摘要
随着自然语言处理技术的不断发展和普及,语言处理领域中的马太效应越来越明显。这种现象主要表现为主流语言获得更多的资源和关注,而低资源语言则面临着信息孤岛和数字鸿沟等问题。由于ChatGPT的训练数据主要来自于主流语言,导致其在低资源语言中表现不佳,也使得低资源语言的语言处理能力和应用受到了限制,从而进一步加剧了马太效应。拥抱ChatGPT可以帮助低资源语言更好地融入数字时代的信息交流中,促进全球语言和文化的多样性。通过群智感知的方式共同推动低资源语言处理技术的进步,促进全球语言和文化的多样性和交流;同时ChatGPT也需要注重数据和信息的质量控制,以及在低资源语言中的适应性和可扩展性。ChatGPT可以促进低资源语言的信息化和自然语言处理能力的提升,但也需要认识到ChatGPT可能会加剧马太效应,因此需要采取措施确保这种技术的发展不会削弱低资源语言的地位,以期保护和推动语言多样性发展。
As the natural language processing technology continues to evolve and proliferate,the Matthew effect in the field of language processing becomes increasingly evident.This phenomenon is primarily characterized by mainstream languages receiving more resources and attention,while low-resource languages face issues such as information silos and the digital divide.Since the training data for ChatGPT mainly comes from mainstream languages,its performance in low-resource languages is subpar.This also limits the language processing capabilities and applications for low-resource languages,thereby further exacerbating the Matthew effect.Embracing ChatGPT can help low-resource languages better integrate into the information exchange of the digital era,promoting global linguistic and cultural diversity.By adopting group intelligence perception,we can jointly advance the progress of low-resource language processing technology and foster global linguistic and cultural diversity and exchange.At the same time,ChatGPT also needs to focus on the quality control of data and information,as well as its adaptability and scalability in low-resource languages.ChatGPT can enhance the informatization and natural language processing capabilities of low-resource languages,but it is also necessary to recognize that ChatGPT may exacerbate the Matthew effect.Therefore,measures need to be taken to ensure that the development of this technology does not undermine the status of low-resource languages,with the aim of protecting and promoting the development of language diversity.
作者
姚登峰
赵源
叶毓睿
饶高琦
阿布都克力木·阿布力孜
YAO Dengfeng;ZHAO Yuan;YE Yurui;RAO Gaoqi;ABULIZI Abudukelimu(Beijing Key Lab of Information Service Engineering(Beijing Union University),Beijing 100101,China;Center for Psychology and Cognitive Science,Tsinghua University,Beijing 100084,China;National Key Laboratory of High-End Server Systems,Beijing 100001,China;Metaverse Industry Working Committee of China Mobile Communications Association,Beijing 100001,China;Beijing Language and Culture University,Beijing 100083,China;School of Information Management,Xinjiang University of Finance and Economics,Urumqi 830012,China)
出处
《乐山师范学院学报》
2024年第8期36-44,共9页
Journal of Leshan Normal University
基金
国家社会科学基金一般项目“中国手语新手势构词理据及其认知神经机制研究”(21BYY106)
国家自然科学基金重点项目“语言理解的认知机理与计算机模型研究”(62036001)。