Advertisers use Sponsored Search to drive traffic to their site at a conversion rate and cost per conversion that provides value to them. However, very often advertisers bid at a constant price on a bundle of keywords, either for lack of enough data to fully optimize their bids at a keyword level, or indirectly by opting into Advanced Matching (AM) that allows an advertiser to reach a large number of queries while explicitly bidding only on a limited number. Then this single bid price reflects the return the advertiser gets from the full bundle. Under these conditions, the advertiser is competing too aggressively for some keyword auctions and with too low bids for others. In this paper, we propose a solution to improve the fairness of each keyword's bid prices within an AM bundle: adjusting the AM keyword bid by the ratio of its conversion rate to the conversion rate it would have reached had it been an Exact Match (EM). First we describe how we measure advertisers' conversion rates despite the opt-in nature of conversion tracking, and illustrate the need for bid adjustment in the context of AM. Then we present our approach to predict conversion rates in a robust manner. Our model uses a number of features capturing the quality of the match between the ad and the query. Then we describe how we adjust keyword bid prices to reflect their value to the advertiser thereby improving (1) the auction through fewer incorrectly high bids in the auction, (2) advertiser return through more auctions won by high value keywords and less by low value keywords, and (3) user satisfaction through higher conversion rate. Finally, we present experimental results from our live system.

This paper summarizes the work of the "Rapid Speech Recognizer Adaptation" team in the workshop held at Johns Hopkins University in the summer of 1998. The project addressed the modeling of dependencies between units of speech with the goal of making more effective use of small amounts of data for speaker adaptation. A variety of methods were investigated and their effectiveness in a rapid adaptation task defined on the Switchboard conversational speech corpus is reported.

Two models of statistical dependence between the acoustic model parameters of a large vocabulary conversational speech recognition (LVCSR) system are investigated for the purpose of rapid speaker- and environment-adaptation from a very small amount of speech: (i) a Gaussian multiscale process governed by a stochastic linear dynamical system on a tree, and (ii) a simple hierarchical tree-structured prior. Both methods permit Bayesian (MAP) estimation of acoustic model parameters without parameter-tying even when no samples are available to independently estimate some parameters due to the limited amount of adaptation data. Modeling methodologies are contrasted, and comparative performance of the two on the Switchboard task is presented under identical test conditions for supervised and unsupervised adaptation with controlled amounts of adaptation speech. Both methods provide significant (1% absolute) gain in accuracy over adaptation methods that do not exploit the dependence between acoustic model parameters.

This paper compares parametric and nonparametric constrained-mean trajectory segment models for large vocabulary speech recognition, extending distribution clustering techniques to handle polynomial mean trajectory models for robust parameter estimation. The parametric model has fewer free parameters and gives similar recognition performance to the nonparametric model, but has higher recognition costs.

Segment models are a generalization of HMMs that can represent feature dynamics and/or correlation in time. We develop the theory of Bayesian and maximum-likelihood adaptation for a segment model characterized by a polynomial mean trajectory. We show how adaptation parameters can be shared and adaptation detail can be controlled at run-time based on the amount of adaptation data available. Results on the Switchboard corpus show error reductions for unsupervised transcription mode adaptation and supervised batch mode adaptation.

Describes a method for clustering multivariate Gaussian distributions using a maximum likelihood criterion. The authors point out possible applications of model clustering, and then use the approach to determine classes of shared covariances for contest modeling in speech recognition, achieving an order of magnitude reduction in the number of covariance parameters, with no loss in recognition performance.>