1 / 16

Developing Recommendation Techniques for Scholarly Papers

Developing Recommendation Techniques for Scholarly Papers. Kazunari Sugiyama. National University of Singapore. Previous Research Topics. Web Information Retrieval (@NAIST) How to Characterize Web page User Adaptive Information Retrieval Disambiguation (@TITECH)

chavez
Download Presentation

Developing Recommendation Techniques for Scholarly Papers

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. Developing Recommendation Techniques for Scholarly Papers Kazunari Sugiyama National University of Singapore

  2. Previous Research Topics • Web Information Retrieval (@NAIST) • How to Characterize Web page • User Adaptive Information Retrieval • Disambiguation (@TITECH) • Personal Name Disambiguation in Web Search Results • Word Sense Disambiguation in Japanese Texts

  3. Scholarly Paper Recommendation (@NUS) • Junior researchers Only one recently published paper without citations • Senior researchers Multiple published papers with citation papers

  4. User Profile Construction (Junior Researchers) Weighting scheme Cosine similarity

  5. User Profile Construction (Senior Researchers) Forgetting factor Weighting scheme Cosine similarity

  6. Feature Vector Construction for Candidate Papers References Weighting scheme Cosine similarity • Basically, TF-IDF • Also use information about citation and reference papers

  7. Is Pruning of Citation and Reference Papers Effective? sim:0.32 sim:0.27 sim:0.42 sim:0.25 sim:0.13 Threshold: 0.3 References sim:0.18 sim:0.58 sim:0.22 sim:0.36 sim:0.45

  8. Is Pruning of Citation and Reference Papers Effective? sim:0.33 sim:0.27 sim:0.42 sim:0.25 sim:0.13 Threshold: 0.3 References Weighting scheme Cosine similarity sim:0.18 sim:0.58 sim:0.22 sim:0.36 sim:0.45

  9. Experiments • Experimental Data • Researchers • 15 junior researchers • 13 senior researchers • NLP and IR researchers who have publication lists • in DBLP • Candidate Papers to Recommend • ACL Anthology Reference Corpus • Includes information about citation and reference papers

  10. Junior Researchers [NDCG@5] Pruning is effective! The most recent paper with pruning its reference papers

  11. Senior Researchers [NDCG@5] When and are small, FF is effective! Past published papers with forgetting factor

  12. Extensions Characterize the target paper using potential papers Serendipitous recommendation

  13. Characterize the target paper using potential papers (‘06) (‘07) (‘09) (‘05) Potential paper that should cite the target paper

  14. Finding potential papers with collaborative filtering Pi (i=1, … ,N): All papers in the dataset Pcj (j=1, … ,N): Papers as citation papers in the dataset 0.368 0.536 0.211 0.472

  15. Characterize the target paper using potential papers (‘06) (‘07) (‘09) (‘05) Potential paper that should cite the target paper

  16. Serendipitous Recommendation User profile generated from history of contents User profile for serendipitous recommendation User 1 User 4 (Sim: 0.16) Weight: 1/(0.16+1) User 10 (Sim: 0.26) Weight: 1/(0.26+1) User profile for serendipitous recommendation User 1 (Sim: 0.32) Weight: 1/(0.32+1) User 2 User 5 (Sim: 0.21) Weight: 1/(0.21+1) User profile for serendipitous recommendation User 1 (Sim: 0.14) Weight: 1/(0.14+1) User 7 (Sim: 0.25) Weight: 1/(0.25+1) User 3 User profile for serendipitous recommendation User 6 (Sim: 0.07) Weight: 1/(0.07+1) User 2 (Sim: 0.12) Weight: 1/(0.12+1) User n

More Related