Abstract:

The integration of Artificial Intelligence (AI) into higher education is redefining pedagogical methodologies, facilitating an effective transition toward technology-enhanced self-learning models. This study investigates the feasibility of transforming recorded theoretical sessions at the Escuela Politécnica Superior de Alcoy (EPSA), part of the Universitat Politècnica de València (UPV), into high-quality technical study materials through the use of automated transcription tools and advanced language models.

Historically, speech transcription has evolved from its origins in the 1950s and the 1970s 'Harpy' system to the current era driven by neural networks and high-capacity processors. However, raw literal transcription often exhibits deficiencies that limit its pedagogical utility. This paper analyzes a theoretical session of approximately 2 hours to evaluate the efficacy of AI in enhancing and structuring content. The methodology employed consisted of a sequence of tasks, including: audio capture and extraction; transcription via Microsoft 365; manual error review; orthographic and structural correction performed by AI; formalization of the linguistic register; and, finally, the generation of a technical glossary.

Initial results revealed that the automated transcription generated a document of 12,887 words with an error rate of 2.09% (269 errors). Although this percentage appears low, the lack of punctuation and the presence of unintelligible paragraphs increased the human review time to 5 hours—an unsustainable cost for educators. By using specific prompts, the AI managed to drastically reduce the number of errors to only 10 and synthesized the document into 3,479 words, eliminating filler words and the colloquial language used by the lecturer.

This post-processing workflow significantly reduced the faculty's review time, decreasing from 300 minutes to just 40 minutes. This optimization allows the content of a two-hour lecture to be assimilated in only 15 minutes of structured technical reading. It is concluded that AI acts not only as a corrector but as an academic editor capable of hierarchizing knowledge and elevating the linguistic standards of reference materials. The study demonstrates that the synergy between institutional recording systems and AI is an essential tool for educational innovation, facilitating the creation of accurate, formal, and accessible notes that reinforce students' autonomous learning.

Keywords:

Artificial Intelligence, Teaching Innovation, Automatic Transcription, Natural Language Processing, Learning Materials, Higher Education.