Creating Training Data for Scientific Named Entity Recognition with Minimal Human Effort

Tchoua, Roselyne; Ajith, Aswathy; Hong, Zhi; Ward, Logan; Chard, Kyle; Belikov, Alexander V.; Audus, Debra J.; Patel, Shrayesh N.; Pablo, Juan J. de; Foster, Ian

doi:10.1007/978-3-030-22734-0_29

Cited by 13 publications

(7 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the social science dataset we apply only FastText word embeddings and lexicon features, as the previous experiment demonstrated that this configuration performed best of the configurations studied. Table 3 compares the precision, recall, and F1 scores of SciNER to our previous work, in which we used a KNN classifier and many manually created rules [29]. Even in this quite different environment, SciNER achieves an F1 score of 0.847, significantly outperforming our rule-based approach that achieves an F1 score of 0.592.…”

Section: Experiments On the Social Science Corpusmentioning

confidence: 99%

See 1 more Smart Citation

SciNER: Extracting Named Entities from Scientific Literature

Hong

Tchoua

Chard

et al. 2020

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

The automated extraction of claims from scientific papers via computer is difficult due to the ambiguity and variability inherent in natural language. Even apparently simple tasks, such as isolating reported values for physical quantities (e.g., "the melting point of X is Y") can be complicated by such factors as domain-specific conventions about how named entities (the X in the example) are referenced. Although there are domain-specific toolkits that can handle such complications in certain areas, a generalizable, adaptable model for scientific texts is still lacking. As a first step towards automating this process, we present a generalizable neural network model, SciNER, for recognizing scientific entities in free text. Based on bidirectional LSTM networks, our model combines word embeddings, subword embeddings, and external knowledge (from DBpedia) to boost its accuracy. Experiments show that our model outperforms a leading domain-specific extraction toolkit by up to 50%, as measured by F1 score, while also being easily adapted to new domains.

show abstract

Section: Experiments On the Social Science Corpusmentioning

confidence: 99%

“…Each paper was reviewed by two expert reviewers to label polymer names, and when disagreement arose a third more senior domain expert made the final decision. The result is a list of 495 polymer names identified by experts in the 100 papers [29].…”

Section: Data Preparationmentioning

confidence: 99%

SciNER: Extracting Named Entities from Scientific Literature

Hong

Tchoua

Chard

et al. 2020

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…The early applications of chemical NER were mainly focused on extraction of drugs and biochemical information to perform more effective document searches ( Corbett and Copestake, 2008 ; Jessop et al., 2011 ; Rocktäschel et al., 2012 ; García-Remesal et al, 2013 ). Recently, chemical NER has shifted toward (in)organic materials and their characteristics ( Swain and Cole, 2016 ; He et al., 2020 ; Weston et al., 2019 ; Shah et al., 2018 ), polymers ( Tchoua et al., 2019 ), nanoparticles ( Hiszpanski et al., 2020 ), synthesis actions and conditions ( Vaucher et al., 2020 ; Hawizy et al., 2011 ; Kim et al., 2017c ; Kononova et al., 2019 ). The methods used for NER vary from traditional rule-based and dictionary look-up approaches to modern methodology built around advanced ML and NLP techniques, including conditional random field (CRF) ( Lafferty et al., 2001 ), long short-term memory (LSTM) neural networks ( Hochreiter and Schmidhuber, 1997 ), and others.…”

Section: Text Mining Of Scientific Literaturementioning

confidence: 99%

Opportunities and challenges of text mining in materials research

et al. 2021

View full text Add to dashboard Cite

Research publications are the major repository of scientific knowledge. However, their unstructured and highly heterogenous format creates a significant obstacle to large-scale analysis of the information contained within. Recent progress in natural language processing (NLP) has provided a variety of tools for high-quality information extraction from unstructured text. These tools are primarily trained on non-technical text and struggle to produce accurate results when applied to scientific text, involving specific technical terminology. During the last years, significant efforts in information retrieval have been made for biomedical and biochemical publications. For materials science, text mining (TM) methodology is still at the dawn of its development. In this review, we survey the recent progress in creating and applying TM and NLP approaches to materials science field. This review is directed at the broad class of researchers aiming to learn the fundamentals of TM as applied to the materials science publications.

show abstract

“…Therefore, the recognition of polymer names is not easy. For such variations, machine learning techniques are effective and have been used for this type of recognition [15,16]. However, because manual preparation of the training data requires high cost, remarkable results have not yet been obtained.…”

Section: Introductionmentioning

confidence: 99%

Machine extraction of polymer data from tables using XML versions of scientific articles

Oka

Yoshizawa

Shindo

et al. 2021

Science and Technology of Advanced Materials: Methods

View full text Add to dashboard Cite

In this study, we examined machine extraction of polymer data from tables in scientific articles. The extraction system consists of five processes: table extraction, data formatting, polymer name recognition, property specifier identification, and data extraction. Tables were first extracted in plain text. XML versions of scientific articles were used, and the tabular forms were accurately extracted, even for complicated tables, such as multi-column, multi-row, and merged tables. Polymer name recognition was performed using a named entity recognizer created by deep neural network learning of polymer names. The preparation cost of the training data was reduced using a rule-based algorithm. The target polymer properties in this study were glass transition temperature (T g ), melting temperature (T m ), and decomposition temperature (T d ), and the specifiers were identified using partial string matching. Through these five processes, 2,181 data points for T g , 1,526 for T m , and 2,316 for T d were extracted from approximately 18,000 scientific articles published by Elsevier. Nearly half of them were extracted from complicated tables. The F-scores for the extraction were 0.871, 0.870, and 0.841, respectively. These results indicate that the extraction system created in this study can rapidly and accurately collect large amounts of polymer data from tables in scientific literature.

show abstract

Creating Training Data for Scientific Named Entity Recognition with Minimal Human Effort

Cited by 13 publications

References 31 publications

SciNER: Extracting Named Entities from Scientific Literature

SciNER: Extracting Named Entities from Scientific Literature

Opportunities and challenges of text mining in materials research

Machine extraction of polymer data from tables using XML versions of scientific articles

Contact Info

Product

Resources

About