Understanding User Goals in Web Search University of Seoul Computer Science Database Lab. Min Mi-young.

Understanding User Goals in Web Search

University of SeoulComputer Science Database Lab.

Min Mi-young

23年 4月 18日 Database Lab. 2

Contents

IntroductionRelated WorkA Framework For Search GoalsAssociating Goals With Queries

Manual query classificationResults

Future workConclusions

Introduction

Focus on in previous worksHow people searchWhat they are searching forNot why they are searching

Suggests “navigational” searches User behavior would be the stream of queries users produceStudying these queries and Optimizing the engines based on such as the length of a typical queryPrevent from looking beyond query“Why” is essential to satisfy user’s information need

23年 4月 18日 Database Lab. 3

Introduction

Goal : “Why are you performing that search?” (cf)“ceramics”For choosing a suitable present for a friendLearning which colleges offer adult education courses in potterySeeing if a favorite author’s new book has been released, etc

If we knew user’s goal…?At the very least, engine provide a user experience tailored toward that goal

23年 4月 18日 Database Lab. 4

IntroductionPremise

Web searches reflect a diverse set of underlying user goalsKnowledge of goals offers the prospect of future improvements to web search engines

Involve three tasks1. Need to create a conceptual framework for user goals2. Need a way for search engines to associate user goals

with queries3. Need to modify the engines in order to exploit the goal

informationFocus on first task, & initial parts of the second

Characterizing user search goals& examining the problem of inferring goals from query behavior

23年 4月 18日 Database Lab. 5

Related Work

Bates [4] look at the different ways in which people performed searches

Later proposed ways to characterize the overall search process [5]

Belkin’s Anomalous States of Knowledge (ASK) framework

Attempt to model the cognitive state of the userAnd then translate this understanding into a practical design for an information retrieval system [6]An analysis of some of the different types of information needs of different users

23年 4月 18日 Database Lab. 6

Related Work

Silverstein et alConducted an analysis of query logs from the AltaVista search engineConfirming some of the original findings of web search use, such as the predominance of very short queries [11]

Ongoing research of Spink and her colleaguesAnalyzed query logs of the Excite search engine from 1997, 1999, and 2991 [13]

23年 4月 18日 Database Lab. 7

Related Work

Prior to the advent of the worldwide web, search engine designers

Assume that users had an “informational” goal in mind

Goals underlying web searches are many and variedAbove these helped us to understand ‘what’ users are searching for and ‘how ‘their information-seeking process works

23年 4月 18日 Database Lab. 8

Related Work

“Broder’s taxomony of web search” [7]Idea

Traditional notion of an “information need” might not adequately describe web searchingCame up with a trichotomy of web search “types” : navigational, informational, transactional

Navigational : are intended to find a specific web site that the user has in mindInformational : are intended to find information about a topicTransactional : are intended to “perform some web-mediated activity”

23年 4月 18日 Database Lab. 9

A Framework For Search Goals

First taskTo understand the space of use goalsCome up with a framework that could identify and organize a manageable set of canonical goal categories

Goal categories must encompass the majority of actual goals users have in mind when searching

Develop the goal framework1.Looked at a sample of queries from the AltaVista

search engine2.Brainstormed a variety of goal possibilities, based on

our own experiences3.Resulted in a flat list of goals

Served as a basis for an initial goal classification framework

23年 4月 18日 Database Lab. 10

A Framework For Search Goals

4. which used to categorize a sample of 100-200 queries

5. Revised the framework to accommodate the results of the classification test

Categories were modified, (or new add, drop, merge, split)

6. Classify-refine process was repeated three times with a new set of queries

23年 4月 18日 Database Lab. 11

A Framework For Search Goals

Resulting goal frameworkGoals of Hierarchical structure Resemble Broder’s trichotomy, more general “resource” category

23年 4月 18日 Database Lab. 12

A Framework For Search Goals

23年 4月 18日 Database Lab. 13

Associating Goals With Queries

Two waysUser identify the goal explicitly through the use interface

Google’s “I’m feeling lucky”, in which users implicitly identify their goal as “navigate to a specific web site”

System can attempt to infer the goal automatically

Involve automatic classification using statistical or machine learning methods

23年 4月 18日 Database Lab. 14

Manual Query Classification

User goals can be deducted from looking at user behavior available to the search engine

Included in behaviorThe query itselfThe results returned by the search engineResults clicked on by the userFurther searches or other actions by the user

To facilitate task of manual classificationCreated a research tool that provided these four types of information for sets of queries taken from the AltaVista query logs

23年 4月 18日 Database Lab. 15

23年 4月 18日 Database Lab. 16

A screen shot of the classification tool interface

Query : “kelly blue book”

Links which lead to the search results that appear when the query is executed

Six seconds later, user entered a new, more specific query on the same topic

Eight seconds later, clicked www.kbb.comwhich is the home page of the Kelly Blue Book

Manual Query Classification

Query : final fantasy (name of popular computer games)User examined the result list for 36 seconds, then visited web site www.ffonline.com, “an unofficial guide to final fantasy”About a minute later, user returned to original query, then chose a diff site, “Eyes on Final Fantasy” www.eyesonoff.comIndicates that user was not interested in buying the game, but simply information about it

23年 4月 18日 Database Lab. 17

Results

23年 4月 18日 Database Lab. 18

User goals can be deducted from looking at user behavior available to the search engine

Included in behaviorThe query itselfThe results returned by the search engineResults clicked on by the userFurther searches or other actions by the user

To facilitate task of manual classificationCreated a research tool that provided these four types of information for sets of queries taken from the AltaVista query logs

23年 4月 18日 Database Lab. 19

Future Work

Certain limitationsHave no way of knowing conclusively whether the goal we inferred for a query is in fact the user’s actual goal

Combine with user studies, including qualitative data

AltaVista user population may not be representative of search engine users in general

This may account for some of the user behavior we sawExtend research to Yahoo!

23年 4月 18日 Database Lab. 20

Conclusion

Need to take into account more knowledge of user behavior – not how, but why!

Achievement Created a framework for understanding the underlying goals of searchDemonstrated the framework can be used to associate goals with queries given limited information

ConclusionNavigational queries appear to much less prevalent than generally believedMany queries appear to be motivated by a unexplored goal involving the need to obtain online and offline resources

23年 4月 18日 Database Lab. 21