Content identification method, model training method, data processing method, system and equipment

Abstract

The embodiment of the invention provides a content identification method, a model training method, a data processing method, system and equipment. The content identification method comprises the following steps of taking content to be identified as input of an application model, and executing the application model to output first result information, determining a content label as an identificationresult based on the first result information, executing a corresponding service operation according to the content label, the application model being obtained after training of a training model is completed, the training model adopting at least two loss functions to calculate at least two loss values after one iteration in the training process, and completing parameter updating based on the at least two loss values. According to the technical scheme provided by the embodiment of the invention, the content identification accuracy is high, and particularly, the method has relatively good discrimination for contents with relatively high similarity, such as phonetic close characters and homophonetic characters.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a content recognition, model training, and data processing method, system, and device. Background technique [0002] Content recognition technology is to allow machines to recognize and understand the content sent by users. For example, speech recognition technology can facilitate both human-to-human communication (HHC) and human-machine communication (HMC). HHC, for example, voice messages sent to others can be converted into text for easy reading, and voice input is also more convenient; HMC, such as voice search, personal intelligent assistant, voice-controlled games, smart home, etc. [0003] Traditional speech recognition technology is highly dependent on manually selected features and has a low accuracy rate. Applying deep learning technology to the field of speech recognition can imitate the brain's pattern of learning and recognizing speech signals, whic...

AI Technical Summary

Problems solved by technology

However, the existing speech recognition technology still has frequent errors in the recognition of similar words or homonyms. Although there is subsequent correction of the speech model, it often fails to achieve the desired effect.

Images