Semantic MemoryKnowledge Acquisition ThroughActive Dialogues

Many language-oriented problemscannotbe solved without semantic memorycontaining descriptions of concepts atdifferent level ofdetails. Automatic creation of semantic memories isa greatchallenge evenforthesimplest knowledge representation methodsbasedonrelations between concepts andkeywords. Semantic memorybasedon such simple knowledge representation facilitates implementation of quite interesting linguistic competences thathavenotyetbeen demonstrated bymoresophisticated ruleorframe-based knowl- edgebases, forexample CYC.Theselinguistic abilities include wordgames,suchasthetwentyquestions game,thatmaybe implemented using semantic memorybuilt onrelational model forknowledge representation. Creation oflarge-scale knowledge baseforsemantic memoryinvolves miningstructured infor- mationsources (ontologies, dictionaries, encyclopedic entries) andfreetexts (textbooks andinternet sources). Quality ofthis knowledge may beimproved usingcollaborative projects in whichsystems thatalready possess somelinguistic competence actively interact withhumanusers, mining their knowledge. In this article three dialog scenarios formining humanknowledge areintroduced, andthedataacquired intosemantic memory structures through suchinteraction isdescribed.