Design And Implementation Of Emotional Consulting Service APP

Today's fast pace of life and heavy work pressures have led to gradual increase problems in people emotional.However,parents have no enough time to pay attention to the mental health of young teenagers in real time.Once teenagers have emotional problems,they usually solve them by self-regulation,which can easily lead to the accumulation of negative emotions,resulting in various extreme events.According to investigation,with the upgrading of mobile technology,teenagers generally use mobile phones to chat to solve their emotional problems.In view to this phenomenon,this paper designs an emotional consulting service APP,which using We Chat mini program as the front end of the APP to interact with users.At the same time,a dialogue model is constructed as the back end of the APP to process the data sent by users and replace the human emotional mentors to reply to the users' emotional problems.The main tasks of this paper are as follows:Firstly,a generative chatbot model is constructed.Using word2 vec and BERT to obtain static and dynamic word vector coding construct seq2seq-Attention dialogue model.The deep learning model GRU is used at the encoder-decoder.Through model training and testing,the two models with the best performance are evaluated.In the evaluation,the recovery correlation of the two models is stable in the range of 0.7-0.8,but in terms of recovery fluency,the Perplexity of BERT model is 0.15 lower than that of word2 vec model,and the average score of the BERT model is 0.28 higher than that of the word2 vec model in terms of artificial evaluation,so the seq2seq-Attention model encoded by BERT is better.Secondly,doc2 vec algorithm and BERT algorithm are used to build retrieval chatbot model.The experimental results show that the retrieval accuracy of doc2 vec model is 0.02 lower than BERT model,but the reaction time is 4.1 s shorter than BERT model.Because BERT model takes up too much resources and user experience is poor,so doc2 vec model is more suitable for retrieval model.Finally,the generation and retrieval technology are combined to build a joint chatbot model and develop an emotional consulting service APP.The joint model first retrieves similar questions through the retrieval model,if the similarity is higher than the threshold,the responses corresponding to the first question will be directly output,otherwise the responses will be added to the candidate answer set,and the responses will be made by generating the model to calculate the Perplexity of the output sentences of the generated model.If the Perplexity is lower than 0.09,the generated model responses can meet the requirements and can be output,otherwise the responses that rank first in the candidate answer set will be output.Then,we design and implement a We Chat small program based on We Chat APP,and use the Flask framework to achieve information interaction,so as to obtain the function of real-time and appropriate solution to user emotional problems.Emotional consulting service APP reduces the waiting time of users,improves the user experience,more accurately understands the user's emotional problems,and makes appropriate responses,so as to encourage,enlighten users,alleviate pressure and prevent extreme events.