We address the problem of linking related documents across languages in a multilingual collection. We evaluate three diverse unsupervised methods to represent and compare documents: (1) multilingual topic model; (2) cross-lingual document embeddings; and (3) Wasserstein distance. We test the performance of these methods in retrieving news articles in Swedish that are known to be related to a given Finnish article. The results show that ensembles of the methods outperform the stand-alone methods, suggesting that they capture complementary characteristics of the documents.
|Title of host publication||Proceedings of the LREC 2020 Workshop on Cross-Language Search and Summarization of Text and Speech|
|Publication date||16 May 2020|
|Publication status||Published - 16 May 2020|
|MoE publication type||A4 Article in conference proceedings|
Zosa, E., Pivovarova, L., & Granroth-Wilding, M. (2020). A Comparison of Unsupervised Methods for Ad hoc Cross-Lingual Document Retrieval. In Proceedings of the LREC 2020 Workshop on Cross-Language Search and Summarization of Text and Speech