Natural Language Processing
Natural Language Techniques in Support of Spacecraft Design (internally called "SHUMI - Support to HUman Machine Interaction") was the first NLP-related Ariadna with the contribution of the University of Roma Tor Vergata.
The aim of the study was to assess and produce an "intelligent advisor" tool for spacecraft designers. The advisor would capture pieces of information needs during the design process, and it would retrieve connecting/relevant data by browsing both internal and external sources of information. The main study tasks were the definition of a suitable system architecture and identification of its main components, and the identification and representation of the spacecraft ontology (i.e. the "knowledge" of the particular domain).
Summary of results
The research team of Roma Tor Vergata created two study description materials, the "Functional Architectural Proposals" and the "Detailed Study" reports. The former consists of the brief description of an architecture found to fit to the study requirements, while the latter, detailed one (including also the architectural proposal itself) gives overviews on the various information access systems, methodologies of syntactic and semantic analysis, methods of ontology learning, and available technological solutions as the future parts of the proposed modular system.
The core part of the suggested architecture is an Information Retrieval (IR) engine and a Document Clusterer. The latter module would provide a topical representation of the retrieved results. An Information Need Extractor would be used to get the information needs by processing implicit or explicit requests (e.g. conversations, study documentations, explicit queries of experts) and to feed them into the IR module. The main modules would perform their tasks by using the mission ontology together with generic linguistic knowledge bases.
During the present study period the first steps of creating the mission knowledge base (KB) were carried out. Possible relevant terms and verbal expressions were extracted from a corpus provided by ACT, and the extracted results were validated (filtered) by experts from ACT, ESA Headquarter and CDF. To make the validation task easier, the research group of Roma Tor Vergata supplied the experts with a validator interface software.
The study responsibles would like to thank all the experts for their contribution to the validation tasks.
Links & references
- Pazienza M. T., Pennacchiotti M., Vindigni M., Zanzotto F. M., "Natural Language Techniques in Support of Spacecraft Design," European Space Agency, the Advanced Concepts Team, Ariadna Final Report 03-5101, 2004.