| With the continuous development of artificial intelligence technology,the application scenarios of AI technology are constantly diversified.In this paper,the classification algorithm of infant crying and the intelligent question-answering algorithm in the field of parenting are studied,and a deep neural network model is proposed to classify the crying of infants and answer parents’ questions.The main work and innovation of this article is as follows:(1)By studying Dunstan’s infant crying classification theory,and in video website collects videos and audios of infants and toddlers crying,manually edits and labels them to get them5 Audio data of infant crying in a variety of emotional categories.And on this basis,the sound is converted into a spectral image to obtain it 5 Image data of infant crying spectrum of various emotional categories.By crawling Baidu Wenku,Douding and other document websites,we can obtain the commonly used parenting knowledge documents in the field of parenting,and carry out data cleaning and manual marking to obtain the Q&A database in the field of parenting.(2)Through the spectral data analysis of infant crying,the Inception-Res Net-v2 model is selected for training,and the structure of the Inception-Res Net-v2 model is introduced in detail,and changes are made on this basis.Compared with the previous Inception-Res Netv2 model,the accuracy rate was greatly improved,and finally compared with the traditional Res Net101 network,it was verified that the model only needed fewer training rounds than the traditional model,and the accuracy rate could be higher.(3)The processing of Q&A data in the field of parenting involves the use of search engine tools Elastic Search Index the data for Q&A.A dictionary of data extraction keywords for multi-domain Q&A data extraction,and a dictionary of data extraction keywords for the field of childcare.Use of questions Elastic Search Conduct recalls and data enhancements to questions,where accuracy of the data is primarily involved EDA Methods,homophone substitution.Data is enhanced after processing the data,and then trained with keyword information BERT model and compare it with the data before processing.Finally,the experimental results are summarized,and the accuracy of the infant cry classification algorithm is significantly improved after the improved Inception-Res Net-v2 network is adopted.The use of the BERT model with keyword information and the operation of data enhancement can improve the accuracy of the question answering system,which is of great significance for the application of intelligent cradle cars. |
