A Knowledge Graph Construction Method for Food Nutrition

Abstract

Food provides indispensable nutrition to sustain people's life activities. Lack of awareness of the nutritional ingredients in food will lead to health issues caused by an unbalanced diet, inadequate nutrition, and nutritional overload. With more and more sugar-free, low-fat products coming onto the market today, there is a growing concern about the nutritional ingredients of foods. In this paper, we propose the knowledge graph of the nutrition ingredient of food constructed by the hybrid model, which helps users to understand the detailed information of nutrient ingredients more clearly. We first pre-process the data according to the type of data crawled. As for structured data, we convert them into triples, which are used to construct the graph. And as for unstructured data, knowledge extraction technology is mainly used. Knowledge extraction mainly focuses on dependency parsing for relation extraction and performs knowledge fusion on the extracted data to calculate the similarity between data points of different categories. Then, we divide the data of different levels into tree clustering structures to find the lowest cost clustering scheme. Finally, the processed data is stored in the Neo4j graphical database for the visual display of the graph, which helps individuals to understand the nutritional ingredients of food more intuitively.