24 March 2022 in Informatics in Medicine Unlocked
1 January 2023 in Bioinformatics
Timetrees depict evolutionary relationships between species and the geological times of their divergence. Hundreds of research articles containing timetrees are published in scientific journals every year. The TimeTree project has been manually locating, curating, and synthesizing timetrees from these articles for almost two decades into a TimeTree of Life, delivered through a unique, user-friendly web interface (timetree.org). The manual process of finding articles containing timetrees is becoming increasingly expensive and time-consuming. So, we have explored the effectiveness of text-mining approaches and developed optimizations to find research articles containing timetrees automatically. We have developed an optimized machine learning (ML) system to determine if a research article contains an evolutionary timetree appropriate for inclusion in the TimeTree resource. We found that BERT classification fine-tuned on whole-text articles achieved an F1 score of 0.67, which we increased to 0.88 by text-mining article excerpts surrounding the mentioning of figures. The new method is implemented in the TimeTreeFinder tool, TTF, which automatically processes millions of articles to discover timetree-containing articles. We estimate that the TTF tool would produce twice as many timetree-containing articles as those discovered manually, whose inclusion in the TimeTree database would potentially double the knowledge accessible to a wider community. Manual inspection showed that the precision on out-of-distribution recently-published articles is 87%. This automation will speed up the collection and curation of timetrees with much lower human and time costs. https://github.com/marija-stanojevic/time-tree-classification Supplementary data are available at Bioinformatics online.
26 November 2018 in Open Forum Infectious Diseases
Immunization programs maintain and improve vaccination coverage to prevent diseases. Immunization program text data provide contextual information necessary to better understand vaccine coverage. However, text data analysis can be labor intensive. Cognitive computing systems address this challenge by systematically processing large volumes of text data. Publicly available data were used. Formal data were gathered using scrapers and parsers to extract information from immunization-related websites, journals, and legislation. Informal data were collected via a social media search platform, Sysomos, from Twitter feeds. All data were preprocessed to remove irrelevant text. Existing algorithms analyzed data and retrieved the most closely related words or paragraphs and produced similarity scores for queries. Additionally, Word2vec and Glove algorithms were used to assess similarity and frequency of occurrence between queried and retrieved information. The system searches by query, date, and jurisdiction. A query can range from a single word to a whole document. The system understands similarities between words, sentences, paragraphs, and documents and retrieves text based on similarities to the query. Results are supplemented by similarity scores, dates, jurisdictions, web-links, and usernames (where applicable). Similarity scores allow for quantitative analysis on text data. The pilot cognitive computing system used algorithms to quickly search formal and informal immunization text data, creating a well-rounded system. The formal data can help identify program activities associated with changes in vaccination coverage. The informal data can help assess information being shared through social media during an outbreak or other emergency. The system will stay relevant as long as new data are continuously incorporated to update the algorithms. All authors: No reported disclosures.