Design And Implementation Of Maternal Health Speech Consultation Platform

Maternal health care plays an important role in ensuring the health of pregnant women and fetuses and reducing the probability of fetal malformation.The scientific implementation of maternal health care activities is a key prerequisite for ensuring maternal and fetal health,and professional maternal health knowledge guidance is the basis for ensuring the smooth progress of health care activities.However,the maternity information service lacks effective ways and means,and the way of acquiring knowledge and its carrier form are relatively lacking.In recent years,mobile Internet has flourished,mobile phones can serve the public as a new information carrier,and the emergence of speech recognition and speech synthesis technology has made it possible for people to interact with terminals through voice conversations.The maternal use of voice to interact with the terminal,using the mobile phone as the carrier of information,through the background processing to get the answer,can facilitate the maternal access to high-quality consulting services,which has great guiding significance for maternal health.This paper designs a maternal health voice consultation platform.The platform uses Android as the client.The maternal interacts with the terminal in a voice manner,and uses powerful background computing capabilities to put functions such as corpus crawling and semantic matching in the background.Processing,the final platform provides counseling responses to pregnant women.The main work of this paper is as follows:(1)On the basis of the Keda Xunfei platform,the core method provided by the SDK is encapsulated,and the interface package library is constructed to complete the speech recognition and synthesis functions.(2)Based on the Scrapy framework and Redis database,a distributed corpus crawler was designed and developed to crawl the Q&A corpus of the maternal health website.The crawler stores the URL in the Redis database,and uses the collection data structure of Redis to implement Request deduplication.The master node sharesthe crawl queue,the fingerprint collection and the child nodes,and distributes the crawl task to the child nodes,compared to the crawl of the single node.The distributed corpus crawler can get the question and answer corpus faster.(3)Aiming at the matching problem of questions and answers,a hybrid semantic model based on deep learning is proposed.By acquiring the context information for the model training,the convolution kernel is used to extract the local semantic information of the sentence,and the attention mechanism is adopted to consider The irrelevant features,the semantic features extracted by the final model can better represent the matching relationship between sentences.Experiments show that compared with a single neural network,the model has higher accuracy in the problem answer matching task.(4)Finally,the maternal health consultation platform was designed and implemented,with mobile phone as the interactive end,WEB application as the background processing end,HTTP as the communication protocol,REST request method to decouple the front and back,Json as the standard interactive format,and finally Implemented the platform and demonstrated it.