DEVELOPMENT OF A SOFTWARE PACKAGE FOR PROBLEMATIC TEXT DATA BASED ON NEURO-FUZZY MODELING

Аннотация

Textual data from sources such as social media, messaging, and OCR often contain problematic elements – they are noisy, ambiguous, or incomplete – posing challenges for traditional natural language processing (NLP) techniques. This paper presents a hybrid neuro-fuzzy modeling approach to robustly process such problematic text data. We combine fuzzy logic’s strength in handling uncertainty and imprecision with neural networks’ learning capability to create a software system that can normalize noisy text, interpret ambiguous language, and make reliable decisions even with incomplete information. We outline the architecture of the proposed system, which integrates fuzzy inference modules (capturing expert linguistic rules and similarity measures) with trainable neural network components that adapt these rules to data. Practical examples demonstrate how the system corrects slang and spelling variations and resolves ambiguities incontext. Keywords: neuro-fuzzy modeling, fuzzy logic, noisy text processing, text normalization, natural language processing (nlp), textclassification, interpretability.