Classification of intelligent speech system and education method based on improved multi label transfer learning model

In recent years, improved multi label learning has been widely used in text classification, protein function prediction, image annotation and other fields. In a variety of multi label applications, the most important thing is to correctly classify each sample and its corresponding label. And the text information and visual information are used to improve the semantic representation of deepening image annotation. In order to further solve the sample problem in the medical field, we introduce the general domain method into the medical transfer learning model to solve the imbalance of single label and multi label sorting data. This paper uses the upstream data sampling method and the method of single label data to multi label data migration to balance the data. We take the intelligent voice system as the research direction, combined with speech recognition technology and high frequency infrared control technology applied to smart home devices, and develop the hardware facility of intelligent voice. As a home human-computer voice interactive terminal, it can search pleasant music, or know today's weather conditions, we can also set a series of functions such as alarm clock; In addition, based on the classification of educational methods, we analyzed the problems existing in the teaching process elements such as teaching type, practical teaching, teaching and campus culture, and carried out the teaching practice of problem-solving concept.