Information Technology Reference
In-Depth Information
Fig. 5.7. Fact searchwiththeInFactCustom Query Generator:InFact translates
the query of Figure 5.6into theInFact Query Language (IQL) and returns a list of
results. IQL operators are fullydocumentedinthe Helppage.
roles of source (or subject), action (orverb), and target (or object). Note that
relationships, by default, aresortedbyrelevance to aquery, but can also be resorted
by date, actionfrequency,oralphabeticallybysource, action ortarget. Each ofthe
relationshipsorfacts in the table is in turn hyperlinkedto theexact locationinthe
source document where it was found,sothe user can quicklyvalidate thefindings
and explore itscontext (Figure 5.10).
Usage logs were the primary driverforthiscustomization effort. The personnel
at GlobalSecurity.org were very helpful and providedus with many monthsofuser
tra c Web logs. We wrotesomesimple scripts to analyze the logs. For example,
westudiedthe 500 mostpopular keyword searches performed onthesite rankedin
order ofpopularity.Next, we began looking for entity types that wouldbe helpful
to the most number ofusers. We found a lot ofuserinterestinweapons, terrorists,
and US o cials, amongst other things. We then setabout creatingontologies for
each ofthese areas. New custom ontologies can easilybe mappedinto the internal
InFact ontology XML format.
5.4.4 Analysis of QueryLogs
We wishto quantify the relative popularity of naturallanguage (Fact)searchversus
keyword search.In addition, we wishtocompare the relativesuccess ofalternative
strategies we adoptedtoovercome usability issues. Thisstudy of log data reflect
Search WWH ::




Custom Search