A new way of search query like knowledge graph and its interpretability

Ying-jie Xie, Guosun Zeng

Abstract: In online search, it has always been a challenge to provide a way of search query that can help users accurately express search intentions and encourage them to type easily. This paper studies a knowledge graph-like search query and its interpretability. Inspired by the excellent expressive capabilities of the knowledge graph in knowledge reasoning and explainable artificial intelligence, we propose a new way of search query named search query knowledge graph, which can help users express their search requests using concepts such as entities, relationships, and facts. Therefore, people can draw a graph simply and interactively as a search query in a drag-and-drop manner. Compared with traditional search query keyword, the search query knowledge graph is not only fully compatible with it but also enriches and expands it. The search query keyword is a special case of the search query knowledge graph. We take the search query knowledge graph as the search input, obtain the corresponding search results in an online search, and then generate another knowledge graph named search result knowledge graph. According to the structural mapping relationship and the semantic matching relationship between the above two knowledge graphs, we innovatively discuss the structural interpretability and semantic interpretability for the search query knowledge graph. Some judgments or criteria about structural interpretability and semantic interpretability are given, and the sufficiency and necessity of the two interpretabilities are addressed. Extensive experimental results show that the search query knowledge graph proposed in this paper is a more effective and accurate way to express users’ search intentions and can provide help for online search.