| 摘要 |
<P>PROBLEM TO BE SOLVED: To improve summarization accuracy. <P>SOLUTION: A summarization device for preparing a summary of an inputted document includes: a sentence importance estimation device 21 storing the weight of the feature amount of a sentence learned beforehand as a set of parameters; a sentence importance estimation part 22 for obtaining the importance weight(U<SB POS="POST">i</SB>) (provided that U<SB POS="POST">i</SB>indicates the i-th sentence of the document) of each sentence of the document using the sentence importance estimation device 21; and a summarization processing part 23 for, when a binary indicating whether or not the i-th sentence of the document includes a word j is defined as m<SB POS="POST">ij</SB>, the weight of the word j in the i-th sentence is defined as w<SB POS="POST">ij</SB>and a binary indicating whether or not the word j in the i-th sentence is included in a summary is defined as z<SB POS="POST">ij</SB>, obtaining the z<SB POS="POST">ij</SB>which maximizes a value for which the product of m<SB POS="POST">ij</SB>, w<SB POS="POST">ij</SB>and z<SB POS="POST">ij</SB>is added together for all i,j possible in the document to prepare a summary. The summarization processing part 23 obtains w<SB POS="POST">ij</SB>to be a larger value when the weight(U<SB POS="POST">i</SB>) is larger and to be a larger value when the importance weight(w<SB POS="POST">j</SB>) of the word j (provided that the w<SB POS="POST">j</SB>indicates a j-th word in a vocabulary constituting the document) is larger. <P>COPYRIGHT: (C)2012,JPO&INPIT |
