We present an investigation of recently proposed character and word sequence kernels for the task of authorship attribution based on relatively short texts. Performance is compare...
Abstract—Modern military and civilian surveillance applications should provide end users with the high level representation of events observed by sensors rather than with the raw...
* We present a method for representing and recognizing visual events using attribute grammars. In contrast to conventional grammars, attribute grammars are capable of describing fe...
Inducing a grammar from text has proven to be a notoriously challenging learning task despite decades of research. The primary reason for its difficulty is that in order to induce...
We describe how simple, commonly understood statistical models, such as statistical dependency parsers, probabilistic context-free grammars, and word-to-word translation models, c...