Text-Audio Automatic Summarization Method Based on Improved PageRank Algorithm

Abstract

The invention provides a method of automatically abstracting text-audio based on the improved PageRank algorithm. The method comprises the steps that audios are transcribed into text, the similarities of the transcription text and the original text are computed; the importance of the sentences in the transcripts and the original texts is computed by utilizing the improved PageRank algorithm and based on the similarities; and according to the importance of the sentences, a text summary is generated. By the improved PageRank algorithm, the sorting of texts is guided, the similarity matrix between the cross-modal sentences is modified, so that the connection direction between two sentences with the same semantics coming from different modes is changed from the original bi-direction to the unidirection, so that when the semantics of the original text sentences and the transcribed text sentences are the same, the sentences are outputted in accordance with the original text, while more important information is assuredly obtained, the problems of grammatical errors and incoherencies brought about by the transcriptions are avoided, and the readability of the automatic abstracts of the text-audios is improved.

Description

technical field [0001] The invention belongs to the technical field of natural language processing and speech processing, and relates to a text-audio automatic summarization method based on an improved PageRank algorithm. Background technique [0002] With the advent of the Internet age, multimedia information including text documents, audio and other information can be seen everywhere. Effectively analyze relevant text and audio information, and obtain as much important information as possible, and present it to users in the form of text summaries, which helps users obtain key information only through text summaries, and realizes convenient and efficient Access to information. For a series of text and audio related to an information event, the current common practice is to first transcribe the audio into text through an automatic speech recognition system, and then combine it with the original text document to generate text through the traditional multi-document automatic ...

AI Technical Summary

Problems solved by technology

Compared with the method of obtaining information from the original text alone, the above method transcribes the relevant information in the voice information, and this information will not be completely covered by the original text document, which helps to increase the amount of information obtained. However, automatic speech recognition exists Many errors, resulting in a less fluent transcription, more grammatical errors, and thus less readability

Therefore, while ensuring as much important information as possible, improving the poor readability of the existing text-audio automatic summarization is a technical problem that needs to be solved at present.

Images