Monolingual and multilingual topic analysis using LDA and BERT embeddings

Highlights：

• The paper analysis topic evolution in monolingual, and multilingual topic similarity relations in library and information science.

• For monolingual, we computed the relation score for topics at prior-and post-year to explore their evolution in scientific publications.

• For multilingual, we calculated the similarity of publication datasets between two different languages in the same year.

• Our approach was able to compare the scientific research frontiers in different languages at a given time.

• By given a topic evolution in the source language with a multilingual similarity relation, we can predict the key topic in the target language.

摘要

•The paper analysis topic evolution in monolingual, and multilingual topic similarity relations in library and information science.•For monolingual, we computed the relation score for topics at prior-and post-year to explore their evolution in scientific publications.•For multilingual, we calculated the similarity of publication datasets between two different languages in the same year.•Our approach was able to compare the scientific research frontiers in different languages at a given time.•By given a topic evolution in the source language with a multilingual similarity relation, we can predict the key topic in the target language.

论文评审过程：Received 20 December 2019, Revised 27 April 2020, Accepted 18 May 2020, Available online 25 June 2020, Version of Record 25 June 2020.