TY - JOUR
T1 - Inductive regression tree and genetic programming techniques for learning user Web search preferences
AU - Pendharkar, Parag C.
N1 - Funding Information:
A search engine’s capability to allow a user to incorporate his or her preferences in the search plays an important role. Because Internet users come from different backgrounds, cultures, and ages, the utility of documents based on seemingly similar query terms is different for different users. For example, a researcher planning to apply for a National Science Foundation (NSF) grant may enter the abbreviation “NSF” in the search engine. The results of the Web search, in addition to the NSF Web site, may show pages related to the National Sanitation Foundation and the National Schizophrenia Fellowship. Although the user may click the NSF Web site, the results of his or her search on NSF will not change in the near future (provided that the user uses same search engine). Ideally, it is desirable to have an intelligent agent that learns from past user behavior. The need to incorporate user preferences has paramount importance when it is desired to protect the user from objectionable material. Several content filtering applications currently exist. Among these filtering applications are Net Nanny®, Cyber Patrol®, and Cybersitter™.
PY - 2006
Y1 - 2006
N2 - Most Web search engines determine the relevancy of Web pages based on query terms, and a few content filtering applications allow consumers to block objectionable material. However, not many Web search engines and content filtering applications learn the user preferences over time. In this study, we proposed two machine-learning approaches that can be used to learn consumer preferences to identify documents that are most relevant to the consumer. We test the proposed machine learning approaches on a few simulated data sets. The results of our study illustrate that data mining approaches can be used to design intelligent adaptive agents that can select the relevant Web pages, given query terms, for the user.
AB - Most Web search engines determine the relevancy of Web pages based on query terms, and a few content filtering applications allow consumers to block objectionable material. However, not many Web search engines and content filtering applications learn the user preferences over time. In this study, we proposed two machine-learning approaches that can be used to learn consumer preferences to identify documents that are most relevant to the consumer. We test the proposed machine learning approaches on a few simulated data sets. The results of our study illustrate that data mining approaches can be used to design intelligent adaptive agents that can select the relevant Web pages, given query terms, for the user.
UR - http://www.scopus.com/inward/record.url?scp=33846012171&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33846012171&partnerID=8YFLogxK
U2 - 10.1207/s15327744joce1603&4_4
DO - 10.1207/s15327744joce1603&4_4
M3 - Article
AN - SCOPUS:33846012171
SN - 1091-9392
VL - 16
SP - 223
EP - 245
JO - Journal of Organizational Computing and Electronic Commerce
JF - Journal of Organizational Computing and Electronic Commerce
IS - 3-4
ER -