A Taxonomy of Web Searches Andrei Broder SIGIR
A Taxonomy of Web Searches Andrei Broder, SIGIR Forum, 2002 Ahmet Yenicag Ceyhun Karbeyaz
Abstract Classic IR is inherently predicated on users search for their information need, that leads them to use an IR system. But the need behind a web search is not only informational. Navigational Transactional In this paper: Analyze the taxonomy of web searches Present some statistics about this taxonomy Show evolution of search engines in light of this taxonomy. 2/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
Taxonomy of Web Searches Classify web queries according to their intent: Navigational: Give me the URL of the site I want to reach. Bilkent => www. bilkent. edu. tr Informational: Find information assumed to be available on the web and no further interaction is predicted except reading. Closest to classical IR What is IR? => IR is … Transactional: Show me sites where I can perform a certain transaction. Shopping, downloading media files (mp 3, video) 3/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
Statistics Prevalence of queries are determined by two methods: - User survey - Alta. Vista query log analysis 4/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
Statistics (continued) User survey: Self selection Most of the queries are non-navigational Hard to distinguish between transactional and informational queries Queries that are neither transactional, nor navigational are assumed to be informational Estimation: Transactional queries are around 36% 5/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
Statistics (continued) Log analysis: Alta. Vista log is analyzed for 1000 random queries. Only English queries are taken into consideration: 400 queries Figure 4. Query Classification 6/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
Evolution of Search Engines Three stages: First generation: uses mostly on-page data and very close to classic IR. (informational) Alta. Vista Second generation: uses off-page, web specific data such as link analysis. (informational and navigational) Google Third generation: Recently emerging, attempts to blend data from multiple sources to answer “the need behind the query”. (informational, transactional and navigational) Ankara => direct links to a hotel reservation page for Ankara 7/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
Conclusion The need behind a query might be: informational, navigational or transactional. Search engines need to deal with all of them. Understanding taxonomy is important for successful web search development. Current engines lack of supporting transactional queries directly. Third generation engines will solve this problem mostly via semantic analyses (understanding what the query is about). 8/8 CS 533, Information Retrieval, Karbeyaz & Yenicag 2/25/2021
- Slides: 8