Beyond Keywords: Semantic Search Framework for Data in Organizations

Next to its essential role of supporting operational and decisional business activities, data also has economic significance. The desire to seek maximum value from their data assets prompts organizations to implement different infrastructures, architectures, governance, and security to facilitate creating and storing huge volume and variety of data. However, these implementations come with challenges as the data is not often the main focus, and this usually makes data management and discovery difficult. The data mesh paradigm advocates for data to be decomposed around domains and served as a product for use by data users. Having data as products is posited to enable ease in data management and discovery. This also enables the application of semantic searches to data ecosystem. Integrating recent advances in artificial intelligence into semantic search mechanisms makes them great candidates to enable business to derive optimal value from their data assets. In this regard, our project’s goal is to develop a framework that utilizes ontology embeddings, vector search, and large language models (LLMs) to improve data discovery and management using semantic search. The result will be capabilities that enable organizations to manage their data products through more intelligent and customized searches of their data.

Faculty Supervisor:

Daniel Amyot

Student:

Partner:

Accenture Inc

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

University of Ottawa

Program: