Predictive article recommendation using natural language processing and machine learning to support evidence updates in domain-specific knowledge graphs

Sharma, Bhuvan; Willis, Van C; Huettner, Claudia S.; Beaty, Kirk; Snowdon, Jane L.; Xue, Shang; South, Brett R.; Jackson, Gretchen Purcell; Weeraratne, Dilhan; Michelini, Vanessa V.

doi:10.1093/jamiaopen/ooaa028

Cited by 3 publications

(6 citation statements)

References 31 publications

(19 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sharma et al [ 95 ] propose IBM PARSe, a paper recommendation system for the medical domain to reduce the number of papers to review for keeping an existing knowledge graph up-to-date. Classifiers identify new papers from target domains, named entity recognition finds relevant medical concepts before papers’ TF-IDF vectors are compared to ones in the knowledge graph.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Only six of the observed papers incorporate knowledge graphs. Only one uses a predefined one, the Watson for Genomics knowledge graph [95]. Most of the approaches build their own knowledge graphs, only one asks users to construct the graphs: Wang et al [109] build two knowledge graphs, one in-domain and one cross-domain graph.…”

Section: Knowledge Graphmentioning

confidence: 99%

“…For application between papers we encountered the possibility of using unspecified embeddings: unspecified word or vector representations of papers [30,48,107,110], papers' key terms or top words [2,98] and key phrases [46]. We found some approaches using vector space model variants: unspecified [59], tf vectors [39,88], tf-idf vectors [42,95,111], dimensionality reduced tf-idf vectors [86] and lastly, tf-idf and entity embeddings [56]. Some approaches incorporated more advanced embedding techniques: SBERT embeddings [5], Doc2Vec embeddings [28], Doc2Vec embeddings with incorporation of their emotional score [109] and NPLM representations [29].…”

Section: Cosine Similaritymentioning

confidence: 99%

“…Even though there are still approaches (not necessarily CF ones) utilising basic TF-IDF representations of papers [2,42,86,95], nowadays this problem can be bypassed by using a text embedding method such as Doc2Vec or BERT.…”

Section: Synonymymentioning

confidence: 99%

“…Hua et al [42] consider users who start diving into a topic which they have not yet researched before. Sharma et al [95] name subject matter experts incorporating articles into a medical knowledge base as their target audience. Shi et al [96] clearly state use cases for their approach which always target users which are unaware of a topic but already have one interesting paper from the area.…”

Section: Target Audiencementioning

confidence: 99%

See 4 more Smart Citations

Scientific paper recommendation systems: a literature review of recent publications

Kreutz

Schenkel

2022

Int J Digit Libr

View full text Add to dashboard Cite

Scientific writing builds upon already published papers. Manual identification of publications to read, cite or consider as related papers relies on a researcher’s ability to identify fitting keywords or initial papers from which a literature search can be started. The rapidly increasing amount of papers has called for automatic measures to find the desired relevant publications, so-called paper recommendation systems. As the number of publications increases so does the amount of paper recommendation systems. Former literature reviews focused on discussing the general landscape of approaches throughout the years and highlight the main directions. We refrain from this perspective, instead we only consider a comparatively small time frame but analyse it fully. In this literature review we discuss used methods, datasets, evaluations and open challenges encountered in all works first released between January 2019 and October 2021. The goal of this survey is to provide a comprehensive and complete overview of current paper recommendation systems.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Section: Knowledge Graphmentioning

confidence: 99%

Section: Cosine Similaritymentioning

confidence: 99%

Section: Synonymymentioning

confidence: 99%

Section: Target Audiencementioning

confidence: 99%

See 3 more Smart Citations

Scientific paper recommendation systems: a literature review of recent publications

Kreutz

Schenkel

2022

Int J Digit Libr

View full text Add to dashboard Cite

show abstract

The use of artificial intelligence for automating or semi-automating biomedical literature analyses: A scoping review

Santos

Silva

Couto

et al. 2023

Journal of Biomedical Informatics

View full text Add to dashboard Cite

Mitigation of Chemical Reporting Liabilities through Systematic Modernization of Chemical Hazard and Safety Data Management Systems

2023

View full text Add to dashboard Cite

In the United States alone, approximately 2 billion tons of hazardous material products are manufactured each year for both household and industrial applications and contribute to thousands of worker chemical exposures with as many as 50,000 deaths from prolonged exposure each year. The potential hazards and impacts of these chemicals for human health and the environment are primarily communicated to the public through Safety Data Sheets (SDSs) from the chemical vendors or distributors. These documents provide a standardized approach for how and what information is provided to product users to assist them with assessment of precautionary measures, hazard mitigation, emergency response or cleanup procedures, and environmental, health, and safety (EHS) management. Despite the criticality for hazard communication (HAZCOM) precision, legacy SDS management and industry business practices leave the overall ability to effectively manage chemicals vulnerable to significant liability through a lack of full constituent disclosure, injection of data quality errors through various handling of SDS information and manual data entry, and the lack of direct SDS-to-product association. Chemical spills and accidents often require individuals to look for the appropriate SDS on a local computer, online, or in workplace binders; each of which results in information returned that is often found to be outdated or incorrect. Workplace HAZCOM violations remain among the top citations during EHS inspections by regulatory agencies. More important, however, is the lack of precise association of SDS to hazardous products that can occur through chemical management lifecycles. Incorrect SDSs can yield significant liability, as subsequent environmental and occupational health analyses and reporting are based upon incorrect and, in some cases, entirely different chemical formulations. This paper focuses on the need for a paradigm shift in our chemical management systems and how a standardized management system and various recent technological advances can be incorporated into Environmental Management System operations to reduce or eliminate these liabilities. The following advancements can be used to enhance the lifecycle management of workplace chemicals, reduce potential exposure and spill risks, reduce workplace hazards, and increase the efficiency and accuracy of environmental reporting through a more streamlined systems approach. EHS system enhancement applications discussed in this paper include the following: the need for a centralized universal SDS repository with full chemical disclosure of all product constituents and a nationally adopted machine language SDS standard. The use of artificial intelligence/machine learning in environmental systems and how they can be used as a medium to transition toward an automated standard by reverse-engineering and partitioning SDS components into machine-encoded text that can be validated and uploaded to a centralized repository. Algorithmic and meta-algorithmic approaches to SDS requirement and data...

show abstract

Predictive article recommendation using natural language processing and machine learning to support evidence updates in domain-specific knowledge graphs

Cited by 3 publications

References 31 publications

Scientific paper recommendation systems: a literature review of recent publications

Scientific paper recommendation systems: a literature review of recent publications

The use of artificial intelligence for automating or semi-automating biomedical literature analyses: A scoping review

Mitigation of Chemical Reporting Liabilities through Systematic Modernization of Chemical Hazard and Safety Data Management Systems

Contact Info

Product

Resources

About