One Country, 700+ Languages: NLP Challenges for Underrepresented Languages and Dialects in Indonesia

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 7226 - 7249 May 22-27, 2022 c⃝2022 Association for Computational Linguistics One Country, 700+ Languages: NLP Challenges for Underrepresented Languages and Dialects in Indonesia Alham Fikri Aji1∗†, Genta Indra Winata2∗, Fajri Koto3∗, Samuel Cahyawijaya4∗, Ade Romadhony5,7∗, Rahmad Mahendra6,7∗, Kemal Kurniawan3,7, David Moeljadi8, Radityo Eko Prasojo9, Timothy Baldwin3,10, Jey Han Lau3, Sebastian Ruder11 1Amazon 2Bloomberg 3The University of Melbourne 4HKUST 5Telkom University 6Universitas Indonesia 7INACL 8Kanda University of International Studies 9Kata.ai 10MBZUAI 11Google Research afaji@amazon.co.uk, gwinata@bloomberg.net, fajri.phd@gmail.com, rahmad.mahendra@cs.ui.ac.id Abstract NLP research is impeded by a lack of re- sources and awareness of the challenges pre- sented by underrepresented languages and di- alects. Focusing on the languages spoken in In- donesia, the second most linguistically diverse and the fourth most populous nation of the world, we provide an overview of the current state of NLP research for Indonesia’s 700+ lan- guages. We highlight challenges in Indonesian NLP and how these affect the performance of current NLP systems. Finally, we provide gen- eral recommendations to help develop NLP technology not only for languages of Indone- sia but also other underrepresented languages. 1 Introduction Research in natural language processing (NLP) has traditionally focused on developing models for En- glish and a small set of other languages with large amounts of data (see Figure 1, bottom right). While the lack of data is generally cited as the key rea- son for the lack of progress in NLP for underrep- resented languages (Hu et al., 2020; Joshi et al., 2020), we argue that another factor relates to the diversity and the lack of understanding of the lin- guistic characteristics of such languages. Through the lens of the languages spoken in Indonesia, the world’s second-most linguistically diverse country, we seek to illustrate the challenges in applying NLP technology to such a diverse pool of languages. Indonesia is the 4th most populous nation glob- ally, with 273 million people spread over 17,508 islands. There are more than 700 languages spo- ken in Indonesia, equal to 10% of the world’s lan- guages, second only to Papua New Guinea (Eber- hard et al., 2021). However, most of these lan- guages are not well documented in the literature; many are not formally taught, and no established standard exists across speakers (Novitasari et al., ∗ These authors contributed equally. † Work done prior to joining Amazon. 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 Year 0 50 100 150 200 Number of papers Indonesian Javanese Sundanese Madurese Minangkabau Buginese Palembang Acehnese Banjarese Balinese Betawi Indonesian Javanese Sundanese Madurese Minangkabau Buginese Palembang Acehnese Banjarese Balinese Betawi Chinese Hindi Arabic Bengali Persian Japanese Punjabi Marathi Filipino Turkish Korean Vietnamese English German Russian French Italian Spanish Polish Ukrainian Romanian Dutch 0 50 100 150 Number of papers / million speakers Indonesian languages Asian languages European languages Figure 1: Following Joshi et al. (2020), we compile …

Content truncated to 500 words