2018/07/02 by Edoardo Maria Ponti, Ponti, Edoardo Maria, Helen O’Horan +13 · 2 citations
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques
paper · pdf · doi:10.48550/arxiv.1807.00914
openalex publication_date 2018/07/02 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Linguistic typology aims to capture structural and semantic variation across\nthe world's languages. A large-scale typology could provide excellent guidance\nfor multilingual Natural Language Processing (NLP), particularly for languages\nthat suffer from the lack of human labeled resources. We present an extensive\nliterature survey on the use of typological information in the development of\nNLP techniques. Our survey demonstrates that to date, the use of information in\nexisting typological databases has resulted in consistent but modest\nimprovements in system performance. We show that this is due to both intrinsic\nlimitations of databases (in terms of coverage and feature granularity) and\nunder-employment of the typological features included in them. We advocate for\na new approach that adapts the broad and discrete nature of typological\ncategories to the contextual and continuous nature of machine learning\nalgorithms used in contemporary NLP. In particular, we suggest that such\napproach could be facilitated by recent developments in data-driven induction\nof typological knowledge.\n