Semantic Search System with Metagraph Knowledge Base and Natural Language Processing

• Main Authors: Anton Kanev, Valery Terekhov
• Format: Article
• Language: English
• Published: FRUCT 2021-01-01
• Series: Proceedings of the XXth Conference of Open Innovations Association FRUCT
• Subjects: information retrieval; semantic search; knowledge base; natural language processing; metagraph
• Online Access: https://www.fruct.org/publications/acm28/files/Ter.pdf

Currently, various investigations are actively carried out to improve the precision and recall of information retrieval. Many authors associate this process with the need to analyze the meaning of words. The authors of this paper have proposed a semantic search method using natural language processing and the metagraph knowledge base. The general model and main algorithms of the proposed method for indexing and information extraction are described. Natural language processing capabilities affect the amount of data available for search, thus, the recall of the information extraction system was measured. Marking up a dataset according to meaning depends on the situation and is subjective. Therefore, the precision of semantic search was assessed on an unlabeled dataset using the methodology proposed by the authors. To increase recall, semantic search is complemented by keyword search, and semantics results are used to change the ranking of user query results. The authors suggested set of queries for this investigation. The ranking order for semantic and regular keyword searches was estimated using the metric proposed by the authors.