abachaa / medquad Goto Github PK
View Code? Open in Web Editor NEWMedical Question Answering Dataset of 47,457 QA pairs created from 12 NIH websites
License: Other
Medical Question Answering Dataset of 47,457 QA pairs created from 12 NIH websites
License: Other
------------------------------------------- MedQuAD: Medical Question Answering Dataset ------------------------------------------- MedQuAD includes 47,457 medical question-answer pairs created from 12 NIH websites (e.g. cancer.gov, niddk.nih.gov, GARD, MedlinePlus Health Topics). The collection covers 37 question types (e.g. Treatment, Diagnosis, Side Effects) associated with diseases, drugs and other medical entities such as tests. We included additional annotations in the XML files, that could be used for diverse IR and NLP tasks, such as the question type, the question focus, its syonyms, its UMLS Concept Unique Identifier (CUI) and Semantic Type. We added the category of the question focus (Disease, Drug or Other) in the 4 MedlinePlus collections. All other collections are about diseases. The paper cited below describes the collection, the construction method as well as its use and evaluation within a medical question answering system. N.B. We removed the answers from 3 subsets to respect the MedlinePlus copyright (https://medlineplus.gov/copyright.html): (1) A.D.A.M. Medical Encyclopedia, (2) MedlinePlus Drug information, and (3) MedlinePlus Herbal medicine and supplement information. -- We kept all the other information including the URLs in case you want to crawl the answers. Please contact me if you have any questions. ------------------- QA Test Collection ------------------- We used the test questions of the TREC-2017 LiveQA medical task: https://github.com/abachaa/LiveQA_MedicalTask_TREC2017/tree/master/TestDataset. As described in our BMC paper, we have manually judged the answers retrieved by the IR and QA systems from the MedQuAD collection. We used the same judgment scores as the LiveQA Track: 1-Incorrect, 2-Related, 3-Incomplete, and 4-Excellent. -- Format of the qrels file: Question_ID judgment Answer_ID The QA test collection contains 2,479 judged answers that can be used to evaluate the performance of IR & QA systems on the LiveQA-Med test questions: https://github.com/abachaa/MedQuAD/blob/master/QA-TestSet-LiveQA-Med-Qrels-2479-Answers.zip ---------- Reference ---------- If you use the MedQuAD dataset and/or the collection of 2,479 judged answers, please cite the following paper: "A Question-Entailment Approach to Question Answering". Asma Ben Abacha and Dina Demner-Fushman. BMC Bioinformatics, 2019. @ARTICLE{BenAbacha-BMC-2019, author = {Asma {Ben Abacha} and Dina Demner{-}Fushman}, title = {A Question-Entailment Approach to Question Answering}, journal = {{BMC} Bioinform.}, volume = {20}, number = {1}, pages = {511:1--511:23}, year = {2019}, url = {https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-019-3119-4} } License - The MedQuAD dataset is published under a Creative Commons Attribution 4.0 International Licence (CC BY). https://creativecommons.org/licenses/by/4.0/ Contact - Asma Ben abacha (abenabacha at microsoft dot com)
Hi,
Thank you for releasing this amazing dataset which is a great contribution to ML in healthcare! I wonder if you have any future plan to release the knowledge source of all those question answer pairs as well to make it a great reading comprehension benchmark like SQuAD. I noticed that many questions can be answered using the webpages whose urls are mentioned in your dataset. I wonder if you are planning to clean up those webpages and extract relevant passages to support the reading comprehension task.
Thanks!
I see that we only have questions pairs but there is no real method to crawl the answer from MedLineDrugs. Is it crawlable? Thank you so much!
Hi Abachaa,
I don't see a common license as part of this github repo. Are you planning to add a license to this github to make it easier to use the datasets?
Feel free to check this repo for scraping (most of) the answers: https://github.com/glicerico/medquad-scraper
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.