1. Introduction

Introduced the application of Large Language Models (LLMs) in the field of artificial intelligence, particularly their importance in language contrastive research. Explored how multilingual models (MLLMs) can be integrated into language contrastive research, as well as their development prospects in terms of application scenarios and potential. Analyzed the difficulties of setting up "comparative third party" (TC) in language comparison research, and proposed semantic computation supported by LLMs as a new solution. Finally, the application scenarios and future development directions of LLMs in language contrastive research were discussed.

2. Issues in Comparative Language Research

Explored the issue of selecting comparative criteria (TC) in language comparative research. TC is mainly divided into two types: formal correspondence and semantic equivalence, among which formal correspondence is difficult to obtain meaningful discoveries in theory and application. Comparative language research usually observes the similarities and differences between languages based on translation equivalence, but in actual translation, there are mistranslations and individual translator style differences, resulting in few translations with strict semantic equivalence. Krzeszowski proposed that translations with high semantic syntactic correspondence should be selected as TC, but this method has limitations in languages with different morphological types. In addition, language comparison studies also attempt to seek comparability in more generalized functional equivalence by quantitatively analyzing bilingual or multilingual texts. However, the selection of TC often carries subjective judgments from researchers and lacks objectivity. To address this issue, this article proposes utilizing the knowledge application capabilities of generative artificial intelligence, particularly large language models (LLMs) and multilingual large language models (MLLMs), to provide more objective judgment criteria for TC settings and support multidimensional and multi-level language comparative analysis.

3. LLMs' multilingual semantic analysis ability

Explored in depth the capabilities of Large Language Models (LLMs) in multilingual semantic analysis. LLMs utilize self attention mechanisms to capture long-range contextual information, thereby demonstrating advantages in tasks such as translation. The self attention mechanism generates more natural and coherent language expressions by measuring the correlation between different words or markers. In addition, LLMs adopt techniques such as shared tagging, sub word propagation, and word embedding correspondence to process cross linguistic context and information, improving the accuracy of language understanding and expression.

4. Application of LLMs in Language Analysis: Scenarios

Explored the application of Large Language Models (LLMs) in the field of language analysis, including vocabulary comparison, discourse and pragmatic analysis, and discourse and rhetorical comparison. In terms of vocabulary comparison, taking ChatGPT-4 as an example, its ability to analyze the distribution of parts of speech in Chinese and English was demonstrated, revealing the differences in the use of different language means when expressing the same semantics between the two languages. Through the theory of cognitive linguistics, concepts such as conceptualization, linguistic relativity, cognitive economy, and information density have been analyzed. In terms of discourse and pragmatic analysis, LLMs can automatically identify and evaluate discourse cohesion and coherence based on specific theoretical models, such as Brown U0026 Yule's discourse analysis theory. ChatGPT-4 is capable of conducting systematic and accurate coherence analysis on given text. In terms of discourse and rhetorical comparison, LLMs are able to identify and extract rhetorical devices such as metonymy, metaphor, exaggeration, and contrast from bilingual texts, and accurately judge stylistic features such as formal/informal, lexical and syntactic variations. These abilities demonstrate higher efficiency than humans when processing large amounts of textual data.

5. A New Paradigm for Comparative Research under the Large Language Model

Explored the application of Large Language Models (MLLMs) in language contrastive research and proposed a new research paradigm. The main content includes:

6. Conclusion

Emphasis was placed on the importance of large language models in language comparison research, pointing out that they can achieve quantifiable comparative analysis between multiple languages and provide judgment results through fine-tuning models. With the popularization of big language models, the sources of parallel corpora will become more diverse, providing multidimensional data for research. In addition, the big language model will promote language comparative research based on semantic correspondence rather than translation equivalence. Its value lies not only in knowledge query and writing assistance, but also in assisting humans in multidisciplinary and multi perspective intellectual activities, realizing the synergy of text analysis, generative artificial intelligence analysis, and human intelligence analysis, and promoting language research to a higher level.