A reinforcement learning-based algorithm for discrete dynamic stochastic recognition of speech dialog emotions

Abstract

At present, the most common speech dialogue emotion discrete dynamic random recognition calculation natural language processing is mostly independent processing of object data, and the recognition efficiency is low, resulting in the infinite increase of the final FRP value. According to the current recognition requirements, speech data resource collection and speech recognition are carried out first, and a multi-level method is adopted to improve the recognition efficiency and realize multi-level processing and sentiment analysis of natural speech. Based on this, a discrete dynamic random recognition calculation model of reinforcement learning speech dialogue emotion was constructed, and the multi-cycle automatic synchronous correction method was used to realize the random recognition processing. The test results show that for the six randomly selected test periods, compared with the improved emotion model random recognition method and the artificial intelligence emotion random recognition method, the final FRP value of the reinforcement learning emotion random recognition method designed this time is well controlled 15%, which indicates that with the assistance of reinforcement learning technology, the designed calculation method is more flexible and changeable. Furthermore, its inherent random recognition mechanism is more comprehensive, efficient, and targeted, thereby rendering it highly valuable and significant for applications under complex background conditions.