Computer Sciences and Information Technology
Title : | Information Retrieval Via Knowledge Graphs Developed for Aircraft Accidents Database and Aircraft Manuals |
Area of research : | Computer Sciences and Information Technology |
Focus area : | Development of Information Retrieval Platform |
Principal Investigator : | Prof. Pushpak Bhattacharyya, Indian Institute of Technology (IIT) Patna |
Timeline Start Year : | 2019 |
Contact info : | pb@cse.iitb.ac.in |
Details
Executive Summary : | Knowledge Graphs are becoming increasingly important in knowledge and data management applications as they afford a semantic structure to the underlying data. They form crucial components of modern web search engines; state-of-the-art question answering systems and are used in a variety of applications especially in finance and healthcare. Knowledge Graphs give a method of representation of the contained knowledge in an interactive, learning and maintainable knowledge layer. Challenges are to tackle the end-to-end extraction and cognitive representation of documents in two related domains. The domains of interest in this project are Aircraft Accident Report Domain and the Aircraft Operational, Maintenance Manual Domain. The information in these domains is largely textual with images and tables. NLP Techniques can be exploited to organize information from these domains. One of the methods is to represent and organize information as knowledge graphs. The first phase of the project shall result in separate knowledge graphs in each domain – the Aircraft Accident Report Domain and the Aircraft/Subsystem Manual Domain. The second phase of the project shall attempt to build an information retrieval platform that would relate these two knowledge graphs for the given set of query/response evaluation sets or use the two knowledge graphs to generate common safety and operational insights for a given set or even generate a dashboard of major clusters of information. Phase 3 of the project shall explore optimization and pruning techniques and methods to add new data sets in both domains. The information retrieval platform will score some of the results of various algorithms and techniques based on a set of evaluation questions and answers. These methods and techniques could be used/reapplied for alternate domains like industrial safety, environment science, Transportation and many more. Honeywell, an industry partner shall benefit in using the information retrieval platform (The deliverable) in analyzing existing aircraft manual effectiveness related to safety domain information and derive strategies to improve safety and operational efficiency of their products, solutions or services via improvement of manuals, refinement in system design, training, maintenance or customer support procedures. |
Total Budget (INR): | 54,00,560 |
Achievements : | The work is in the early stages. Knowledge graph construction and use in a domain is the main research done. The methods are translatable to other domains. Publications and patents based on the methodology developed will be made |
Organizations involved