This paper proposes a word sense language model based method for information retrieval. This method, differing from most of traditional ones, combines word senses defined in a thes...
An N-gram language model aims at capturing statistical word order dependency information from corpora. Although the concept of language models has been applied extensively to handl...
The recent availability of large collections of text such as the Google 1T 5-gram corpus (Brants and Franz, 2006) and the Gigaword corpus of newswire (Graff, 2003) have made it po...
We measure the effects of a weak language model, estimated from as little as 100k words of text, on unsupervised acoustic model training and then explore the best method of using ...
Statistical language models estimate the probability of a word occurring in a given context. The most common language models rely on a discrete enumeration of predictive contexts ...
John Blitzer, Kilian Q. Weinberger, Lawrence K. Sa...