2025 ASEE Annual Conference & Exposition

Instructors and students alike face challenges when it comes to acquiring targeted, high-quality educational materials for their courses. Both instructors and students often benefit from well-structured, customized notes or textbook material that aligns closely with the teaching and learning objectives. Nevertheless, textbook focus may not align with specific course objectives and finding relevant material within them can be cumbersome. Instructors may need to tailor their courses on short notice to address student feedback or emerging needs, while students often seek supplementary reading to deepen their understanding of specific topics. Both scenarios require quick and precise access to relevant educational resources, but manually searching through large volumes of textbook material is time-consuming and inefficient.

Online methods can provide a faster searching experience by recommending materials based on the user’s specific queries. Existing online platforms, such as library websites, recommend books using keyword search. While keyword search can be effective, it is limited to direct matches or predefined synonyms, often resulting in irrelevant results due to typos or misworded queries. Furthermore, users may not always know the precise terms they are looking for but still seek related concepts. To address this limitation, semantic search is a potential solution.

TextMatch is an implementation of semantic search on a large existing database of books. Through contextual interpretation of user queries, TextMatch reranks results from an initial broad keyword search and returns a refined list of recommendations. This refinement allows for more accurate and meaningful results compared to what would be returned by a simple keyword search mechanism.

TextMatch was evaluated on a variety of metrics, which were compared to those from a search algorithm that only performed keyword search. Evaluation of TextMatch demonstrated improved Precision@10 and Mean Reciprocal Rank (MRR) scores compared to basic keyword search, while maintaining low search times. This highlights the significance of incorporating semantic search capabilities to provide students and instructors with contextually relevant educational materials that enhance the learning experience.