Audio text alignment method and system based on Doc2Vec

Abstract

The invention discloses an audio text alignment method and system based on Doc2Vec; the method comprises the steps: carrying out the threshold estimation based on an AIC-FCM optimized through a simulated annealing genetic algorithm, segmenting an audio with the length of a book into short audios with sentences as dimensions, carrying out the voice recognition of the short audios, and outputting short texts with sentences as dimensions; carrying out paragraph extraction on the electronic book based on a Doc2Vec model to obtain the paragraph texts with paragraphs as dimensions; and performing text similarity matching on the short text and the paragraph text based on a dynamic matching method of a threshold prediction method to complete text alignment. Compared with a traditional audio text alignment algorithm, long audio segmentation is closer to an ideal segmentation result, the alignment effect is basically equal to that of Doc2vec, and the time complexity is reduced by about 35%.

Description

technical field [0001] The invention relates to the technical field of natural language processing, and more specifically relates to the problem of alignment of audio texts, so as to improve the efficiency and quality of audio comparison e-book production. Background technique [0002] Based on the association between audiobooks and e-books, e-books that can produce sounds can be produced. This kind of e-book has important practical value in education, especially in language education scenarios. However, currently, the production of such e-books is not widely. The reason is that such books mostly rely on manual annotation for production, which limits their development to a large extent. As the technical feasibility of this type of book, audio text technology still has some problems. First of all, in terms of audio processing, the audio accompanying the book is generally read and recorded by professionals according to the text of the e-book, so it has the ability to match th...

AI Technical Summary

Problems solved by technology

If the longer audio is directly recognized, it will not only bring a large decoding time consumption, but also reduce the accuracy of the recognition output

The second is the alignment algorithm. At present, various alignment algorithms are quite perfect, but there are still problems such as time complexity and high space complexity.

Images