Minimum Error Rate Training (MERT) is an effective means to estimate the feature function weights of a linear model such that an automated evaluation criterion for measuring syste...
Wolfgang Macherey, Franz Josef Och, Ignacio Thayer...
Abstract. Synchronous programs may contain cyclic signal interdependencies. This prohibits a static scheduling, which limits the choice of available compilation techniques for such...
Conditional random fields for sequence labeling offer advantages over both generative models like HMMs and classifiers applied at each sequence position. Among sequence labeling...
Boolean operations and feature-based modeling are closely linked domains. For example, several, sometimes many Boolean operations are involved during a feature instantiation or mo...
This paper presents a formal framework for designing search algorithms which can identify target images by the spatial distribution of color, edge and texture attributes. The fram...