{"doi":"10.17760/d20467219","title":"What does the evidence say? Automatic information extraction from medical trial reports","abstract":"What does WebMD say you should do when you type in your symptoms? What would your primary care physician say? How do medical practitioners determine what the best intervention is for a given patient? Part of effective medicine derives from the intuition that clinicians develop with years of experience, and part derives from understanding the information available in evidence from external research. Unfortunately, the standard avenue for disseminating new results from clinical trials is in free-text articles published in medical journals. The sheer volume of published articles imposes an enormous barrier to collecting, parsing, and aggregating the relevant results to provide an evidence-based answer to even a single clinical question. Organizations such as the Cochrane Collaborate are dedicated to producing such systematic reviews, but it still requires a team of trained professionals and one to two years in order to finish a single review.Anywhere there is an overwhelming amount of data we need to process, there is an opportunity for automation. To help reduce the manual burden of extracting recommendations from the totality of the available evidence, we turn to Natural Language Processing methods in order to produce structured representations of otherwise unstructured trial reports. We aim to identify all of the relevant information needed to search for and retrieve relevant articles, and characterize the relationships between them in order to summarize the key results and findings in each article. To overcome the difficulties of this highly technical and idiosyncratic domain we develop new corpora, design novel model architectures, and ultimately release our system in order to lower the barrier between primary documents and those who would make use of the information.--Author's abstract","journal":null,"year":2022,"id":314219,"datarank":0.0,"base_score":0.0,"endowment":0.0,"self_citation_contribution":0.0,"citation_network_contribution":0.0,"self_endowment_contribution":0.0,"citer_contribution":0.0,"corpus_percentile":null,"corpus_rank":null,"citation_count":0,"citer_count":0,"citers_with_citation_signal":0,"citers_with_endowment":0,"datacite_reuse_total":0,"is_dataset":false,"is_dataset_confidence":0.9502,"is_data_producer":false,"deposit_databanks":null,"is_oa":true,"file_count":0,"downloads":0,"has_version_chain":false,"published_date":"2022-01-01","fair_score":null,"fair_percentile":null,"algorithm_id":"datarank_citation_only_1hop_v6","ranking_scope":"data_only","authors":[{"id":1013142,"name":"Benjamin D. Nye","orcid":"0000-0002-5902-9196","position":0,"is_corresponding":true}],"reference_count":117,"raw_metadata":null,"created_at":"2026-07-19T00:33:52.047924Z","pmid":null,"pmcid":null,"fwci":null,"citation_percentile":null,"influential_citations":0,"oa_status":null,"license":null,"views":0,"total_file_size_bytes":0,"version_count":0,"fair_f":null,"fair_a":null,"fair_i":null,"fair_r":null,"fair_zscore":null,"fair_rationale":null,"fair_model":null,"fair_agent_version":null,"fair_fulltext_source":null,"fair_has_llm":null,"fair_computed_at":null,"clinical_trials":[],"software_tools":[],"db_accessions":[],"linked_datasets":[],"topics":[]}