Low-resource languages like Sundanese continue to challenge Automatic Speech Recognition (ASR) systems owing to dialectal variation, limited data, and phonetic variance, which lower transcription quality. This research proposes a Type-2 fuzzy logic-based post-processing system to increase transcription accuracy in such cases. First, we transcribe using Whisper, a cutting-edge multilingual ASR model. These outputs are improved by fuzzy language rules based on regular Sundanese phonological and morphological faults. In particular, our technique addresses ASR ambiguity for Sundanese words with complicated meanings, consonant alterations, and reduplications. Interval-valued membership functions let the Type-2 fuzzy system turn approximation or uncertain phrases into more correct linguistic ones. The recommended correction approach decreases Word Error Rate (WER) by 25% on average in 30 typical samples using a publicly accessible Sundanese speech dataset. This method is unique in its interpretable and rule-extendable design. It's ideal for underrepresented languages, unlike post-editing or statistical adjustments. This work supports language inclusion in speech technologies by showing how fuzzy logic in the ASR pipeline may enhance transcription quality in circumstances with little linguistic data, and promotes speech technology linguistic inclusion.