《网络搜索和挖掘关键技术 Web Search and Mining》课程教学资源（PPT讲稿）Lecture 10 Query expansion

Relevance Feedback and Query Expansion Web Search and Mining Lecture 10: Query expansion

Relevance Feedback and Query Expansion Lecture 10: Query expansion Web Search and Mining 1

Relevance Feedback and Query Expansion Recap of the last lecture Evaluating a search engine Benchmarks Precision and recall Results summaries

Relevance Feedback and Query Expansion Recap of the last lecture ▪ Evaluating a search engine ▪ Benchmarks ▪ Precision and recall ▪ Results summaries 2

Relevance Feedback and Query Expansion Recap: Unranked retrieval evaluation Precision and recall Precision fraction of retrieved docs that are relevant P(relevant retrieved Recall fraction of relevant docs that are retrieved P(retrieved relevant Relevant Nonrelevant Retrieved Not Retrieved fn Precision P= tp/tp fp) Recall r=tp/tp+ fn)

Relevance Feedback and Query Expansion 3 Recap: Unranked retrieval evaluation: Precision and Recall ▪ Precision: fraction of retrieved docs that are relevant = P(relevant|retrieved) ▪ Recall: fraction of relevant docs that are retrieved = P(retrieved|relevant) ▪ Precision P = tp/(tp + fp) ▪ Recall R = tp/(tp + fn) Relevant Nonrelevant Retrieved tp fp Not Retrieved fn tn

Relevance Feedback and Query Expansion Recap: A combined measure: F Combined measure that assesses precision/recall tradeoff is F measure weighted harmonic mean F (B2+1)PR a+(1-a) BP+R R People usually use balanced F, measure e,withβ=1orω= Harmonic mean is a conservative average

Relevance Feedback and Query Expansion 4 Recap: A combined measure: F ▪ Combined measure that assesses precision/recall tradeoff is F measure (weighted harmonic mean): ▪ People usually use balanced F1 measure ▪ i.e., with  = 1 or  = ½ ▪ Harmonic mean is a conservative average P R PR P R F + + = + − = 2 2 ( 1) 1 (1 ) 1 1    

Relevance Feedback and Query Expansion This lecture Improving results For high recall E.g. searching for aircraft doesn't match with plane; nor thermodynamic with heat Options for improving results Local methods Relevance feedback Pseudo relevance feedback Global methods Query expansion thesaurus Automatic thesaurus generation

Relevance Feedback and Query Expansion This lecture ▪ Improving results ▪ For high recall. ▪ E.g., searching for aircraft doesn’t match with plane; nor thermodynamic with heat ▪ Options for improving results… ▪ Local methods ▪ Relevance feedback ▪ Pseudo relevance feedback ▪ Global methods ▪ Query expansion ▪ Thesaurus ▪ Automatic thesaurus generation 5

Relevance Feedback and Query Expansion LOCAL METHOD: RELEVANCE FEEDBACK

Relevance Feedback and Query Expansion LOCAL METHOD: RELEVANCE FEEDBACK 6

Relevance Feedback and Query Expansion Local method: Relevance Feedback Relevance feedback Relevance feedback: user feedback on relevance of docs in initial set of results User issues a(short simple query The user marks some results as relevant or non -relevant The system computes a better representation of the information need based on feedback Relevance feedback can go through one or more iterations Idea: it may be difficult to formulate a good query when you dont know the collection well, so iterate

Relevance Feedback and Query Expansion Relevance Feedback ▪ Relevance feedback: user feedback on relevance of docs in initial set of results ▪ User issues a (short, simple) query ▪ The user marks some results as relevant or non-relevant. ▪ The system computes a better representation of the information need based on feedback. ▪ Relevance feedback can go through one or more iterations. ▪ Idea: it may be difficult to formulate a good query when you don’t know the collection well, so iterate Local Method: Relevance Feedback 7

Relevance Feedback and Query Expansion Local method: Relevance Feedback Relevance feedback We will use ad hoc retrieval to refer to regular retrieval without relevance feedback We now look at some examples of relevance feedback that highlight different aspects

Relevance Feedback and Query Expansion Relevance feedback ▪ We will use ad hoc retrieval to refer to regular retrieval without relevance feedback. ▪ We now look at some examples of relevance feedback that highlight different aspects. Local Method: Relevance Feedback 8

Relevance Feedback and Query Expansion Local method: Relevance Feedback Relevance Feedback: Example Image search engine http://nayana.ece.ucsbedu/imsearch/imsearch.html @ New Page 1-Netscape file Edit View Go Bookmarks Tools Window Help 9③⑤ httpinayana.eceucsbeduiv A Home Browsing and Shopping related 607,000 images are indexed and classified in the database Only One keyword is allowed! Search Designed by Baris Sumengen and Shawn Newsam powered by jLAMP2000 ava, Linux, Apache, Msgi, Perl, Windows2000)

Relevance Feedback and Query Expansion Relevance Feedback: Example ▪ Image search engine http://nayana.ece.ucsb.edu/imsearch/imsearch.html Local Method: Relevance Feedback 9

Relevance Feedback and Query Expansion Local method: Relevance Feedback Results for Initial Quel Browse Search Prev Next Random 测 (144473,16458) (144457,252140) (144456,262857 144456,262863) 144457,252134) (144483,265154) 0.0 00 0.0 00 0.0 0.0 00 0.0 00 00 0.0 00 0.0 00 00 0.0 44483.264644 (144483,265153) (144518,257752) (144538,525937) 4456,249611)(14446,250049 0.0 00 0.0 0.0 0.0 00 00 0.0 0.0 00 0.0 0.0

Relevance Feedback and Query Expansion Results for Initial Query Local Method: Relevance Feedback 10