210 likes | 334 Views
A case for Intent Motivated Search Technique. Presented By: Raymond Gao Wednesday, January 02, 2008. Presentation Roadmap. Why is relevant Search experience difficult to provide? The evolution of Search Engine My demo (validating Intent Motivated Search Technique).
E N D
A case for Intent Motivated Search Technique Presented By: Raymond Gao Wednesday, January 02, 2008
Presentation Roadmap • Why is relevant Search experience difficult to provide? • The evolution of Search Engine • My demo (validating Intent Motivated Search Technique). • Discussing my application – Technical Aspect.
Opening Question • What is a common feature provided by website A-Z, from Amazon, to Ebay, to Google, to Zillow.com? • Search is everywhere
An illustration of un-intended search results? • Query 1a - Where to buy an Amazon book? • Top results are on Amazon.com (Not what I intended. I want books about the Amazon Jungle.)
Illustration #2 • Query 2a - Where to buy Mozart Kugel in Salzburg? • Results are on Cholocate. (Not my 1st choice. I want to find a store in Salzburg to buy cheap Mozart Kugel.)
What did I just show? • When a word has multiple meanings, the Search Engine does not know what is intended context. Thus, it fails to return the correct result.
AI & Beyond Search by Reference Search by Text Search by Meta-Data Personalized Search The Evolution of Search Engine Intent Motivated Search Relevance of the result Complexity of the Search Engine
The Search Evolution • Stage I – Search by Matching Text – Altavista, Excite, Yahoo (pre 2000) • Stage II – Search by Reference – Google, Ask.com • Stage III – Search by Meta-data – Metacrawler, Ice Rocket • Stage IV – Personalized Search – iGoogle, • Stage V – Intent Motivated Search– Mindset.yahoo.com • Stage VI - AISearch Engine & beyond
What, How, Which • What I have built? • I have built a search engine front-end that validates the concept of Intent Motivated Search Experience • How is it built? • Aqualogic Portal’s Search Engine • Rule Engine • AJAX and custom code • Which data sources? • See next page
Data Sources • Regular files, • Web pages, • Other Portals – M.S. SharePoint, • Group Wares - Lotus Notes, • Other Content Management Systems (Documentum, Vignette, …)
Intent Motivated Search (Demo) • User Profiling (Rule drive) • + Meta-Tagging • + User Dictionary • = A Prototype for Intent Motivated Search Experience • User’s search query is affected by his/her group membership, previous experience, &many other factor. • Proposed Solution: an search engine that understands the user’s vocabulary. • Extend the reach of meta-tags, user custom dictionary, with the aid of user profiling.
Business Logic • Student has little money, Wikipedia is free • Teacher wants accurate content, & library books form the basis for researched • Amazon.com Reward Member has $50 credit. He prefers to redeem the credit for free items. • (Default) General public has a $25 spending limit. Which source does not matter.
Query Process Flow Weighted Query Profiling Interpret Student Teacher Decision Profile Gift Card Holder General Public
The Benefit of Applying Rule Engine • Business Rule are stored in XML format -> facilitating easy editing; • The Rule Engine can be dynamically executed and used to weight the returned search result; • Business Rule can be readjusted independent of the search indexing engine.
Intent & Action • Distinction between Spoken vs. Unspoken intent. • Spoken Intent is from the original query string. • Unspoken Intent maybe inferred from the user attributes, e.g. group membership, profession, age, gender, … • If 80% of Actions are caused by an Intent, by capturing user’s Intention, the Search Experience could improve.
Reference • http://www.seomoz.org/blog/segmenting-search-intent • http://Igoogle.com • http://mindset.yahoo.com • http://www.marketingvox.com/archives/2005/05/31/yahoo_mindset_search_results_reflect_searcher_intent/
My Contact Information Email address: raygao@verizon.net rgao@bea.com Mobile Phone: 972-765-4423