In this paper, we address the problems of deformable object matching (alignment) and segmentation with cluttered background. We propose a novel hierarchical log-linear model (HLLM) which represents both shape and appearance features at multiple levels of a hierarchy. This model enables us to combine appearance cues at multiple scales directly into the hierarchy and to model shape deformations at short-range, medium range, and long-range. We introduce the structure-perceptron algorithm to estimate the parameters of the HLLM in a discriminative way. The learning is able to estimate the appearance and shape parameters simultaneously in a global manner. Moreover, the structure-perceptron learning has a feature selection aspect (similar to AdaBoost) which enables us to specify a class of appearance/shape features and allow the algorithm to select which features to use and weight their importance. This method was applied to the tasks of deformable object localization, segmentation, matching (alignment), and parsing. We demonstrate that the algorithm achieves the state of the art performance by evaluation on public dataset (horse and multi-view face).

Opinion retrieval is a task of growing interest in social life and academic research, which is to find relevant and opinionate documents according to a user's query. One of the key issues is how to combine a document's opinionate score (the ranking score of to what extent it is subjective or objective) and topic relevance score. Current solutions to document ranking in opinion retrieval are generally ad-hoc linear combination, which is short of theoretical foundation and careful analysis. In this paper, we focus on lexicon-based opinion retrieval. A novel generation model that unifies topic-relevance and opinion generation by a quadratic combination is proposed in this paper. With this model, the relevance-based ranking serves as the weighting factor of the lexicon-based sentiment ranking function, which is essentially different from the popular heuristic linear combination approaches. The effect of different sentiment dictionaries is also discussed. Experimental results on TREC blog datasets show the significant effectiveness of the proposed unified model. Improvements of 28.1% and 40.3% have been obtained in terms of MAP and [email protected] respectively. The conclusion is not limited to blog environment. Besides the unified generation model, another contribution is that our work demonstrates that in the opinion retrieval task, a Bayesian approach to combining multiple ranking functions is superior to using a linear combination. It is also applicable to other result re-ranking applications in similar scenario.