Natural Language GenerationSlide 2NLG – Application TypesCanned TextTemplate FillingSlide 6Phrase Based SystemsSlide 8Feature Based SystemsSlide 10CycSlide 12Knowledge BaseInference EngineNL Processing SubsystemSlide 16Slide 17Slide 18Research CycSlide 20The Problems with Meaningful GenerationSCIGenCommunications from ElsewhereOther ExamplesReferencesNatural Language GenerationEric ClarkCSC 9010-002Natural Language GenerationAn essay on Natural Language GenerationAre you sitting comfortably? I really, really like Natural Language Generation. Indispensable to homosapians today, it is yet to receive proper recognition for laying the foundations of democracy. It is estimated that that Natural Language Generation is thought about eight times every day by socialists, who form the last great hope for our civilzation. In the light of this I will break down the issues in order to give each of them the thought that they fully deserve.NLG – Application TypesCanned TextTemplate FillingPhrase Based SystemsFeature Based SystemsCanned TextDownload Complete“PC Load Letter”. What does that mean?Template FillingCongratulations Eric . You have been pre-approved for a low 6.9 % interest rate on your new Citibank Gold Plus Titanium Super Mastercard.Template FillingMail MergeSmartform.comAestivaPhrase Based SystemsGeneralized TemplatesPhrase Pattern matched at top level[SUBJECT VERB OBJECT][NP VP][Det Adj Noun]Etc.Phrase Based SystemsCreates sentences effectivelyLimited to one sentenceScalability limited by interrelationshipsFeature Based SystemsDictated by the characteristics of the sentence desiredPositive/NegativeInterrogative/DeclarativePast/PresentFeature Based SystemsSimplisticLanguage creation defined by the featuresNew distinctions are just added to the list of other featuresCycCreated by Cycorp, Inc.Austin, TX“The Cyc Knowledge Server is a very large, multi-contextual knowledge base and inference engine”CycKnowledge BaseInference EngineNL Processing SubsystemKnowledge BaseFactsRules of ThumbHeuristicsTermsAssertionsInference EngineLogical DeductionInference MechanismsBest-first searchNL Processing SubsystemLexiconSyntactic ParserSemantic InterpreterNL Processing SubsystemLexiconNL Processing SubsystemSyntactic Parser{:SENTENCE {:NP {:DETP {#$Determiner [the]}} {:N-BAR {#$SimpleNoun [man]}}} {:VP {#$Verb [saw]} {:NP {:DETP {#$Determiner [the]}} {:N-BAR {#$SimpleNoun [light]}} {:PP {#$Preposition [with]} {:NP {:DETP {#$Determiner [the]}} {:N-BAR {#$SimpleNoun [telescope]}}}}}}}}The man saw the light with the telescope.NL Processing SubsystemSemantic Interpreter “Mary believes that the blue hat is pretty“(#$believes :SUBJECT :CLAUSE)Research CycBased on OpenCycOpen SourceLinux BasedHas NLG capabilitiesHybrid Phrase/Feature BasedResearch Cyc“ResearchCyc is a version of the Cyc technology designed specifically for use by researchers”The Problems with Meaningful GenerationLimited to specific domainsNot ready for PrimetimeSome useful “dumb” applications are still availableSCIGenComputer Science Research Paper Generator“Our aim here is to maximize amusement, rather than coherence.”CSC 2180Communications from ElsewhereBad PoetryBand NamesPost ModernismOther ExamplesList of Work in NLGReferencesBateman, J. A. and Teich, E. (1995), `Selective information presentation in an integrated publication system: an application of genre-driven text generation', Information Processing and Management 31(5), 753-767.Bhattacharya, S, ‘Natural Language Generation’, accessed 2005 – available via Web: http://www.mla.iitkgp.ernet.in/~monojit/slides/NLG.pptUszkoreit, H ‘Language Generation’ accessed 2005 – available via Web: http://cslu.cse.ogi.edu/HLTsurvey/ch4node2.htmlNatural language generation : third international conference, INLG 2004, Brockenhurst, UK, July 14-16, 2004 : proceedings / Anja Belz, Roger Evans, Paul Piwek (eds.)ResearchCyc – Cycorp Inc., accessed 2005 – available via Web:
View Full Document