Exploring the Linguistic Capabilities and Limitations of AI for Endangered Language preservation

Abstract

The main purpose of the present study was to investigate the crossway of semantics and computational linguistics, aiming how methods of computational linguistics support to understand the process of modeling meaning in natural language. This study opts a qualitative and exploratory approach to examine the capability of AI language models to cater the dimensions of semantics and pragmatics in the context of language documentation. For the current study, the data was collected through the different sources such as real world linguistics examples to understand the nature of the data. The published documentation of under sourced languages, such as digital archive (ELAR, DOBES) grammar and fields notes were also became the source for collecting the data. The data was analyze through thematic linguistic analysis. The study opted semantic and evaluation criteria to check the lexical accuracy in the form of synonyms and antonyms use. For pragmatic, the pragmatic evaluation was done to check the contextual interpretation and appropriateness of speech act. The study finds gap between semantic and pragmatic modeling in AI language system, having a greater disparity evident in under documented languages. On the basis of these findings researchers presented some recommendations for future researchers and scholars.