Matching Literature Heritage Entities From Heterogeneous Data Sources Based On The Textual Description

Abstract

The paper focuses on problem of short text matching for literature heritage entities alignment from heterogeneous data sources. The overview of existing methods showed that all of them works well for long texts. The paper proposes modification of Jacquard similarity metric for solving the problem based on similarity of unique text tokens adjusted to the specifics of literature heritage domain. Achieved results were evaluated on the literature heritage of the A.S. Pushkin gathered from the various heterogeneous sources (datasets, full works compilations. Encyclopedia of A.S. Pushkin) and shown high accuracy of finding corresponding entities within the system by developed method.

Keywords

• nlp text matching literary heritage jaccard similarity entity matching