Dataset Information

Forty-two Million Ways to Describe Pain: Topic Modeling of 200,000 PubMed Pain-Related Abstracts Using Natural Language Processing and Deep Learning-Based Text Generation.

ABSTRACT:

Objective

Recent efforts to update the definitions and taxonomic structure of concepts related to pain have revealed opportunities to better quantify topics of existing pain research subject areas.

Methods

Here, we apply basic natural language processing (NLP) analyses on a corpus of >200,000 abstracts published on PubMed under the medical subject heading (MeSH) of "pain" to quantify the topics, content, and themes on pain-related research dating back to the 1940s.

Results

The most common stemmed terms included "pain" (601,122 occurrences), "patient" (508,064 occurrences), and "studi-" (208,839 occurrences). Contrarily, terms with the highest term frequency-inverse document frequency included "tmd" (6.21), "qol" (6.01), and "endometriosis" (5.94). Using the vector-embedded model of term definitions available via the "word2vec" technique, the most similar terms to "pain" included "discomfort," "symptom," and "pain-related." For the term "acute," the most similar terms in the word2vec vector space included "nonspecific," "vaso-occlusive," and "subacute"; for the term "chronic," the most similar terms included "persistent," "longstanding," and "long-standing." Topic modeling via Latent Dirichlet analysis identified peak coherence (0.49) at 40 topics. Network analysis of these topic models identified three topics that were outliers from the core cluster, two of which pertained to women's health and obstetrics and were closely connected to one another, yet considered distant from the third outlier pertaining to age. A deep learning-based gated recurrent units abstract generation model successfully synthesized several unique abstracts with varying levels of believability, with special attention and some confusion at lower temperatures to the roles of placebo in randomized controlled trials.

Conclusions

Quantitative NLP models of published abstracts pertaining to pain may point to trends and gaps within pain research communities.

SUBMITTER: Tighe PJ

PROVIDER: S-EPMC7685694 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundThe Enteropathogen Resource Integration Center (ERIC; http://www.ericbrc.org) has a goal of providing bioinformatics support for the scientific community researching enteropathogenic bacteria such as Escherichia coli and Salmonella spp. Rapid and accurate identification of experimental conclusions from the scientific literature is critical to support research in this field. Natural Language Processing (NLP), and in particular Information Extraction (IE) technology, can be a significant aid to this process.DescriptionWe have trained a powerful, state-of-the-art IE technology on a corpus of abstracts from the microbial literature in PubMed to automatically identify and categorize biologically relevant entities and predicative relations. These relations include: Genes/Gene Products and their Roles; Gene Mutations and the resulting Phenotypes; and Organisms and their associated Pathogenicity. Evaluations on blind datasets show an F-measure average of greater than 90% for entities (genes, operons, etc.) and over 70% for relations (gene/gene product to role, etc). This IE capability, combined with text indexing and relational database technologies, constitute the core of our recently deployed text mining application.ConclusionOur Text Mining application is available online on the ERIC website (http://www.ericbrc.org/portal/eric/articles). The information retrieval interface displays a list of recently published enteropathogen literature abstracts, and also provides a search interface to execute custom queries by keyword, date range, etc. Upon selection, processed abstracts and the entities and relations extracted from them are retrieved from a relational database and marked up to highlight the entities and relations. The abstract also provides links from extracted genes and gene products to the ERIC Annotations database, thus providing access to comprehensive genomic annotations and adding value to both the text-mining and annotations systems.

Project description:BackgroundCollaboration among investigators has become critical to scientific research. This includes ad hoc collaboration established through personal contacts as well as formal consortia established by funding agencies. Continued growth in online resources for scientific research and communication has promoted the development of highly networked research communities. Extending these networks globally requires identifying additional investigators in a given domain, profiling their research interests, and collecting current contact information. We present a novel strategy for building investigator networks dynamically and producing detailed investigator profiles using data available in PubMed abstracts.ResultsWe developed a novel strategy to obtain detailed investigator information by automatically parsing the affiliation string in PubMed records. We illustrated the results by using a published literature database in human genome epidemiology (HuGE Pub Lit) as a test case. Our parsing strategy extracted country information from 92.1% of the affiliation strings in a random sample of PubMed records and in 97.0% of HuGE records, with accuracies of 94.0% and 91.0%, respectively. Institution information was parsed from 91.3% of the general PubMed records (accuracy 86.8%) and from 94.2% of HuGE PubMed records (accuracy 87.0). We demonstrated the application of our approach to dynamic creation of investigator networks by creating a prototype information system containing a large database of PubMed abstracts relevant to human genome epidemiology (HuGE Pub Lit), indexed using PubMed medical subject headings converted to Unified Medical Language System concepts. Our method was able to identify 70-90% of the investigators/collaborators in three different human genetics fields; it also successfully identified 9 of 10 genetics investigators within the PREBIC network, an existing preterm birth research network.ConclusionWe successfully created a web-based prototype capable of creating domain-specific investigator networks based on an application that accurately generates detailed investigator profiles from PubMed abstracts combined with robust standard vocabularies. This approach could be used for other biomedical fields to efficiently establish domain-specific investigator networks.

Project description:The Immune Epitope Database (IEDB) project manually curates information from published journal articles that describe immune epitopes derived from a wide variety of organisms and associated with different diseases. In the past, abstracts of scientific articles were retrieved by broad keyword queries of PubMed, and were classified as relevant (curatable) or irrelevant (not curatable) to the scope of the database by a Naïve Bayes classifier. The curatable abstracts were subsequently manually classified into categories corresponding to different disease domains. Over the past four years, we have examined how to further improve this approach in order to enhance classification performance and to reduce the need for manual intervention.Utilizing 89,884 abstracts classified by a domain expert as curatable or uncuratable, we found that a SVM classifier outperformed the previously used Naïve Bayes classifier for curatability predictions with an AUC of 0.899 and 0.854, respectively. Next, using a non-hierarchical and a hierarchical application of SVM classifiers trained on 22,833 curatable abstracts manually classified into three levels of disease specific categories we demonstrated that a hierarchical application of SVM classifiers outperformed non-hierarchical SVM classifiers for categorization. Finally, to optimize the hierarchical SVM classifiers' error profile for the curation process, cost sensitivity functions were developed to avoid serious misclassifications. We tested our design on a benchmark dataset of 1,388 references and achieved an overall category prediction accuracy of 94.4%, 93.9%, and 82.1% at the three levels of categorization, respectively.A hierarchical application of SVM algorithms with cost sensitive output weighting enabled high quality reference classification with few serious misclassifications. This enabled us to significantly reduce the manual component of abstract categorization. Our findings are relevant to other databases that are developing their own document classifier schema and the datasets we make available provide large scale real-life benchmark sets for method developers.