cognitive ability to clean tables, but existing crowdsourcing-based solutions suffer from cleaning quality. For most crowd are not experts, the difficulty degree of cleaning tasks will seriously affect the cleaning result. To help people clean web tables effectively and efficiently, it is important to reduce the overall difficulty of tasks. In this paper, we introduce a difficulty-aware crowdsourcing task optimization system CrowdDA, which aims to recommend the best task execution order from easy to difficult for crowd and support various kinds of cleaning tasks for web tables. CrowdDA takes both latency and space constraints into account for task optimization and generates the task execution order that minimizes the overall difficulty of tasks under two constraints. Furthermore, CrowdDA adopts partition strategies for large tables to improve system efficiency, and introduces independent task sequence to tolerate crowd’s inconsistent answers for system robustness. The experiments based on the real-world datasets demonstrate the performance superiority of CrowdDA for improving the cleaning quality of web tables.

Identifying Notable Tuples in Multi-Concept Web Tables

Identifying notable tuples in a web table is of great help for table understanding and table summarization. However, existing document-internal feature-based methods are inappropriate for identifying notable tuples in web tables. Additionally, for the web table describing multiple concepts, the notability evaluation of a tuple needs to take into account multiple entities as well as their importance in this tuple. In this paper, we investigate the task of identifying notable tuples in a multi-concept web table and propose a framework that includes three tasks: (1) identify multiple entity columns and their importance weights by building a column correlation graph based on types and relationships in the table; (2) obtain fine-grained entity notability scores based on entity link graph and provide solution for entity link failure and entity domain neglection; and (3) evaluate tuple notability by a weighted sum of notability scores of all entities in the tuple. Comprehensive evaluation of our approach is based on real-world web tables. The results demonstrate that our approach outperforms the state-of-the-art baselines by 4.6% on the precision of detecting multiple entity columns and by 12.5% on the metric normalized discounted cumulative gain (NDCG) of evaluating entity notability.

RECA: Related Tables Enhanced Column Semantic Type Annotation Framework

Understanding the semantics of tabular data is of great importance in various downstream applications, such as schema matching, data cleaning, and data integration. Column semantic type annotation is a critical task in the semantic understanding of tabular data. Despite the fact that various approaches have been proposed, they are challenged by the difficulties of handling wide tables and incorporating complex inter-table context information. Failure to handle wide tables limits the usage of column type annotation approaches, while failure to incorporate inter-table context harms the annotation quality. Existing methods either completely ignore these problems or propose ad-hoc solutions. In this paper, we propose Related tables Enhanced Column semantic type Annotation framework (RECA), which incorporates inter-table context information by finding and aligning schema-similar and topic-relevant tables based on a novel named entity schema. The design of RECA can naturally handle wide tables and incorporate useful inter-table context information to enhance the annotation quality. We conduct extensive experiments on two web table datasets to comprehensively evaluate the performance of RECA. Our results show that RECA achieves support-weighted F1 scores of 0.853 and 0.937 with macro average F1 scores of 0.674 and 0.783 on the two datasets respectively, which outperform the state-of-the-art methods.

Column-Type Prediction for Web Tables Powered by Knowledge Base and Text

Web tables are essential for applications such as data analysis. However, web tables are often incomplete and short of some critical information, which makes it challenging to understand the web table content. Automatically predicting column types for tables without metadata is significant for dealing with various tables from the Internet. This paper proposes a CNN-Text method to deal with this task, which fuses CNN prediction and voting processes. We present data augmentation and synthetic column generation approaches to improve the CNN’s performance and use extracted text to get better predictions. The experimental result shows that CNN-Text outperforms the baseline methods, demonstrating that CNN-Text is well qualified for the table column type prediction.

Russian Web Tables: A Public Corpus of Web Tables for Russian Language Based on Wikipedia

Corpora that contain tabular data such as WebTables are a vital resource for the academic community. Essentially, they are the backbone of any modern research in information management. They are used for various tasks of data extraction, knowledge base construction, question answering, column semantic type detection and many other. Such corpora are useful not only as a source of data, but also as a base for building test datasets. So far, there were no such corpora for the Russian language and this seriously hindered research in the aforementioned areas. In this paper, we present the first corpus of Web tables created specifically out of Russian language material. It was built via a special toolkit we have developed to crawl the Russian Wikipedia. Both the corpus and the toolkit are open-source and publicly available. Finally, we present a short study that describes Russian Wikipedia tables and their statistics.

Popularity sensitive and domain-aware summarization for web tables

Table summarization can be of great help, which generates a concise and informative overview of a table to assist users to understand the table easily and unambiguously. A high-quality summary needs to have two desirable properties: presenting notable entities in the table and achieving broad coverage and high diversity on domains. However, notability and domain are often neglected in table summarization. Thus in this paper, we present a framework of domain-aware table summarization that is able to: (1) identify notable entities using a popularity-sensitive notability evaluation algorithm, (2) find core domains with a measurement of domain centrality, (3) and output the final high-quality summary using a three-phase clustering based algorithm. The experimental results show that our summarization method outperforms state-of-the-art methods by 9.62%, 2.78% and 6.77% on metrics coverage, diversity, and notability, respectively. We also conduct a user study to demonstrate that people can improve the accuracy of understanding tables by 17% with the help of our summarization technique.

Hybrid Metadata Classification in Large-scale Structured Datasets

Metadata location and classification is an important problem for large-scale structured datasets. For example, Web tables \cite{wt_corpus} have hundreds of millions of tables, but often have missing or incorrect labels for rows (or columns) with attribute names. Such errors \cite{wtitles} significantly complicate all data management tasks such as {\em query processing, data integration, indexing}, etc. Different sources or authors position metadata rows/columns differently inside a table, which makes its reliable identification challenging.In this work we describe our scalable, hybrid two-layer Deep- and Machine-learning based ensemble, combining Long Short Term Memory (LSTM) and Naive Bayes Classifier to accurately identify Metadata-containing rows or columns in a table. We have performed an extensive evaluation on several datasets, including an ultra large-scale dataset containing more than 15 million tables coming from more than 26 thousands of sources to justify scalability and resistance to variety, stemming from a large number of sources. We observed superiority of this two-layer ensemble, compared to the recent previous approaches and report an impressive 95.73\text{\%} accuracy at scale with our ensemble model using regular LSTM.

TURL

Relational tables on the Web store a vast amount of knowledge. Owing to the wealth of such tables, there has been tremendous progress on a variety of tasks in the area of table understanding. However, existing work generally relies on heavily-engineered task-specific features and model architectures. In this paper, we present TURL, a novel framework that introduces the pre-training/fine-tuning paradigm to relational Web tables. During pre-training, our framework learns deep contextualized representations on relational tables in a self-supervised manner. Its universal model design with pre-trained representations can be applied to a wide range of tasks with minimal task-specific fine-tuning.

CWTs: A Public Large-Scale Chinese Web Table Data Set

As the world’s largest information sharing platform, web pages contain a large amount of table data. However, these table data emphasize the presentation of data to users in a more intuitive way, which cannot be understood by computers. How to obtain large-scale web table data and carry out in-depth applications has become the focus of data science. This paper designs an extraction framework for Chinese Web table data. This framework extracts Web table data containing Chinese from the Web and constructs an open Chinese Web table data set CWTs (Chines Web Tables). The data set is quantitatively analyzed in multiple dimensions such as scale and data distribution. The results show that the data set obtained in this paper can provide a preliminary data basis for table data query, question answering system, and knowledge base construction, at the same time provide an optional channel for building a large-scale machine-understandable Chinese knowledge base.

A clustering approach to extract data from HTML tables

HTML tables have become pervasive on the Web. Extracting their data automatically is difficult because finding the relationships between their cells is not trivial due to the many different layouts, encodings, and formats available. In this article, we introduce Melva, which is an unsupervised domain-agnostic proposal to extract data from HTML tables without requiring any external knowledge bases. It relies on a clustering approach that helps make label cells apart from value cells and establish their relationships. We compared Melva to four competitors on more than 3000 HTML tables from the Wikipedia and the Dresden Web Table Corpus. The conclusion is that our proposal is 21.70% better than the best unsupervised competitor and equals the best supervised competitor regarding effectiveness, but it is 99.14% better regarding efficiency.

Knowledge graphs 2021

Providing machines with comprehensive knowledge of the world's entities and their relationships has been a long-standing vision and challenge for AI. Over the last 15 years, huge knowledge bases, also known as knowledge graphs, have been automatically constructed from web data, and have become a key asset for search engines and other use cases. Machine knowledge can be harnessed to semantically interpret texts in news, social media and web tables, contributing to question answering, natural language processing and data analytics. This position paper reviews these advances and discusses lessons learned. It highlights the role of "DB thinking" in building and maintaining high-quality knowledge bases from web contents. Moreover, the paper identifies open challenges and new research opportunities. In particular, extracting quantitative measures of entities (e.g., height of buildings or energy efficiency of cars), from text and web tables, presents an opportunity to further enhance the scope and value of knowledge bases.

Automatic construction of RDF with web tables

Being a metadata model, the Resource Description Framework (RDF) has been widely used in the context of semantic extraction, unified organization and intelligent processing of a large number of data because RDF’s machine understandability. Knowledge graph based on RDF model, for example, is predominantly generated from massive data with various models and formats. While many efforts have been carried out in converting databases and XML into RDF, the generation of RDF from Web tables currently receives less attention due to their flexible and complex structure as well as difficult understanding. Given that billions of Web tables are widely used in real-world applications, in this paper, we devote to the automatic construction of RDF with diverse Web tables. We recognize total six representative types of Web tables that have been extensively applied in Web-based applications, including Web tables with simple structure and complex structure. For each type of Web tables, we propose our approach for converting it into RDF. Our approaches can well identify the semantics of Web tables with complex structure and then generate the corresponding RDF. We implement an RDF generation system and experimental results show the feasibility of our proposed approaches.

Improving Distant Supervised Relation Extraction with Noise Detection Strategy

Distant supervised relation extraction (DSRE) is widely used to extract novel relational facts from plain text, so as to improve the knowledge graph. However, distant supervision inevitably suffers from the noisy labeling problem that will severely damage the performance of relation extraction. Currently, most DSRE methods are mainly focused on reducing the weights of noisy sentences, ignoring the bag-level noise where all sentences in a bag are wrongly labeled. In this paper, we present a novel noise detection-based relation extraction approach (NDRE) to automatically detect noisy labels with entity information and dynamically correct them, which can alleviate both instance-level and bag-level noisy problems. By this means, we can extend the dataset from the Web tables without introducing more noise. In this approach, to embed the semantics of sentences from corpus and web tables, we firstly propose a powerful sentence coder that employs an internal multi-head self-attention mechanism between the piecewise max-pooling convolutional neural network. Second, we adopt a noise detection strategy, which is expected to dynamically detect and correct the original noisy label according to the similarity between sentence representation and entity-aware embeddings. Then, we aggregate the information from corpus and web tables to make the final relation prediction. Experimental results on a public benchmark dataset demonstrate that our proposed approach achieves significant improvements over the state-of-the-art baselines and can effectively reduce the noisy labeling problem.

Merging Web Tables for Relation Extraction with Knowledge Graphs

We propose methods for extracting triples from Wikipedia’s HTML tables using a reference knowledge graph. Our methods use a distant-supervision approach to find existing triples in the knowledge graph for pairs of entities on the same row of a table, postulating the corresponding relation for pairs of entities from other rows in the corresponding columns, thus extracting novel candidate triples. Binary classifiers are applied on these candidates to detect correct triples and thus increase the precision of the output triples. We extend this approach with a preliminary step where we first group and merge similar tables, thereafter applying extraction on the larger merged tables. More specifically, we propose an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">observed schema</i> for individual tables, which is used to group and merge tables. We compare the precision and number of triples extracted with and without table merging, where we show that with merging, we can extract a larger number of triples at a similar precision. Ultimately, from the tables of English Wikipedia, we extract 5.9 million novel and unique triples for Wikidata at an estimated precision of 0.718.

TURL

Relational tables on the Web store a vast amount of knowledge. Owing to the wealth of such tables, there has been tremendous progress on a variety of tasks in the area of table understanding. However, existing work generally relies on heavily-engineered task-specific features and model architectures. In this paper, we present TURL, a novel framework that introduces the pre-training/fine-tuning paradigm to relational Web tables. During pre-training, our framework learns deep contextualized representations on relational tables in an unsupervised manner. Its universal model design with pre-trained representations can be applied to a wide range of tasks with minimal task-specific fine-tuning. Specifically, we propose a structure-aware Transformer encoder to model the row-column structure of relational tables, and present a new Masked Entity Recovery (MER) objective for pre-training to capture the semantics and knowledge in large-scale unlabeled data. We systematically evaluate TURL with a benchmark consisting of 6 different tasks for table understanding (e.g., relation extraction, cell filling). We show that TURL generalizes well to all tasks and substantially outperforms existing methods in almost all instances.

Lists of Chemical Warfare Agents and Precursors from International Nonproliferation Frameworks: Structural Annotation and Chemical Fingerprint Analysis.

To support efforts to stem the proliferation of chemical weapons (CWs), we have curated and structurally annotated CW-control lists from three key international nonproliferation frameworks: the Chemical Weapons Convention (CWC), the Australia Group (AG), and the Wassenaar Arrangement. The curated lists are available as web tables at the Costanzi Research website (https://costanziresearch.com/cw-control-lists/). The annotations include manually curated 2D structural images, which provide a means to appreciate at a glance the similarities and differences between different entries, as well as downloadable 2D structures, in two different formats and three different structural identifiers, namely, simplified molecular-input line-entry system, standard InChI, and standard InChIKey, which are intended to provide a platform for cheminformatics analyses. The tables also include links to National Center for Biotechnology Information's PubChem and National Institute of Standards and Technology's Chemistry WebBook cards, hence providing prompt access to a wealth of physicochemical, analytical chemistry, and toxicological information. To showcase the importance of structural annotations, we discuss a discrepancy in a CW-control list covering the defoliant Agent Orange, which we identified through our curation process, and propose a solution to address it. Moreover, we present the results of chemical fingerprinting analyses, through which we clustered the entries of the three CW-control lists under study into structurally related groups and studied the overlaps between the three lists. As an application of this study, we examine the recent updates of CWC Schedule 1 and the AG precursors list, highlighting the relationships between the two amendments and proposing the possible addition of further chemicals. Our research is intended to facilitate the communication between scientific advisors and policymakers as well as the work of chemists and cheminformaticians involved in the CW nonproliferation field. Ultimately, we seek to provide tools to bolster the control of CWs and support the global efforts to rid the world of this category of weapons.

Learning cell embeddings for understanding table layouts

There is a large amount of data on the web in tabular form, such as Excel sheets, CSV files, and web tables. Often, tabular data is meant for human consumption, using data layouts that are difficult for machines to interpret automatically. Previous work uses the stylistic features of tabular cells (such as font size, border type, and background color) to classify tabular cells by their role in the data layout of the document (top attribute, data, metadata, etc.). In this paper, we propose a deep neural network model which can embed semantic and contextual information about tabular cells in a low-dimensional cell embedding space. We pre-train this cell embedding model on a large corpus of tabular documents from various domains. We then propose a classification technique based on recurrent neural networks (RNNs) to use our pre-trained cell embeddings, combining them with stylistic features introduced in previous work, in order to improve the performance of cell type classification in complex documents. We evaluate the performance of our system on three datasets containing documents with various data layouts, in two settings: in-domain and cross-domain training. Our evaluation result shows that our proposed cell vector representations in combination with our RNN-based classification technique significantly improve cell type classification performance.

Web Table Extraction, Retrieval, and Augmentation

Tables are powerful and popular tools for organizing and manipulating data. A vast number of tables can be found on the Web, which represent a valuable knowledge resource. The objective of this survey is to synthesize and present two decades of research on web tables. In particular, we organize existing literature into six main categories of information access tasks: table extraction, table interpretation, table search, question answering, knowledge base augmentation, and table augmentation. For each of these tasks, we identify and describe seminal approaches, present relevant resources, and point out interdependencies among the different tasks.