From Words to Meaning

The article explores the application of large language models for interpreting the results of topic modeling in historical texts. The author proposes a novel approach to address the issues of subjectivity and labor-intensiveness associated with traditional methods of interpreting topics obtained through machine learning algorithms. The study uses the mid-19th century diary of gymnasium student V. I. Chemezov as its primary material. The methodology includes constructing a topic model of the diary using the Latent Dirichlet Allocation algorithm, developing a specialized prompt for language models, and conducting a comparative analysis of interpretations obtained from various language models (GPT 4, Claude, Llama 3.1). The results demonstrate the high effectiveness of the proposed approach: language models successfully interpret complex and ambiguous topics, taking into account the historical context and emotional subtext of the entries. The author notes that using language models significantly accelerates the process of topic interpretation while maintaining high-quality analysis. The article also discusses the limitations of the method. The author emphasizes that the proposed approach does not replace expert knowledge of historians but complements researchers' toolkit, creating a new paradigm in historical research where artificial intelligence technologies act as intellectual assistants. The study concludes that the development of this approach could lead to significant progress in the methodology of historical research, opening new possibilities for analyzing and interpreting historical sources in the digital age. The author highlights the potential of this method to overcome subjective biases in interpretation and its ability to process large volumes of textual data efficiently. By combining the strengths of machine learning algorithms with the contextual understanding capabilities of large language models, this approach offers a promising avenue for enhancing the depth and breadth of historical analysis in the era of digital humanities

Epistemology · Linguistics · Meaning (existential · Natural language processing · Computational and Text Analysis Methods · Computer Science · Philosophy