DiSCoMaT: Distantly Supervised Composition Extraction from Tables in Materials Science Articles (2022-07-03T00:00:00.000000Z)

TL;DR

This work defines a novel NLP task of extracting compositions of materials from tables in materials science papers, and presents DiSCoMaT, a strong baseline that combines multiple graph neural networks with several task-specific regular expressions, features, and constraints.

Abstract

A crucial component in the curation of KB for a scientific domain (e.g., materials science, food & nutrition, fuels) is information extraction from tables in the domain’s published research articles. To facilitate research in this direction, we define a novel NLP task of extracting compositions of materials (e.g., glasses) from tables in materials science papers. The task involves solving several challenges in concert, such as tables that mention compositions have highly varying structures; text in captions and full paper needs to be incorporated along with data in tables; and regular languages for numbers, chemical compounds, and composition expressions must be integrated into the model. We release a training dataset comprising 4,408 distantly supervised tables, along with 1,475 manually annotated dev and test tables. We also present DiSCoMaT, a strong baseline that combines multiple graph neural networks with several task-specific regular expressions, features, and constraints. We show that DiSCoMaT outperforms recent table processing architectures by significant margins. We release our code and data for further research on this challenging IE task from scientific tables.

Authors

Tanishq Gupta

1 papers

Mohd Zaki

1 papers

Devanshi Khatsuriya

1 papers

TL;DR

Abstract

Authors

References73 items

Component repair cost functions in Indian context for seismic loss estimation of reinforced concrete buildings

BatteryBERT: A Pretrained Language Model for Battery Database Enhancement

Quantifying the advantage of domain-specific pre-training on named entity recognition tasks in materials science

Elucidating the Influence of Structure and Ag+-Na+ Ion-Exchange on Crack-Resistance and Ionic conductivity of Na3Al1.8Si1.65P1.8O12 Glass Electrolyte

Automated pipeline for superalloy data by text mining

Evaluation of the nutritional value of seaweed products for broiler chickens’ nutrition

MatSciBERT: A materials domain language model for text mining and information extraction

Topic Transferable Table Question Answering

TAPEX: Table Pre-training via Learning a Neural SQL Executor

Scientific Language Models for Biomedical Knowledge Base Completion: An Empirical Study

CLTR: An End-to-End, Transformer-Based System for Cell-Level Table Retrieval and Table Question Answering

TABBIE: Pretrained Representations of Tabular Data

Capturing Row and Column Semantics in Transformer Based Question Answering over Tables

Donkey milk and fermented donkey milk: are there differences in the nutritional value and physicochemical characteristics?

Artificial intelligence and machine learning in glass science and technology: 21 challenges for the 21 st century

Neural Relation Extraction on Wikipedia Tables for Augmenting Knowledge Graphs

Extracting a Knowledge Base of Mechanisms from COVID-19 Papers

Transformers: State-of-the-Art Natural Language Processing

DeepTable: a permutation invariant neural network for table orientation classification

INFOTABS: Inference on Tables as Semi-structured Data

TaBERT: Pretraining for Joint Understanding of Textual and Tabular Data

Understanding the role of post-indentation recovery on the hardness of glasses: Case of silica, borate, and borosilicate glasses

TaPas: Weakly Supervised Table Parsing via Pre-training

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks.

Answering Conversational Questions on Structured Data without Logical Forms

TabNet: Attentive Interpretable Tabular Learning

Named Entity Recognition and Normalization Applied to Large-Scale Information Extraction from the Materials Science Literature

The Materials Science Procedural Text Corpus: Annotating Materials Synthesis Procedures with Shallow Semantic Structures

A Machine Learning Approach to Zeolite Synthesis Enabled by Automatic Literature Data Extraction

Information extraction from scientific articles: a survey

AllenNLP: A Deep Semantic Natural Language Processing Platform

Graph Attention Networks

ChemDataExtractor: A toolkit for automated extraction of chemical information from the scientific literature

Understanding the Semantic Structures of Tables with a Hybrid Deep Neural Network Architecture

Numerical Relation Extraction with Minimal Supervision

Effects of Al3 +, W6 +, Nb5 + and Pb2 + on the structure and properties of borotellurite glasses

Cyclical Learning Rates for Training Neural Networks

KnowLife: a versatile approach for constructing a large knowledge graph for biomedical sciences

An overview of the BIOASQ large-scale biomedical semantic indexing and question answering competition

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Adam: A Method for Stochastic Optimization

Influence of NaX (X=I or Cl) additions on GeS2–Ga2S3 based glasses

The structural role of tellurium dioxide in lead borophosphate glasses

Open-domain quantity queries on web tables: annotation, response, and consensus models

Chemical composition and antinociceptive effect of aqueous extract from Rourea induta Planch. leaves in acute and chronic pain models.

Structural investigations and calorimetric dissolution of manganese phosphate glasses

Understanding Tables in Context Using Standard NLP Toolkits

Commentary: The Materials Project: A materials genome approach to accelerating materials innovation

Annotating and searching web tables using entities, types and relationships

Kerr studies of several tellurite glasses

Distant supervision for relation extraction without labeled data

Structural manifestations of the mixed-alkali effect: NMR studies of sodium rubidium borate glasses

Compositional dependence of ultraviolet fluorescence intensity of Ce3+ in silicate, borate, and phosphate glasses

Spectroscopic investigation of MoO3–Fe2O3–P2O5 and SrO–Fe2O3–P2O5 glasses. Part I

X-ray absorption studies of the local strontium environments in borosilicate waste glasses

Optimization of Dy3+-doped Ge–Ga–As–S–CsBr glass composition and its 1.31 μm emission properties

Local atomic order of Ge–Se–Br glasses

Spectroscopy and thermal properties of Ga2S3 based glasses

XInfoTabS: Evaluating Multilingual Tabular Natural Language Inference

Trans-KBLSTM: An External Knowledge Enhanced Transformer BiLSTM Model for Tabular Reasoning

Effect of fuel nuclide composition on the fuel lifetime of the RITM-200 reactor unit

Effect of fuel composition and EGR on spark-ignited engine combustion with LPG fueling: Experimental and numerical investigation

Natural Language Inference over Tables: Enabling Explainable Data Exploration on Data Lakes

Elucidating the Influence of Structure and Ag + -Na + Ion-Exchange on Crack-Resistance and Ionic Conductivity of Na 3Al 1.8Si 1.65P 1.8O 12 Glass Electrolyte

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

A Primal Dual Formulation For Deep Learning With Constraints

URL https://www

International glass database system

Querying biomedical Linked Data with natural language questions

Dropout: a simple way to prevent neural networks from overfitting

Epam/sciglass

Japan New Glass Forum

Field of Study

Journal Information

Name