Text sentence segmentation method and system

Abstract

The invention discloses a text sentence segmentation method and system. The method comprises following steps: pre-collecting a small amount of textual data and corresponding speech data, constructinga long-term memory segmentation model based on text segmentation features and acoustic segmentation features; when the text is segmented, obtaining the text of the sentence to be segmented and corresponding speech data; extracting text segmentation features and the acoustic segmentation features according to the to-be-segmented text and the speech data corresponding to the to-be-segmented text, respectively; according to the extracted text segmentation features, acoustic segmentation features, and the long-term memory segmentation model, segmenting the to-be-segmented text. The invention can effectively improve the accuracy of text segmentation.

Description

technical field [0001] The invention relates to the field of natural language processing, in particular to a text sentence segmentation method and system. Background technique [0002] In recent years, with the practicality of speech recognition technology and the rapid development of hardware storage, more and more people are accustomed to using storage devices to record their voices, and using transcription tools to convert recorded voice data into text data for information preservation. Instead of the traditional way of manual recording while listening and taking notes to record important information. However, when performing speech recognition on speech data to obtain the corresponding recognized text, the text data is often continuous and uninterrupted, which is not conducive to the user's reading and understanding. The traffic pressure is very high, and parking is difficult. On the other hand, in this window, in this administrative service center, we can see that ther...

AI Technical Summary

Problems solved by technology

[0003] The existing sentence segmentation methods generally use the word vector information of the text data to segment the sentence directly through the method of sequence labeling. However, the word vector can only describe the text data, and cannot describe the relevant information of the text data corresponding to the speech data, so that the sentence segmentation The accuracy is low; in addition, the prior art generally uses a sequence tagging model to segment sentences, and the sequence tagging model can only remember less historical information, and cannot remember the future information of each word, further reducing the accuracy of sentence segmenting

For example, "how should I do something that will make her change her mind", the current word of the constructed sequence model is "thing", if the model cannot remember the historical information "how" of "thing", then the sentence sentence judgment at the "thing" , it is very likely that errors will occur; another example is "the word you said is a question word", if the model cannot remember the future information of the word "?", it will also make mistakes when judging the sentence at the word "?"

Images