1 / 15

Ranked Queries over sources with Boolean Query Interfaces without Ranking Support

Ranked Queries over sources with Boolean Query Interfaces without Ranking Support. Vagelis Hristidis , Florida International University Yuheng Hu , Arizona State University Panos Ipeirotis , New York University. Motivation: PubMed (and USPTO, and Linked In, and…).

michi
Télécharger la présentation

Ranked Queries over sources with Boolean Query Interfaces without Ranking Support

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. Ranked Queries over sources with Boolean Query Interfaces without Ranking Support Vagelis Hristidis, Florida International UniversityYuhengHu, Arizona State UniversityPanosIpeirotis, New York University

  2. Motivation: PubMed(and USPTO, and Linked In, and…) • PubMed offers only ranking by date, author, title, or journal • Usually, user like ranking by relevance • Measured by IR ranking function, like tf-idf

  3. Problem Definition • Input • Query Q contains term t1,…tn • Database D contain documents d1,…,dm • Output • Top-kdocuments ranked according to a relevance score function • Example of ranking function: tf.idf • Baseline: Submit a disjunctive query with all query keywords, retrieve all the documents, locally re-rank • Problems with Baseline method: Too many results! • “immunodeficiency virus structure”  1,451,446 results

  4. Query Relaxation Approach • A tf.idf query has OR semantics • Using queries will AND semantics returns promising documents earlier on • Gradual query relaxation allows fast execution • Key questions: • Which (conjunctive) queries to execute? • When to stop?

  5. Problem Setting and Challenges • Boolean query interface, (e.gPubMed) • Limited data access through web service (quota per day) • No useful ranking functions • No indices to rely on • No statistics exported from database

  6. idf, (easy part) tf, (challenging part) tf parameter of Poissonfor the term in database Probabilistic Approach • Document Score • Estimate tf (and scores) probabilistically: • The tf of the terms in a database tend to follow a Poisson distribution • Document scores also follow a Poisson

  7. The k-th highest score so far Query candidate Probabilistic top-k with query relaxation • Querying strategy – How to pick a good query candidate? • A good query should have good “benefit” • Benefit: Probability that document in results for relaxed query qin top-k. Pr{ScoreQ(D,q) > τ} Score follows Poisson, function of the λ parameters of query terms in Q We choose the query candidate q with maximum probability

  8. Estimation of Poisson Parameters • Sample-based estimation: Fetch documents, construct sample, use estimates from sample • Need very extensive sampling size for reliable estimates • Query-based estimation: Combine sampling and query execution • Every query generates a sample and provides candidate top-k docs • Main challenge: Adjust estimates to compensate for querying bias(we are looking for top-k documents, we do not perform random sampling)

  9. Query-based Sampling • Document sample returned for each query is not random! • Sample is “conditional” on query terms (guaranteed to appear) • Need to acknowledge in estimates that queries are trying to find the top-k, not intended for random sampling • Without correction, estimates significantly off

  10. Top-kalgorithm using query relaxation • Send conjunctive query to the database with all terms • Update statistics for each termusing estimates from the biased sample • Compute benefits for each possible query relaxation • If benefit (i.e., probability of finding top-k document) belowthreshold, stop; else go to step 1

  11. Experiments • Datasets • PubMed • TREC • Quality Measure • Spearman’s Footrule • Algorithms • Baseline • Summary-based • Query-based

  12. Experiments: Quality • Compared footrule distance compared to baseline (baseline = retrieve everything, fetch locally, rerank) • Lower values better • Query-based sampling consistently better than alternatives

  13. Experiments: Efficiency • Measured #documents, queries, and execution time of alternative techniques

  14. Conclusion • Technique for top-k queries on top of document databases without ranking support • Introduction of an exploration-exploitation framework for building necessary statistics on-the-fly, during query execution • Order-of-magnitude efficiency improvements, small losses in quality

  15. Thank you ! Questions?

More Related