1 / 15

eEvidence : Information Seeking Support for Evidence-based Practice: An Implementation Case Study

eEvidence : Information Seeking Support for Evidence-based Practice: An Implementation Case Study. Jin Zhao, Min-Yen Kan, Paula M. Procter, Siti Zubaidah , Wai Kin Yip, Goh Mien Li. Evidence-based Practice (EBP). Advantages Reliable, extensive and updated. Intervention A.

arlen
Download Presentation

eEvidence : Information Seeking Support for Evidence-based Practice: An Implementation Case Study

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. eEvidence: Information Seeking Support for Evidence-based Practice:An Implementation Case Study Jin Zhao, Min-Yen Kan, Paula M. Procter, SitiZubaidah, Wai Kin Yip, Goh Mien Li

  2. Evidence-based Practice (EBP) • Advantages • Reliable, extensive and updated Intervention A ResearchArticles Intuition Intervention B Experience Intervention C WING, NUS

  3. Two Stages of EBP Crucial yet difficult! • Stage 1: Search and Appraise • Stage 2: Apply and Evaluate Article Collections ResearchArticles Interventions Interventions WING, NUS

  4. Issues in the Search & Appraise Stage • Accessibility problems • Collection-specific specialized search engines • Subscription barrier • Usability gap in Search Engines • Key information not displayed or searchable • Elements of well-form clinical questions • E.g., demographics of the patient, intervention and results • Strength of evidence • E.g., study design • Different search patterns • Active v.s. passive WING, NUS

  5. eEvidence System for EBP • Key features • Harvesting EBP resources by periodic crawling • Address accessibility issue and ensures up-to-date coverage • Automated article classification and key information extraction • Provides crucial information to assist search and appraisal • Dual active/passive user interface • Caters for different search patterns WING, NUS

  6. System Architecture EBPresources WING, NUS

  7. Harvesting EBP Resources by Periodic Crawling • Two stages • Selection of EBP resources by experts • Periodic crawling using Nutch • Advantages • Able to harvest from any type of EBP resources • Always up-to-date WING, NUS

  8. Automated Article Classification • Supervised machine learning pipeline • Three categories: full text / abstract / others • Maximum entropy classifier • Text, webpage and formatting features • Advantages • Filters out useless webpages • Allows users to zoom to subsets of articles WING, NUS

  9. Automated Key Information Extraction • Year of publication, article type, time added, URL • Key sentence / keywords • Advantages • Provides crucial information to assist search and appraisal WING, NUS

  10. Extraction of Key Sentences and Keywords • Cast as two-stage pipelined soft-classification Sex Keywords Patient Sentences Race Keywords Intervention Sentences Age Keywords Keyword classification Sentence Classification Result Sentences Sentences Condition Keywords Research Goal Sentences Intervention Keywords Study Design Sentences Study Design Keywords WING, NUS

  11. Extraction of Key Sentences and Keywords • 1-against-all MaxEnt classifiers for each class based on various text features • Key Sentences Extraction • Precise but much room for improvement in recall • Due to the variety of sentence structure and the shortness of information • Acceptable in the context of Information Retrieval • Keywords Extraction • Good performance in general • Difficult to extract keywords belonging to classes with a diverse vocabulary or lexical variations Performance for key sentences extraction Performance for keywords extraction WING, NUS

  12. Dual Interface (Read) • Recommend relevant articles based on user profile WING, NUS

  13. Dual Interface (Search) • Allow users to search with complex query and filters • Search history in place of profile WING, NUS

  14. Discussion • Iterative development methodology • Interview users for requirements • Design and implement features • Gather user feedbacks of the system • Current feedbacks • Able to find freely available full text articles • Search history useful for writing of search methodology • Collection size still limited for practical task • Future work • Improve key information extraction • Extend collection and perform full-fledge evaluation • Discover user- and group- specific features WING, NUS

  15. Conclusion • eEvidence: a literature retrieval system with novel features to facilitate EBP • Harvesting EBP resources by periodic crawling • Automated article classification and key information extraction • Dual active/passive user interface WING, NUS

More Related