Design And Implementation Of Knowledge Graph Based Question Answering And Recommendation System In Diet For The Elderly

Balance diet is one of the best ways for the aged and infirm to maintain health and prevent chronic diseases.At present,the number of senior citizens in China is growing rapidly,and the dietary health problems of their aging are gradually highlighted.On the one hand,with the increase of the aging population and the increasing emphasis on dietary health,the demand for this crucial moment nutritional information of the elderly in society is growing rapidly.On the other hand,older people lack access to comprehensive dietary knowledge,and it is difficult to distinguish the mixed information on the Internet.Based on this,this thesis designs and implements a question answering and recommendation system based on knowledge graph in diet for the elderly.This system ensures easy operability,which accurately answers questions for the aged and their families in real-time,facilitates users to quickly query dietary knowledge and recommends suitable recipes and specific practices.It offers an important solution,especially for the aging population,to food habit challenges in everyday life.The main research work of this thesis is as follows:1.Developed a dietary information graph to take into consideration of senior citizens.To design pattern layer here used combining domain knowledge with expert opinion;and Scrapy crawler framework is used to capture a large number of diet-related data and build a labeled data set.Through comparative experiments,BERT-IDCNN-CRF and BERTBiLSTM models are selected for named entity recognition and relationship extraction to construct knowledge graph triples,and Neo4j graph database is used for storage.2.Research on question answering and recommendation algorithm.(ⅰ)Question answering procedure is developed based on semantic analysis.Through comparative experiments of user question data set,BERT and BERT-BiLSTM-CRF models are selected for intention recognition and named entity recognition.The results are retrieved in the knowledge graph after logical conversion.(ⅱ)The KGCN recommendation model has improved,and the user history question information is introduced to propose the KGCN-UHQ model.(ⅲ)In order to guarantee the exact response to questions and to satisfy the users’ individual desires,the question answers algorithm focused on semantical analysis is innovatively combined with the recommendation algorithm.3.This question answering and recommendation system is designed and implemented in user friendly way.The front-end and the back-end separation architecture is used to realize the PC and mobile terminals of the system.The system is equipped with the voice recognition interface and realize inputting questions through voice and text,replying the corresponding answers,voice broadcast,purchasing ingredients,storing user historical interactive data and so on.