Presentation is loading. Please wait.

Presentation is loading. Please wait.

Discovering Key Concepts in Verbose Queries Michael Bendersky and W. Bruce Croft University of Massachusetts SIGIR 2008.

Similar presentations


Presentation on theme: "Discovering Key Concepts in Verbose Queries Michael Bendersky and W. Bruce Croft University of Massachusetts SIGIR 2008."— Presentation transcript:

1 Discovering Key Concepts in Verbose Queries Michael Bendersky and W. Bruce Croft University of Massachusetts SIGIR 2008

2 Objective “Discovering Key Concepts in Verbose Queries”

3 Objective “Discovering Key Concepts in Verbose Queries” Number 829 Spanish Civil War support Provide information on all kinds of material international support provided to either side in the Spanish Civil War

4 Objective “Discovering Key Concepts in Verbose Queries” Number 829 Spanish Civil War support Provide information on all kinds of material international support provided to either side in the Spanish Civil War

5 Objective “Discovering Key Concepts in Verbose Queries” Use of key concepts?

6 Objective “Discovering Key Concepts in Verbose Queries” Use of key concepts? Combine with current IR model

7 Retrieval Model Conventional Language Model: score(q,d) = p(q|d) =

8 Retrieval Model Conventional Language Model: score(q,d) = p(q|d) = New Model: score(q,d) = p(q|d) = =

9 Final Retrieval Function score(q,d) =

10 Final Retrieval Function score(q,d) = Language Model

11 Final Retrieval Function score(q,d) = Key Concepts

12 What is a Concept? Noun phrase in a query

13 What is a Concept? Noun phrase in a query Number 829 Spanish Civil War support Provide information on all kinds of material international support provided to either side in the Spanish Civil War

14 What is a Concept? Noun phrase in a query Number 829 Spanish Civil War support Provide information on all kinds of material international support provided to either side in the Spanish Civil War

15 Finding ‘Key’ Concepts Rank concepts by p(c i |q)

16 Finding ‘Key’ Concepts Rank concepts by p(c i |q) Compute p(c i |q) by frequency? Number 829 Spanish Civil War support Provide information on all kinds of material international support provided to either side in the Spanish Civil War

17 Finding ‘Key’ Concepts Approximate p(c i |q) by machine learning h(c i ) is c i ’s query-independent importance score p(c i |q) = h(c i ) /  ci  q h(c i ) cici AdaBoost.M1 h(ci)h(ci)

18 Features of a Concept is_cap : is capitalized tf : in corpus idf : in corpus ridf : idf modified by Poisson model wig : weighted information gain; change in entropy from corpus to retrieved data g_tf : Google term frequency qp : number of times the concept appears as a part of a query in MSN Live qe : number of times the concept appears as exact query in MSN Live

19 TREC Corpus

20 Exp 1: Identifying Key Concept Cross-validation on corpus Each fold has 50 queries Check whether the top concept is a key concept Assume 1 key concept per query during annotation

21 Exp 1: Identifying Key Concept

22 Better than idf ranking

23 Exp 2: Information Retrieval score(q,d) = Use only the top 2 concepts for each query q is the entire section = 0.8

24 Exp 2: Information Retrieval KeyConcept[2] : author’s method SeqDep : include all bigrams in query

25 Exp 2: Information Retrieval

26 What to take home? Singling out key concepts improves retrieval


Download ppt "Discovering Key Concepts in Verbose Queries Michael Bendersky and W. Bruce Croft University of Massachusetts SIGIR 2008."

Similar presentations


Ads by Google