Iterative Ontology Development Tool
  • Agency: NASA Ames SBIR Phase I
  • Duration: January 2004 - June 2004.

   Capabilities Being Demonstrated:
  • Extracting ontological concepts from natural language text
  • Bootstrapping support for ontology creation and reuse

Under this SBIR project we developed a prototype semi-automated, cluster-based information extraction tool that enables analysts to rapidly create structured representations of the information present in stylized natural language text, thereby rendering the content accessible to semantically aware query, manipulation, and reasoning tools.

Large amounts of unstructured but stylized natural language text exist in today's data and knowledge repositories. While simple text-based search techniques can often retrieve documents that are at least somewhat relevant to a human analyst's needs, the information contained in those documents is effectively opaque to knowledge-based systems and databases. This means that the information is also inaccessible to analysts who wish to employ query, manipulation, and reasoning tools based on the semantic content of the text.

IOD focuses on providing an infrastructure to aid iterative building of ontologies from existing semi-structured aviation-related incident/accident reports, in a manner that can enable automated analysis of these reports in the long-term. A number of airlines have Flight Operational Quality Assurance (FOQA) programs that analyze archived flight data. Although this analysis process is extremely useful for assessing airline concerns in the areas of aviation safety, operations, training, and maintenance, looking at flight data in isolation does not always provide the context necessary to support a comprehensive analysis. To improve the analysis process, the Aviation Data Integration Project (ADIP) at NASA has focused on developing techniques for integrating flight data with auxiliary sources of relevant aviation data. ADIP has developed an aviation data integration system (ADIS) comprised of a repository and associated integration middleware that provides rapid and secure access to various data sources, such as, weather data, airport operating condition (ATIS) reports, radar data, runway visual range data, and navigational charts. Significant challenges exist in analyzing the various sources of aviation data because of the significant volume and transient nature of the data, as well as difficulties in correlating data across multiple sources.

In Phase I of this project we successfully explored methods and techniques for building an ontology for supporting aviation-related problem-solving activities, such as, incident and accident investigation and safety assessment. A wide variety of information sources, such as, safety reports, pilot logs, weather data, air traffic control information, flight recorder data, maintenance data, and FAA advisories are typically referenced for these problem-solving activities. Since a single unified resource does not exist to perform the various problem-solving activities, it is necessary to build an infrastructure to extract an aviation-related ontology across these resources so that automated support can be provided for the analysts and they can perform their job more efficiently. Knowledge inside and across such systems need to be captured in ontologies, at appropriate abstraction levels and in a reliable manner, so that they can have automated reasoners to analyze the data automatically and allow domain experts to interpret the analysis results holistically. The prototype “Iterative Ontology Development” (IOD) toolkit can aid creation of aviation ontology from various aviation-related data sources.

The difference between existing approaches to building an ontology and our approach lies in the fact that we are using preexisting resources to build the ontology. Our reverse engineering approach promises to (a) extract concepts and their subtle relationships in the context of their usage (b) save an ontology designer the time and effort of building the ontology from scratch by giving him a jump start from the existing documents, and (c) tailor the ontology to his/her viewpoint in an iterative fashion. Besides, it addresses a cost benefit aspect by reusing existing artifacts and increasing their utilization. IOD demonstrates the feasibility of building a semantic mediation toolkit that focuses on providing various types of ontological engineering aids during knowledge entry, which can facilitate automated analysis of large knowledge repositories thus leading to long-term reliability of NASA systems.

« Back | Top
©2007 Pragati Synergetic Research, Inc. Research Focus | Products | Projects | Publications | Market Sectors | About Us