Tonality analysis as a financial analysis tool: Problems and prospects

Subject. The article deals with the analysis of the tonality of annual reports of companies. Objectives. The purpose is to devise a methodology for integrating the analysis of unstructured textual information into traditional financial analysis, using artificial intelligence technologies. Methods. I employed NLP technologies for text preprocessing, including tokenization, lemmatization, and removal of stop words, using my specialized vocabulary of tonality I developed for the energy industry (766 terms). To do this, I used automated data collection and analysis tools based on uploaded Excel and Word reporting files from the websites of the Federal Tax Service of Russia and Interfax, including AI Claude Sonnet 4, Google Colab, and SerpAPI. I also performed a comparative analysis of results of traditional financial analysis, the tone of corporate reporting, and external media sources. Results. Using the PAO LESK case, the paper presents advantages and disadvantages of integrating the analysis of annual financial statements’ text part tonality into the traditional financial analysis methodology. It identifies discrepancies between objective financial indicators and moderately positive tone of corporate reporting, on the one hand, and between the tone of corporate reporting and negative tone of external sources, on the other hand. This indicates problems with communication efficiency with external stakeholders. Conclusions. The inclusion of sentiment analysis in the traditional financial analysis methodology expands the analysis context available to analysts and enables to assess the presence or absence of potential problems related to corporate governance, reputational risks, and effectiveness of corporate communications. The results require validation on a broader sample of enterprises.