Research And Application Of Speech And Text Automatic Alignment Technology Based On Text Similarity Algorithm

Audio books are based on e-books with audio effects,to a certain extent to relieve the boredom of e-books.However,the production of audio books is time-consuming and inefficient.In order to further improve the production efficiency of audio books,this paper studies the automatic alignment technology of speech and text based on text similarity algorithm based on existing audio and text corpus.Based on the direction of speech recognition,this paper is divided into two steps:the first is corpus preprocessing.In order to reduce the influence of text noise,the corpus can be cleaned by regular expression.In order to further improve the efficiency of alignment and reduce the error rate,the alignment based on sentences is defined,and a threshold segmentation method based on the idea of divide and conquer is proposed to extract feature sentences.Different segmentation methods are adopted according to different sentence lengths to further improve the efficiency of alignment.The normalized audio is used for statement endpoint detection by doublethreshold method.Because of the sensitivity of short-time energy to high level,the average amplitude is used instead,which achieves better detection effect on statement endpoint.The second is text similarity calculation.The similarity between reference text and recognition text is calculated by combining vector space model and position vector,the weight factor is determined by experiment,and three evaluation criteria are obtained by statistical training of results,so as to realize automatic result evaluation.Finally,under the weight factor of 0.8 and automatic evaluation standard,the accuracy rate is about 85.21%.After research,the efficiency of the producing of audio books from existing audio and text is further improved.In this paper,based on the research results of alignment technology,the realization of the e-book reader,the realization of reading,look up words and other functions.In the performance scheme,the method of"interface+method" cache is proposed to accelerate the response speed.After the method cache is added in the test,the average response time is increased by about 36.1%,and the performance is improved.