From words to paragraphs: modeling sentiment dynamics in ‘notes from underground’ with GPT-4 via descriptive methods and differential equations

This study examines how the sentiment values in the first part of the book entitled as “Underground” of Fyodor Dostoevsky’s “Notes from Underground” change from words to sentences to paragraphs. Using the GPT-4 language model, we conducted a descriptive analysis of standardized sentiment values and calculated cumulative sentiment trajectories over the text. We then created differential equation models to model the sentiment tones using regression analysis. Our findings suggest that sentiment becomes less negative from words to paragraphs, indicating that context moderates negativity. Paragraph sentiment was also more stable with lower variability. There was a narrative arc of initial decline followed by an upward turn in sentiment. Paragraphs had the highest baseline sentiment, suggesting that they are able to capture more nuanced context. Paragraphs lost short-term sentiment quickly but retained longterm sentiment longest, aligning with paragraphs maintaining overall text sentiment over time. These findings suggest that there are complex dynamics between linguistic units contributing to perceived stability of sentiment. Quantitative decay rates are useful indicators but do not fully characterize sentiment stability.

1. Misuraca M., Forciniti A., Scepi G., Spano M. (2020) Sentiment Analysis for Education with R: packages, methods and practical applications. arXiv:2005,12840.

2. Dietz-Uhler B. & Hurn E.J. (2013) Using learning analytics to predict (and improve) student success: A faculty perspective. Journal of Interactive Online Learning, 12, pp. 17–26.

3. Liu B. (2012) Sentiment Analysis and Opinion Mining. https://www.cs.uic.edu/~liub/FBS/SentimentAnalysis-and-OpinionMining.pdf.

4. Wankhade M., Rao A.C.S. & Kulkarni C. (2022) A survey on sentiment analysis methods, applications, and challenges. Artif Intell Rev 55, pp. 5731–5780. https://doi.org/10.1007/s10462-022-10144-1.

5. Xing FZ, Cambria E., Welsch RE (2018) Natural language based financial forecasting: a survey. Artif Intell Rev 50(1), pp. 49–73.

6. Collomb A., Costea C., Joyeux D., Hasan O., Brunie L. (2014) A Study and Comparison of Sentiment Analysis Methods for Reputation Evaluation. Rapport de Recherche RR-LIRIS-2014-002.

7. Hemmatian F., Sohrabi MK (2019). A survey on classification techniques for opinion mining and sentiment analysis. Artif Intell Rev 52(3), pp. 1495–1545. https://doi.org/10.1007/s10462-017-9599-6.

8. Pontiki M. et al. (2016) SemEval-2016 task 5: aspect based sentiment analysis. In: Proceedings of the 10th international workshop on semantic evaluation (SemEval-2016), Association for Computational Linguistics, San Diego, CA, pp. 19–30.

9. Yadav A., Vishwakarma DK (2020) Sentiment analysis using deep learning architectures: a review. Artif Intell Rev 53(6), pp. 4335–4385.

10. Do H.H., Prasad P., Maag A., Alsadoon A. (2019) Deep learning for aspect-based sentiment analysis: a comparative review. Expert Syst Appl 118, pp. 272–299.

11. Zhang L., Wang S., Liu B. (2018) Deep learning for sentiment analysis: a survey. Wiley Interdiscip Rev 8(4), pp. e1253.

12. Maglogiannis I. et al. (2020) A deep learning approach to aspect-based sentiment prediction. In: Artificial intelligence applications and innovations16th IFIP WG 125 international conference, AIAI 2020, Neos Marmaras, Greece, June 5–7, 2020. Proceedings, Part I 583, pp. 397–408.

13. Meškelė D., Frasincar F. (2020) Aldonar: a hybrid solution for sentence-level aspect-based sentiment analysis using a lexicalized domain ontology and a regularized neural attention model. Inf Process Manag 57(3), pp. 102211. https://doi.org/10.1016/j.ipm.2020.102211.

14. Schouten K. & Frasincar F. (2018) Ontology-driven sentiment analysis of product and service aspects. In 15th Extended Semantic Web Conference (ESWC 2018), LNCS, Springer International Publishing ,vol. 10360, pp. 608–623.

15. Birjali M., Kasri M., Beni-Hssane A. (2021) A comprehensive survey on sentiment analysis: approaches, challenges and trends. Knowl-Based Syst 226, pp. 107134.

16. Leippold, M. (2023) Sentiment spin: Attacking financial sentiment with GPT-3. Finance Research Letters. https://doi.org/10.1016/j.frl.2023.103957.

17. Fan, David P., Cook, R. Dennis (2003) A differential equation model for predicting public opinions and behaviors from persuasive information: Application to the index of consumer sentiment. The Journal of Mathematical Sociology, 27(1), pp. 29–51. doi:10.1080/00222500305886.

18. Casillo M. et al. (2020) Chat-bot: a cultural heritage aware teller-bot for supporting touristic experiences. Pattern Recogn Lett 131, pp. 234–243. https://doi.org/10.1016/j.patrec.2020.01.003.

19. Chang M. et al. (2020) Building ontology-driven tutoring models for intelligent tutoring systems using data mining. IEEE Access 8, pp. 48151–48162. https://doi.org/10.1109/ACCESS.2020.2979281.

20. Colace F. de Santo M., Greco L. (2014) Safe: a sentiment analysis framework for e-learning. Int J Emerg Technol Learn 9(6), pp. 37–41. https://doi.org/10.3991/ijet.v9i6.4110.

21. D’Aniello G., Gaeta M. & La Rocca I. (2022) KnowMIS-ABSA: an overview and a reference model for applications of sentiment analysis and aspect-based sentiment analysis. Artif Intell Rev 55, pp. 5543–5574. https://doi.org/10.1007/s10462-021-10134-9.

22. Chen Z., Liu Y. & Sun H. (2021) Physics-informed learning of governing equations from scarce data. Nat Commun 12. https://doi.org/10.1038/s41467-021-26434-1.

23. Belsley D.A., Kuh E. & Welsch R.E. (1980) Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York: Wiley.

24. Alpar R. (2022) Uygulamalı İstatistik ve Geçerlik Güvenirlik. Ankara: Detay Yayıncılık.

25. Schmidt M.D. & Lipson H. (2009) Distilling free-form natural laws from experimental data. Science 324, pp. 81–5.

26. Xue L. & Zou H. (2006) Parameter Estimation for Differential Equation Models Using a Framework of Measurement Error in Regression Models. Journal of the American Statistical Association, 101(475), pp. 1570– 1583. https://doi.org/10.1198/016214506000000691.