tech,4-1-I05-5009,ak evaluation </term> . This paper presents an <term> evaluation method </term> employing a <term> latent variable
model,8-1-I05-5009,ak evaluation method </term> employing a <term> latent variable model </term> for <term> paraphrases </term> with their
other,12-1-I05-5009,ak <term> latent variable model </term> for <term> paraphrases </term> with their <term> contexts </term> .
other,15-1-I05-5009,ak <term> paraphrases </term> with their <term> contexts </term> . We assume that the <term> context
other,4-2-I05-5009,ak contexts </term> . We assume that the <term> context </term> of a <term> sentence </term> is indicated
other,7-2-I05-5009,ak assume that the <term> context </term> of a <term> sentence </term> is indicated by a <term> latent variable
other,12-2-I05-5009,ak <term> sentence </term> is indicated by a <term> latent variable </term> of the <term> model </term> as a <term>
model,16-2-I05-5009,ak <term> latent variable </term> of the <term> model </term> as a <term> topic </term> and that the
other,19-2-I05-5009,ak </term> of the <term> model </term> as a <term> topic </term> and that the <term> likelihood </term>
measure(ment),23-2-I05-5009,ak as a <term> topic </term> and that the <term> likelihood </term> of each <term> variable </term> can be
other,26-2-I05-5009,ak the <term> likelihood </term> of each <term> variable </term> can be inferred . A <term> paraphrase
other,1-3-I05-5009,ak variable </term> can be inferred . A <term> paraphrase </term> is evaluated for whether its <term>
other,7-3-I05-5009,ak </term> is evaluated for whether its <term> sentences </term> are used in the same <term> context
other,13-3-I05-5009,ak sentences </term> are used in the same <term> context </term> . Experimental results showed that
measure(ment),11-4-I05-5009,ak proposed method achieves almost 60 % <term> accuracy </term> and that there is not a large performance
model,24-4-I05-5009,ak performance difference between the two <term> models </term> . The results also revealed an <term>
measure(ment),5-5-I05-5009,ak </term> . The results also revealed an <term> upper bound </term> of <term> accuracy </term> of 77 % with
measure(ment),8-5-I05-5009,ak revealed an <term> upper bound </term> of <term> accuracy </term> of 77 % with the method when using
other,18-5-I05-5009,ak % with the method when using only <term> topic information </term> . This article is devoted to the
skrij podrobnosti