Simplifying syntactic and semantic parsing of NL based queries in advanced application domains
ISSN der Zeitschrift
Natural language processing and information systems
Regular Research Papers
Gesellschaft für Informatik e.V.
The paper aims at presenting a natural (sub)language based querying approach (MDDQL) for SQL (relational, object-relational) databases, which relies on an ontology driven, interactive query construction mechanism. This guides the user to the construction of queries that are semantically compliant with the application domain semantics. To this extent, syntactic and semantic parsing of a query is done implicitly, during the query construction, rather than syntactically and semantically parsing the query after its formulation. Given also that the vocabulary terms are represented as objects having properties and not as simple natural language words, it is possible to cope with the intentional meaning of terms as well as with homonyms, i.e., the same word meaning different things, during the query construction. In addition, it is possible to express the query terms in a different natural language without changing or providing another syntactic and semantic parsing mechanism. Therefore, the generated SQL-statement reflects not only the application domain semantics but is also identically inferred from multi-lingual queries. This querying approach is currently applied to a Swiss national registry, which consists of, at the moment, twelve data repositories of patients' records, but it can also be applied to any scientific or technical domain with an advanced and hardly understood terminology.