. 4bstraet : Smartphone,σ1〃励 oo た o ηゴ 励 1θ'PC 尠 c1・se the distance between human and computer device . Bblt there are many users who can ' t use that Personal Digita114 ∬ ' ∫ご ant ∈ が σ ' ε η 珈 η -の 2 need to 吻 ワrove its 9ρ θ r α わ' 1 め ゐ 伽 θ〃igent agent system has been attrae 珈 9 attention to SO ' ve such a problem ・ In order to devel 〔 ) P intelligent agent system ,' t is necessar : y to impr( 冫V ε 伽 π ヨ CO9 η ' tion rate 〔 -プ a text conversation ノ " or these users ・ 跏 this paper, we propose

more »

... a mathematical 〃 10 ∂b' ノ br estimating the sentenee patterns , 加 澀 the } vord order qプ conversation text 伽 Japanese ) . Me use Hi ' attribortes a 刀 d basic eomponents a ∬ ubject , y8r わ, and adjective . ThUS } ve can infer stochast ' ca め ・ word order ( ゾ わα∫' Ccomponents ノ 沁 η3 pattern qプposipositional in the text .脆 are の ・ ' ' 0 ψ } 1' 0 ' e labeling Japanese conversation ' 餌 〃 ∫ ' H 膩 M 跏 ' 加 3 跏 φ田 V ・ modeis , the ・ model ・ cont 伽 ' わoth ・the ・sente 刀 0 θ 肋 肋 伽 3 η ・ 'C ・ 脚 in・a s 吻 e・t, ・and ・ the・sentence ・ C ・脚 ' 廨 9 ・ S 吻 ect, and the m ・del ・ f ・ η り ・ the sen ' ence ・ wh ' ch ・ d・es ・ n ・ ' contain a sublect . wereprepared , and comparat 加 θ 即 顔 〃 ients were conducted . Keyivords ' Hidden Markov ルlo ( lel . vaterb ' 490 厂舳 〃1, word order , Morphological Analysis , Twitter ハ伽耀