sigir00 poster

Uploaded from authorPOINTLite
Views:
 
Category: Entertainment
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Auto-Construction of a Live Thesaurus from Search Term Logs for Interactive Web Search: 

Auto-Construction of a Live Thesaurus from Search Term Logs for Interactive Web Search

Our Thesaurus: 

Our Thesaurus Travel Recreation Trip Tour Hiking Backpacking Camping Broader Term Synonym Similar Terms Relevant Terms Travel Agency Air Flight VISA Lodging Passport Rental Car Travel Computer A Subject Taxonomy

Interactive Web Search: 

Interactive Web Search Short Query on Web Search The Approach:Interactive Term Suggestion

Research Progress: 

Research Progress Integration of Human & Machine Categorized Core Terms Subject Taxonomy Search Term Log Categorized Search Terms Manual Categorization of Core Terms Structuring of Subject Taxonomies Live Thesaurus

Existence of Core Terms: 

Existence of Core Terms Coverage comparison between D-1998 and G-1999 (by the number of distinct queries) Some “core” search terms exist without effects of time D-1998: Search log of Dreamer within 3 months in 1998 G-1999: Search log of GAIS within 2 weeks in 1999

New Term Categorization: 

New Term Categorization Take into account of new term t’s co-occurrence terms’ categories Wtc : t’s co-occurrence terms in V with category c fwDt : frequency of w in Dt Ranking score of category c: ..t w.. ..w..t t...t.w.. .w…t.. .t..w..w t..w.. ..t.. ..t.. ..t.. ..t.. Dt Dw V : all categorized core terms The whole document set

Preliminary Experiment on New Term Categorization: 

Preliminary Experiment on New Term Categorization Take a real search engine as the whole document set Experimental Result SE term t Dt ..w.. ..w.. ..w..w.. ….. ….. …..

Obtained Performance on Observing Information Needs: 

Obtained Performance on Observing Information Needs

Possible Applications of Live Thesaurus: 

Possible Applications of Live Thesaurus Web Infometrics Web Page/Site Classification Term Translation Extraction Interactive Search via Voice Relevant Term Extraction Live Thesaurus

Live Thesaurus ?: 

Live Thesaurus ? Thesaurus A set of terms plus a set of relations on subject relevance between them Live Thesaurus A thesaurus automatically constructed from online resources such as web pages and search logs Log-based Live Thesaurus A live thesaurus constructed from search term logs Travel A Query Term Net Recreation Trip Tour Hiking Backpacking Camping Broader Term Synonym Similar Terms Relevant Terms Travel Agency Air Flight VISA Lodging Passport Rental Car

Interactive Web Search: 

Interactive Web Search Short Query on Web Search 2.35 words (Altavista, 1998) 3.55 Chinese characters (Our observation, 1999) Not enough information for search Interactive Term Suggestion SE Query Terms

The Approach: 

The Approach The overall design of the on-going research

Research Progress: 

Research Progress Integration of Human & Machine Categorized Core Terms Subject Taxonomy Search Term Log Categorized Search Terms Manual Categorization of Core Terms Structuring of Subject Taxonomies Live Thesaurus

Existence of Core Terms: 

Existence of Core Terms Coverage comparison between D-1998 and G-1999 (by the number of distinct queries) Some “core” search terms exist without effects of time D-1998: Search log of Dreamer within 3 months in 1998 G-1999: Search log of GAIS within 2 weeks in 1999

New Term Categorization: 

New Term Categorization Categorize a new term t, given categorized core terms V as feature set Experimental Result SE term t Dt Dw ..w.. ..w.. ..w..w.. ….. ….. ….. Wtc : t’s co-occurrence terms in V with category c fwDt : frequency of w in Dt Ranking score of category c:

Obtained Performance on Observing Information Needs: 

Obtained Performance on Observing Information Needs Comparison between the distributions of manually and auto. categorized core terms

Possible Applications of Live Thesaurus: 

Possible Applications of Live Thesaurus Web Infometrics Web Page/Site Classification Term Translation Extraction Interactive Search via Voice Relevant Term Extraction Live Thesaurus