Automated Rationalization of Subsurface Software Portfolios in Oil & Gas Using Fine-Tuned Sentence-Similarity Transformers

This paper presents an innovative approach for optimizing application landscapes in the oil and gas industry, employing fine-tuned sentence similarity algorithms to rationalize subsurface software portfolios. We developed a unique framework that integrates linguistic analysis with industry-specific nuances, addressing gaps identified in current methodologies. Through a meticulous process of data collection, preprocessing, and algorithm selection, we tailored our approach to effectively process technical language specific to the oil and gas sector. The results from our experiments, which involved rigorous testing and training phases, demonstrated a high degree of accuracy in identifying redundant and overlapping applications, with a ~7% increase in Spearman Correlation post fine-tuning. A case study application in a real-world scenario illustrated the practical benefits of our method, showing a potential of up to ~25% reduction in software redundancies and an improvement in IT resource allocation. Our approach not only enhances the efficiency of software utilization but also sets a new precedent in applying natural language processing techniques for IT portfolio management in the oil and gas industry.