Math-aware Similarity of Papers in Digital Mathematics Libraries

SOJKA, Petr. Math-aware Similarity of Papers in Digital Mathematics Libraries. 2014.

Basic information

• Original name: Math-aware Similarity of Papers in Digital Mathematics Libraries
• Authors: SOJKA, Petr (203 Czech Republic, guarantor, belonging to the institution).
• Edition: 2014.

Other information

• Original language: English
• Type of outcome: Conference abstract
• Field of Study: 10201 Computer sciences, information science, bioinformatics
• Country of publisher: Czech Republic
• Confidentiality degree: is not subject to a state or trade secret
• WWW: Extended abstract
• RIV identification code: RIV/00216224:14330/14:00077987
• Organization unit: Faculty of Informatics
• Keywords (in Czech): podobnost dokumentů; digitální matematická knihovna; EuDML; DML-CZ
• Keywords in English: document similarity; digital mathematics library; EuDML; DML-CZ
• Tags: International impact

Abstract

The exploratory, semantic similarity searching is becoming widespread in digital libraries, and math ones are no exception. For working mathematicians and their use of digital mathematical libraries (DML) as the Czech Digital Mathematics Library DML-CZ or European Digital Mathematics Library (EuDML) we have designed and implemented math-aware similarity computation framework based on leading edge topic modelling techniques implemented by Gensim software package. Studies on the classification of math papers done for DML-CZ have been tested and deployed in EuDML, where for given paper ten most semantically similar papers are computed and shown. In the latest experiments we are evaluating several possible representations of mathematical formulae to get the semantically similar papers. Quality of similarity is measured by comparation to the similarity matrix induced from the Mathematical Subject Classifications every paper is marked up by. In the talk we will report a) about the evaluation of the similarities computed by several different methods, b) on the experience from 20 months of deployment in EuDML and more than 5 years in DML-CZ, c) about the importance of representing formulae even for paper similarity computations, d) on setting up Gensim for the math-aware use in DML projects.

Links

• LG13010, research and development project: Name: Zastoupení ČR v European Research Consortium for Informatics and Mathematics (Acronym: ERCIM-CZ)

Investor: Ministry of Education, Youth and Sports of the CR

• 250503, interní kód MU: Name: The European Digital Mathematics Library (Acronym: EuDML)

Investor: European Union