Chapter Eleven

Conclusion

As large quantities of data are becoming more accessible via the WWW and P2P file-sharing networks, search is beginning to play a vital role in today’s society. As such, it is important to continue to improve the quality of a user’s search experience by identifying and intelligently exploiting “hidden” information (or signals) in these databases. One such signal is user context information.

This book examined the scalable use of user context information to improve search quality in both the WWW and P2P networks. We have presented scalable algorithms and detailed mathematical analysis for personalized search in both domains.

There are several future challenges for the personalization of search.

User interface design is one of the key challenges in personalized search. Some of the key issues here are non-intrusiveness and consistency of experience. It is important not to have personalization degrade the quality of the user’s search experience, and likewise, it is important that a user’s overall experience maintain a consistency even despite her changing interests. Ideally, personalized search should be as seamless and transparent as unpersonalized search. All of these considerations point to a subtle introduction of personalization into search results.

Another challenge in personalized search is the utilization of implicit measures of user preferences. In the spirit of making the user experience seamless and transparent, a user should not have to make his interests explicit. There are several signals that could be used to indicate a user’s preferences, for example, their search or browsing history. Correctly using these implicit measures of user preferences remains a fertile area for research in personalization and machine learning.

It would also be interesting and useful to explore approaches to personalized search other than link-based methods. Personalization methods that exploit text on a page offer much promise, as do methods that exploit user visit behavior where that information is available.

There are broad privacy implications of personalized search in a world where search plays a central role in daily life. My own opinion is that any implementation of personalization where the search engine collects user data should build in transparency (the search engine should show the user all the data it uses to personalize search results), control (the user should be able to delete data that she doesn’t want used in personalization, and data portability (any user data collected by the search engine should belong to the user, not the search engine, and the user should be able to download that data and bring it to another search engine). An in-depth discussion on privacy and personalization should be the topic of its own book and is outside the scope of this technical work.

There is immense potential for personalized search not just in ranking, but in many other aspects of the search engine, such as user interfaces, corpora selection in federated search, and different models of “push search,” like recommendations and alerts. With Web-enabled mobile phones more common, impoverished input devices will usher in the need for personalization in query formulation and latency reduction. And with the proliferation of social networks, social search will become a more dominant method of search. All of these will require novel approaches, novel algorithms, and novel mathematical underpinnings.

Indeed, the scope of future work in this area is broad, and, in the context of today’s data-driven society, the potential value is tremendous.