Risk Identification in the Electronic Information Manufacturing Industry Using Annual Report Texts: A Comparative Analysis of LDA and BER Topic Topic Modeling Methods

With the rapid technological evolution of the electronic information manufacturing industry, intensified market competition, and increasing complexity of global industrial and supply chain environments, enterprises are facing increasingly diversified types of risks. Annual reports of listed companies contain abundant unstructured textual information related to operational risks, providing an important data foundation for industry risk identification. This study investigates risk-related texts extracted from the annual reports of listed electronic information manufacturing companies in the Shanghai and Shenzhen A-share markets during 2021–2024. A risk text corpus was constructed, and based on unified text preprocessing procedures, the Latent Dirichlet Allocation (LDA) model and BERTopic, a topic modeling approach based on pretrained language models, were employed to identify and comparatively analyze risk topics. The differences between the two models were evaluated from the perspectives of topic identification results, topic quality, and interpretability.The results show that LDA generates a relatively concentrated risk category structure and identifies seven risk topics, including macroeconomic risk, investment risk, market competition risk, supply chain risk, core talent risk, accounts receivable risk, and operational management risk, making it suitable for analyzing the overall risk structure of the industry. In contrast, BERTopic further refines risk information through contextual semantic representations and identifies 28 valid risk topics. It outperforms LDA in terms of both topic coherence and topic diversity, enabling more accurate differentiation of specific risk issues such as technological R&D risk, core talent loss risk, exchange rate fluctuation risk, and raw material price fluctuation risk. The findings demonstrate that different topic modeling approaches have distinct applicable scenarios for risk identification: LDA is more suitable for macro-level risk classification, whereas BERTopic is more effective for fine-grained risk discovery. This study provides methodological insights for risk text analysis in the electronic information manufacturing industry and offers a foundation for enterprises to establish a hierarchical risk identification framework of “overall risk categories–specific risk issues.”

Electronic Information Manufacturing Industry: Risk Identification; Annual Report Text; Topic Modeling; LDA; BERTopic

1. SODHI M S, LEE S. An analysis of sources of risk in the consumer electronics industry[J]. Journal of the Operational Research Society, 2007, 58(11): 1430-1439. [Google Scholar] [Crossref]

2. LIAO Z W, TANTAI B Y, ABDUL-HAMID A Q, et al. Exploring resilience in the downstream supply chain of the semiconductor industry: The mediating roles of risk mitigation, process simplification, and flexibility[J]. International Journal of Production Economics, 2025, 281: 109530. [Google Scholar] [Crossref]

3. MOKTADIR M A, REN J. Global semiconductor supply chain resilience challenges and mitigation strategies: A novel integrated decomposed fuzzy set Delphi, WINGS and QFD model[J]. International Journal of Production Economics, 2024, 273: 109280. [Google Scholar] [Crossref]

4. CAMPBELL J L, CHEN H, DHALIWAL D S, et al. The information content of mandatory risk factor disclosures in corporate filings[J]. Review of Accounting Studies, 2014, 19(1): 396-455. [Google Scholar] [Crossref]

5. JIN S C, LEE H W, XIE Y A. Risk disclosures in annual reports and firms' default risk[J]. Journal of Corporate Accounting & Finance, 2024, 35(3): 192-212. [Google Scholar] [Crossref]

6. DYER T, LANG M, STICE-LAWRENCE L. The evolution of 10-K textual disclosure: Evidence from Latent Dirichlet Allocation[J]. Journal of Accounting and Economics, 2017, 64(2-3): 221-245. [Google Scholar] [Crossref]

7. ZHU X Q, WU H D, CHANG Y P, et al. Accounting fraud detection through textual risk disclosures in annual reports: From the perspective of SEC guidelines[J]. Accounting & Finance, 2025, 65(2): 1837-1862. [Google Scholar] [Crossref]

8. KOBEISY A N, ROBERTS R W. Corporate governance and risk disclosure: Evidence from textual analysis of annual reports. Journal of Accounting and Public Policy, 2020, 39(6): 106753. [Google Scholar] [Crossref]

9. BLEI D M, NG A Y, JORDAN M I. Latent Dirichlet allocation[J]. Journal of Machine Learning Research, 2003, 3: 993-1022. [Google Scholar] [Crossref]

10. BIANCHI F, TERRAGNI S, HOVY D. Pre-training is a hot topic: Contextualized document embeddings improve topic coherence[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 2: Short Papers). Online: Association for Computational Linguistics, 2021: 759-766. [Google Scholar] [Crossref]

11. DIENG A B, RUIZ F J R, BLEI D M. Topic modeling in embedding spaces[J]. Transactions of the Association for Computational Linguistics, 2020, 8: 439-453. [Google Scholar] [Crossref]

12. GROOTENDORST M. BERTopic: Neural topic modeling with a class-based TF-IDF procedure[EB/OL]. (2022-03-11)[2026-10-06]. [Google Scholar] [Crossref]

13. REIMERS N, GUREVYCH I. Sentence-BERT: Sentence embeddings using Siamese BERT-networks[C]//Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. Hong Kong, China: Association for Computational Linguistics, 2019: 3982-3992. [Google Scholar] [Crossref]

14. ANGELI G, MANANDHAR S, NEPAL S.Topic modeling using contextual embeddings: A comparative analysis of traditional and neural approaches. Information Processing & Management, 2022, 59(6): 103091. [Google Scholar] [Crossref]

15. GAD M, PARK G, RAWSTHORNE S, et al. When methods matter: How implementation choices shape topic discovery in financial text[J]. Accounting and Business Research, 2026. [Google Scholar] [Crossref]

16. ABDELRAZEK A, EID Y, GAWISH E, et al. Topic modeling algorithms and applications: A survey[J]. Information Systems, 2023, 112: 102131. [Google Scholar] [Crossref]

17. MISHRA M. Comparative analysis of BERTopic versus LDA: Topic modelling in marketing research[J]. Australasian Marketing Journal, 2026, 34(2): 162-173. [Google Scholar] [Crossref]

• Students’ Cognitive and Affective Responses to the National Education Policy 2020: Evidence from Indian Higher Education Institutions
• Rice Husk Ash as a Suitable Aggregates for Concrete Production and Sustainable Building Material in Enugu State Urban, Southeast Nigeria
• Digital Marketing: Examining Students’ Purchase Decision-Making in Private Universities