Information Extraction and Anomaly Detection from Turkish Financial Reports' Footnotes T rk Finansal Raporlari Dipnotlarindan Bilgi ikartma ve Anomali Tespiti

Atasever O., Yesilyurt S., Akgul Y. S., Aydemir S.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Türkiye, 25 - 28 Haziran 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu66497.2025.11111996
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: Anomaly Detection, Footnote Information Extraction, Natural Language Processing, Question-Answering Model, Random Forest, XGBoost
  • Sivas Cumhuriyet Üniversitesi Adresli: Evet

Özet

Financial statement footnotes contain critical information that complements numerical data; however, their text-heavy nature often leads to them being overlooked. This study aims to extract information from the financial footnotes of publicly traded companies listed on Borsa Istanbul to support financial analysis processes and detect anomalies. Within the study, 10 different questions were answered - five using natural language processing-based question-answering models and five using machine learning techniques. Various techniques were applied to analyze the consistency of financial data and detect anomalies, but challenges such as OCR errors, table structure loss, and API costs were encountered. The results demonstrate that NLP and ML methods are effective in financial text analysis and contribute to risk assessments by identifying anomalies.