Method and system for building a topic specific language model for use in automatic speech recognition

摘要: An automatic speech recognition method includes at a computer having one or more processors and memory for storing programs to be executed by the processors, obtaining plurality of corpus categories through classifying calculating raw corpus; classified language models that respectively correspond model training applied on each category; an interpolation implementing weighted merging interpolated models; constructing decoding resource in accordance with acoustic model; input using resource, outputting character string highest probability as result speech.