?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Learning+Analogies+and+Semantic+Relations&rft.creator=Turney%2C+Peter&rft.creator=Littman%2C+Michael&rft.subject=Language&rft.subject=Computational+Linguistics&rft.subject=Semantics&rft.subject=Machine+Learning&rft.description=We+present+an+algorithm+for+learning+from+unlabeled+text%2C+based+on+the+%0AVector+Space+Model+(VSM)+of+information+retrieval%2C+that+can+solve+verbal+%0Aanalogy+questions+of+the+kind+found+in+the+Scholastic+Aptitude+Test+(SAT).+%0AA+verbal+analogy+has+the+form+A%3AB%3A%3AC%3AD%2C+meaning+%22A+is+to+B+as+C+is+to+D%22%3B+%0Afor+example%2C+mason%3Astone%3A%3Acarpenter%3Awood.+SAT+analogy+questions+provide+%0Aa+word+pair%2C+A%3AB%2C+and+the+problem+is+to+select+the+most+analogous+word+%0Apair%2C+C%3AD%2C+from+a+set+of+five+choices.+The+VSM+algorithm+correctly%0Aanswers+47%25+of+a+collection+of+374+college-level+analogy+questions+%0A(random+guessing+would+yield+20%25+correct).+We+motivate+this+research+by+%0Arelating+it+to+work+in+cognitive+science+and+linguistics%2C+and+by+applying+%0Ait+to+a+difficult+problem+in+natural+language+processing%2C+determining+%0Asemantic+relations+in+noun-modifier+pairs.+The+problem+is+to+classify+a+%0Anoun-modifier+pair%2C+such+as+%22laser+printer%22%2C+according+to+the+semantic+%0Arelation+between+the+noun+(printer)+and+the+modifier+(laser).+We+use+a+%0Asupervised+nearest-neighbour+algorithm+that+assigns+a+class+to+a+given+%0Anoun-modifier+pair+by+finding+the+most+analogous+noun-modifier+pair+in+%0Athe+training+data.+With+30+classes+of+semantic+relations%2C+on+a+collection+%0Aof+600+labeled+noun-modifier+pairs%2C+the+learning+algorithm+attains+an+F+%0Avalue+of+26.5%25+(random+guessing%3A+3.3%25).+With+5+classes+of+semantic%0Arelations%2C+the+F+value+is+43.2%25+(random%3A+20%25).+The+performance+is+%0Astate-of-the-art+for+these+challenging+problems.&rft.date=2003&rft.type=Departmental+Technical+Report&rft.type=NonPeerReviewed&rft.format=application%2Fpdf&rft.identifier=http%3A%2F%2Fcogprints.org%2F3084%2F1%2FNRC-46488.pdf&rft.identifier=++Turney%2C+Peter+and+Littman%2C+Michael++(2003)+Learning+Analogies+and+Semantic+Relations.++%5BDepartmental+Technical+Report%5D+++++&rft.relation=http%3A%2F%2Fcogprints.org%2F3084%2F