This research investigates the feasibility of a pre-processing framework designed to enhance text information extraction (TIE) from source code using the T5 (Text-to-Text Transfer Transformer) model. Through systematic analysis of various pre-processing techniques, novel approaches that improve the accuracy and efficiency of T5-based TIE are identified. It involves a comprehensive review of existing methods, empirical analysis using benchmark datasets, and the development of a novel pre-processing framework. The result indicates improvements in code summarization, noise reduction, and tokenization. The proposed framework demonstrates enhanced readability, maintainability, and productivity in software development processes. By automating documentation and simplifying code reviews, it facilitates better onboarding of new developers and efficient bug fixing. This research underscores the importance of pre-processing in improving machine learning models’ performance and offers a robust solution for handling the complexities of text extracted from diverse codebases.