Corpus Management System Based On B/S Architecture

Training acoustic model in Speech Recognition needs a lot of audio data.Institutions in speech recognition field often have their own data servers.The stand or fall of acoustic model is directly related to the level of the speech recognition technology.And,according to the different application scenarios,the acoustic models are different.This is one reason why there are more data in speech recognition.Because of the increasing number of data,the management faces enormous challenge.There is complicated and enormous work in managing and using audio data.According to the above problem,this paper puts forward the data into the corpus,and using the WEB system based on B/S architecture to manage them.This paper prepares some Linux servers used to store audio data.This system guides the corpus administrators set related characteristics of corpus,and then the audio data and its labeled results are stored to the database,then the WEB system,according to the characters of corpus,judges and classify the audio format.This WEB system provides data management of corpus and adding,deleting,modifying,querying and other basic operations on its audio data contained in corpus.At the same time,the system also provides audio annotation,feature extraction;generate the training set and test set,and other functions.This system contains all speech recognition data preparation work;his implementation of data management has the very vital significance.