A significant amount of the world's knowledge is stored in relational databases. However, the ability for users to retrieve facts from a database is limited due to a lack of understanding of query languages such as SQL. We propose Seq2SQL, a deep neural network for translating natural language questions to corresponding SQL queries. Our model leverages the structure of SQL queries to significantly reduce the output space of generated queries. Moreover, we use rewards from in-the-loop query execution over the database to learn a policy to generate unordered parts of the query, which we show are less suitable for optimization via cross entropy loss. In addition, we will publish WikiSQL, a dataset of 80654 hand-annotated examples of questions and SQL queries distributed across 24241 tables from Wikipedia. This dataset is required to train our model and is an order of magnitude larger than comparable datasets. By applying policy-based reinforcement learning with a query execution environment to WikiSQL, our model Seq2SQL outperforms attentional sequence to sequence models, improving execution accuracy from 35.9% to 59.4% and logical form accuracy from 23.4% to 48.3%.

/pdf/seq2sql-generating-structured-queries-from-natural-language-rp21m0b93s.pdf

Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning

Semantic Web in data mining and knowledge discovery

We describe a deployed scalable system for organizing published scientific literature into a heterogeneous graph to facilitate algorithmic manipulation and discovery. The resulting literature graph consists of more than 280M nodes, representing papers, authors, entities and various interactions between them (e.g., authorships, citations, entity mentions). We reduce literature graph construction into familiar NLP tasks (e.g., entity extraction and linking), point out research challenges due to differences from standard formulations of these tasks, and report empirical results for each task. The methods described in this paper are used to enable semantic features in this http URL

Construction of the Literature Graph in Semantic Scholar

Existing question answering datasets focus on dealing with homogeneous information, based either only on text or KB/Table information alone. However, as human knowledge is distributed over heterogeneous forms, using homogeneous information alone might lead to severe coverage problems. To fill in the gap, we present HybridQA, a new large-scale question-answering dataset that requires reasoning on heterogeneous information. Each question is aligned with a Wikipedia table and multiple free-form corpora linked with the entities in the table. The questions are designed to aggregate both tabular information and text information, i.e., lack of either form would render the question unanswerable. We test with three different models: 1) a table-only model. 2) text-only model. 3) a hybrid model that combines heterogeneous information to find the answer. The experimental results show that the EM scores obtained by two baselines are below 20%, while the hybrid model can achieve an EM over 40%. This gap suggests the necessity to aggregate heterogeneous information in HybridQA. However, the hybrid model’s score is still far behind human performance. Hence, HybridQA can serve as a challenging benchmark to study question answering with heterogeneous information.

/pdf/hybridqa-a-dataset-of-multi-hop-question-answering-over-q3uai0uxq9.pdf

HybridQA: A Dataset of Multi-Hop Question Answering over Tabular and Textual Data

The problem of verifying whether a textual hypothesis holds based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are mainly restricted to dealing with unstructured evidence (e.g., natural language sentences and documents, news, etc), while verification under structured evidence, such as tables, graphs, and databases, remains under-explored. This paper specifically aims to study the fact verification given semi-structured data as evidence. To this end, we construct a large-scale dataset called TabFact with 16k Wikipedia tables as the evidence for 118k human-annotated natural language statements, which are labeled as either ENTAILED or REFUTED. TabFact is challenging since it involves both soft linguistic reasoning and hard symbolic reasoning. To address these reasoning challenges, we design two different models: Table-BERT and Latent Program Algorithm (LPA). Table-BERT leverages the state-of-the-art pre-trained language model to encode the linearized tables and statements into continuous vectors for verification. LPA parses statements into programs and executes them against the tables to obtain the returned binary value for verification. Both methods achieve similar accuracy but still lag far behind human performance. We also perform a comprehensive analysis to demonstrate great future opportunities. The data and code of the dataset are provided in \url{this https URL}.

TabFact: A Large-scale Dataset for Table-based Fact Verification

Web tables form a valuable source of relational data. The Web contains an estimated 154 million HTML tables of relational data, with Wikipedia alone containing 1.6 million high-quality tables. Extracting the semantics of Web tables to produce machine-understandable knowledge has become an active area of research.

A key step in extracting the semantics of Web content is entity linking EL: the task of mapping a phrase in text to its referent entity in a knowledge base KB. In this paper we present TabEL, a new EL system for Web tables. TabEL differs from previous work by weakening the assumption that the semantics of a table can be mapped to pre-defined types and relations found in the target KB. Instead, TabEL enforces soft constraints in the form of a graphical model that assigns higher likelihood to sets of entities that tend to co-occur in Wikipedia documents and tables. In experiments, TabEL significantly reduces error when compared to current state-of-the-art table EL systems, including a $$75\%$$ error reduction on Wikipedia tables and a $$60\%$$ error reduction on Web tables. We also make our parsed Wikipedia table corpus and test datasets publicly available for future work.

/pdf/tabel-entity-linking-in-web-tables-5dvu4nq3of.pdf

TabEL: Entity Linking in Web Tables

Knowledge bases extracted automatically from the Web present new opportunities for data mining and exploration. Given a large, heterogeneous set of extracted relations, new tools are needed for searching the knowledge and uncovering relationships of interest. We present WikiTables, a Web application that enables users to interactively explore tabular knowledge extracted from Wikipedia.In experiments, we show that WikiTables substantially outperforms baselines on the novel task of automatically joining together disparate tables to uncover "interesting" relationships between table columns. We find that a "Semantic Relatedness" measure that leverages the Wikipedia link structure accounts for a majority of this improvement. Further, on the task of keyword search for tables, we show that WikiTables performs comparably to Google Fusion Tables despite using an order of magnitude fewer tables. Our work also includes the release of a number of public resources, including over 15 million tuples of extracted tabular data, manually annotated evaluation sets, and public APIs.

https://dl.acm.org/doi/pdf/10.1145/2501511.2501516

Methods for exploring and mining tables on Wikipedia

The advancements of search engines for traditional text documents have enabled the effective retrieval of massive textual information in a resource-efficient manner. However, such conventional search methodologies often suffer from poor retrieval accuracy especially when documents exhibit unique properties that behoove specialized and deeper semantic extraction. Recently, AlgorithmSeer, a search engine for algorithms has been proposed, that extracts pseudo-codes and shallow textual metadata from scientific publications and treats them as traditional documents so that the conventional search engine methodology could be applied. However, such a system fails to facilitate user search queries that seek to identify algorithm-specific information, such as the datasets on which algorithms operate, the performance of algorithms, and runtime complexity, etc. In this paper, a set of enhancements to the previously proposed algorithm search engine are presented. Specifically, we propose a set of methods to automatically identify and extract algorithmic pseudo-codes and the sentences that convey related algorithmic metadata using a set of machine-learning techniques. In an experiment with over 93,000 text lines, we introduce 60 novel features, comprising content-based, font style based and structure-based feature groups, to extract algorithmic pseudo-codes. Our proposed pseudo-code extraction method achieves 93.32% F1-score, outperforming the state-of-the-art techniques by 28%. Additionally, we propose a method to extract algorithmic-related sentences using deep neural networks and achieve an accuracy of 78.5%, outperforming a Rule-based model and a support vector machine model by 28% and 16%, respectively.

Deep Learning-based Extraction of Algorithmic Metadata in Full-Text Scholarly Documents

Distributed representations of words have been shown to capture lexical semantics, as demonstrated by their effectiveness in word similarity and analogical relation tasks. But, these tasks only evaluate lexical semantics indirectly. In this paper, we study whether it is possible to utilize distributed representations to generate dictionary definitions of words, as a more direct and transparent representation of the embeddings' semantics. We introduce definition modeling, the task of generating a definition for a given word and its embedding. We present several definition model architectures based on recurrent neural networks, and experiment with the models over multiple data sets. Our results show that a model that controls dependencies between the word being defined and the definition words performs significantly better, and that a character-level convolution layer designed to leverage morphology can complement word-level embeddings. Finally, an error analysis suggests that the errors made by a definition model may provide insight into the shortcomings of word embeddings.

Thanapon Noraset

Papers

TabEL: Entity Linking in Web Tables

Methods for exploring and mining tables on Wikipedia

Deep Learning-based Extraction of Algorithmic Metadata in Full-Text Scholarly Documents

Definition Modeling: Learning to define word embeddings in natural language

Definition modeling: Learning to define word embeddings in natural language