Shyam Kapur

Document preprocessing for naive Bayes classification and clustering with mixture of multinomials

Naive Bayes classifier has long been used for text categorization tasks. Its sibling from the unsupervised world, the probabilistic mixture of multinomial models, has likewise been successfully applied to text clustering problems. Despite the strong independence assumptions that these models make, their attractiveness come from low computational cost, relatively low memory consumption, ability to handle heterogeneous features and multiple classes, and often competitiveness with the top of the line models. Recently, there has been several attempts to alleviate the problems of Naive Bayes by performing heuristic feature transformations, such as IDF, normalization by the length of the documents and taking the logarithms of the counts. We justify the use of these techniques and apply them to two problems: classification of products in Yahoo! Shopping and clustering the vectors of collocated terms in user queries to Yahoo! Search. The experimental evaluation allows us to draw conclusions about the promise that these transformations carry with regard to alleviating the strong assumptions of the multinomial model.

Unity: relevance feedback using user query logs

The exponential growth of the Web and the increasing ability of web search engines to index data have led to a problem of plenty. The number of results returned per query is typically in the order of millions of documents for many common queries. Although there is the benefit of added coverage for every query, the problem of ranking these documents and giving the best results gets worse. The problem is even more difficult in case of temporal and ambiguous queries. We try to address this problem using feedback from user query logs. We leverage a technology called Units for generating query refinements which are shown as Also try queries on Yahoo! Search. We consider these refinements as sub-concepts which help define user intent and use them to improve search relevance. The results obtained via live testing on Yahoo! Search are encouraging.

Online advertisement campaign optimisation

Like any marketing campaigns, online advertisement campaigns need to be monitored, analysed and optimised. The quantitative methods are more crucial to online campaigns because of their dynamic pricing and highly interactive nature. Not only can marketing effectiveness be measured almost instantly in terms of measures such as click through rate and/or the acquisition/conversion rate, but a rich set of user data can also be collected and used by learning algorithms. The huge sets of dynamic data raise many challenging problems. In order to run a successful campaign, any serious advertiser, publisher or ad exchange network need a system that combines forecasting, data mining and optimisation techniques. In this paper, we propose such a methodology for a systematic analysis of the relevant problems and describe techniques that work on real world data as satisfactory solutions.

Online Advertisement Campaign Optimization

Like any marketing campaigns, online advertisement campaigns need to be monitored, analyzed and optimized. It is more so for online campaigns because online advertisements are usually sold in auction style. Prices can change very dynamically; the creatives, the landing pages and the targeting profiles can all be changed frequently to improve the effectiveness of a campaign. Also, the effectiveness can be measured almost instantly, usually in terms of click through rate and/or the acquisition/conversion rate. It raises many challenging problems in forecasting, data mining and optimization and entails an optimization system for any serious advertisers, publishers or ad networks.