A Diverse Set of Freely Available Linguistic Resources for Turkish

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 13739–13750 July 9-14, 2023 ©2023 Association for Computational Linguistics A Diverse Set of Freely Available Linguistic Resources for Turkish Duygu Altinok Deepgram duygu.altinok@deepgram.com Abstract This study presents a diverse set of freely avail- able linguistic resources for Turkish natural lan- guage processing, including corpora, pretrained models and education material. Although Turk- ish is spoken by a sizeable population of over 80 million people, Turkish linguistic resources for natural language processing remain scarce. In this study, we provide corpora to allow prac- titioners to build their own applications and pretrained models that would assist industry re- searchers in creating quick prototypes. The pro- vided corpora include named entity recognition datasets of diverse genres, including Wikipedia articles and supplement products customer re- views. In addition, crawling e-commerce and movie reviews websites, we compiled several sentiment analysis datasets of different genres. Our linguistic resources for Turkish also in- clude pretrained spaCy language models. To the best of our knowledge, our models are the first spaCy models trained for the Turkish lan- guage. Finally, we provide various types of education material, such as video tutorials and code examples, that can support the interested audience on practicing Turkish NLP. The ad- vantages of our linguistic resources are three- fold: they are freely available, they are first of their kind, and they are easy to use in a broad range of implementations. Along with a thorough description of the resource creation process, we also explain the position of our resources in the Turkish NLP world. 1 Introduction In recent years, with the development of transform- ers, natural language processing has experienced a dramatic breakthrough. Previously, learning archi- tectures offered state-of-art solutions to many NLP tasks, such as sequence tagging and text classifi- cation. Accordingly, data-driven approaches have become the dominant technique used to process lan- guage data. This has made availability of large and high-quality language data an essential resource for the development of NLP models. Turkish is spoken by over 80 million people, both in Turkey and across Europe, Cyprus, and Asia1. However, despite this abundance of Turkish speakers, the number of available Turkish linguistic resources does not compare to the corresponding amount of resources available for well-studied languages such as English. Turkish is an agglutinative language with complex morphology (Göksel and Kerslake, 2005), and its morphosyntactic characteristics are challenging to handle in NLP applications. Similar challenges arise in creating linguistic resources for Turkish. In this paper, we present a new set of Turkish linguistic resources, including corpora, pretrained spaCy models, and education material. The cor- pora comprise named entity recognition datasets of diverse genres, including Wikipedia articles and supplement products customer reviews. We also compiled several sentiment analysis datasets of dif- ferent genres created via crawling e-commerce and movie reviews websites. The key characteristic of our corpora is their availability, as all corpora are easily accessible via their Github repos. With re- gard to spaCy pretrained models, …

Content truncated to 500 words