Optimization of an Engineering Information Retrieval System using Topic Models and Knowledge Graphs

The oil and gas industry is in the top five largest sectors in the world in terms of dollar value, generating an estimated $3.32 trillion in revenue annually. It is estimated that 40-60% of workforce in the oil and gas industry will retire in the next five years, so preserving the knowledge stored in documents is an important objective. The industry partner (WESI — Waterford Energy Services Inc.) has been developing a document management system to support this objective. In this project we focus on enhancing IR system for historical well reports by exploring different topic modelling and ontology building approaches. Topic modelling (TM) has a wide range of applications in domains such as social media, microblogging data, historical documents, biomedical data, etc. However, the practice has shown that there is no “one-fits-all” algorithm. Standard models usually do not perform well in domain- specific settings and our goal is to evaluate and develop recommendations for best-fitting TM algorithms for the well reports.

Faculty Supervisor:

Vlado Keselj

Student:

Goutham Narravula

Partner:

Waterford Energy Services Inc

Discipline:

Computer science

Sector:

Construction and infrastructure

University:

Dalhousie University