Food Matching Recommendation Based On Component Feature Extraction And Latent Semantic Analysis

There are kinds of nutrient in food,and the interaction between them is changeable.Therefore,it is difficult for people to understand and master the gold match of all kinds of food.In this thesis,a food matching recommendation strategy based on component feature extraction and latent semantic analysis is proposed in order to reduce the cognitive burden of gold between foods and provide theoretical support for food matching recommendation system.Based on the analysis of the syntactic relationship between food and nutrition,the feature extraction of food nutrition is realized,and the reasonableness of food matching is judged by the sentiment tendency of predicate verbs.Then the matrix extracted from the component feature is decomposed by singular value,and the dimension is reduced based on the result of singular value decomposition.The latent semantic relationship between food matching is found,and the semantic space of food matching is formed,and the similarity of food matching is calculated to provide a basis for food matching recommendation.The main contents of this study are as follows:First,component features of the food matching are extracted.Firstly,natural language processing is used to analyze Chinese word segmentation and dependency parsing to obtain the grammatical relationship between words.Then,the dependency parsing tree is traversed by postorder-traversal to obtain the nutrition components corresponding to foods in the sentence;the dependency parsing tree is traversed by preorder-traversal to obtain the predicate verbs which represent the sentence emotion.The positive and negative meanings of predicate verbs are taken as the basis for judging whether food matching is beneficial or harmful to human health.Finally,the corresponding component of the food nutrition vector is constructed by judging whether the food contains the current nutritional components.The mean values of multiple food nutrition vectors in the same dimension were calculated and used as the components of the food matching vector.Second,food matching recommendation based on the latent semantic.First of all,the complementary food matching vector is extracted.The feature matrix of recommendation-requirement is constructed by these vectors,which use the food matching as a row,nutrition as a column.Then,the feature matrix is decomposed by the singular value to reduce the vector dimension and map the latent semantic relationship between the rows and rows of the matrix into the latent semantic space of food matching.The experimental results show that the recommendation model proposed in this paper can realize the recommendation of similar food matching by constructing the latent semantic relationship,which provides an effective mean for consumers to choose food matching reasonably.In this paper,the matrix decomposition technology is innovatively applied to the feature processing of daily diet.This not only reduces the amount of data computation,but also excavates the latent semantic relationship between food matching and nutrients.The introduction of the recommendation system provides a reference about daily food matching for diners.