Master's Thesis Laura Altamirano Sainz

Applying lexical knowledge to improve search quality for a German legal information database

Abstract

Search support mechanisms help to accomplish more effective searches in databases.
Legal practitioners could truly benefit from these mechanisms since having good access
to legal information is crucial for them to perform their job.
To conduct a search, one needs to formulate a query to introduce into a search
system. This query can be composed of one or more terms but figuring out these
terms is not always a straightforward task. Lexical knowledge refers to the known
information about a word represented through its related terms. If we take a search
term and retrieve its lexical information, we might have an opportunity to enhance
already known search support mechanisms with it. This information can offer new
options to the users to reformulate their queries.
This thesis focuses on integrating the lexical knowledge extracted from a German
lexical database into a search support mechanism for a German-legal-information
database.
The scope of this work is narrowed to legal documents due to the advantages
acquired from them since these documents tend to be uniform and to lack of distracting
colloquialisms. Nonetheless, this does not shadow their relevance or vastness.
A needs assessment is performed to figure out the real user requirements as part of
the user-centered design approach. Also, an investigation about currently used search
support mechanisms is carried out to identify which ones are common in the legal
domain and which ones can be improved by the integration of the lexical knowledge.
Based on the gathered information, a search system is built to express the lexical
information as a query reformulation tool (i.e. a search support mechanism). For
this, three lexical relations are selected to be integrated in the system: hypernymy,
hyponymy and sibling terms. Users can interact with this information and expand or
narrow their queries in the system. Moreover, GermaNet database is chosen to serve
as the lexical source for German vocabulary and searches are performed on a set of
German law documents.
Finally, the system implementation and favorable empirical results are reported in
this thesis.