A Content Matching Method Based on Speech Recognition

Abstract

The invention discloses a content matching method based on speech recognition, comprising the following steps: 1. Correspondingly encoding the name of the content in the content database into a first sound code, and combining the first sound code with the ID of the content in the content database The corresponding relationship is stored in the sound code mapping library; 2. Convert the sound into text, and encode the text into a second sound code; 3. Search for a matching first sound code in the sound code mapping database according to the second sound code After finding the matching first sound code, the ID of the corresponding content is retrieved through the first sound code, and the matching content is searched in the content database by the ID of the corresponding content. The present invention encodes the text recognized by the voice into a sound code, and at the same time converts the content name in the content database into a sound code by means of sound format coding, adopts a sound code matching method, removes the influence of polyphonic characters and fuzzy sounds, and solves the problem of pronunciation Questions that are similar but with different text cannot be matched.

Description

A Content Matching Method Based on Speech Recognition technical field The invention belongs to the field of speech recognition, and relates to a content matching method based on speech recognition. Background technique With the explosion of artificial intelligence technology and the gradual maturity of voice technology, the human-like dialogue interaction based on dialogue flow [1]. Interaction has become a new type of interaction, and the mainstream smart device interaction mode is changing from remote control, touch interaction to Convergence or transformation of voice interaction. [0003] At present, smart speakers, televisions with voice input, and set-top boxes have gradually become new ways for users to access media content. By saying the content name, the user can play a song, play a video, start an APP, etc. recorded on the computer The content name in the system is in the form of text, and the voice is a sound method. The pronunciation is close, but the ...

AI Technical Summary

Problems solved by technology

Since the name of the content recorded in the computer system is in the form of text, and speech is a form of sound, the converted text after speech recognition is close in sound, but cannot be matched precisely in text. For example, for Hu Yifei, the pronunciation is "" fu yi hui" may be recognized as "Fu Yihui". At this time, the correct content cannot be found through the traditional text matching method.

Images