| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To improve processing speed of a morpheme analysis system using NFA. <P>SOLUTION: The morpheme analysis system has an analysis processing part 20, which retrieves words that compose natural language sentences from a word dictionary 30, retrieves concatenation rules to be used for concatenation of the words from a grammar dictionary 40, generates an automaton based on the words and the concatenation rules, and acquires solutions for morpheme analysis. The grammar dictionary 40 stores concatenation rules, which are unified with concatenation rules that express non-space transition and concatenation rules that a transition origin of the concatenation rules is set to be a transition target to express space transition. To generate an automaton, an optimal solution retrieval part 23 of the analysis processing part 20 generates only states necessary for transition according to concatenation rules corresponding to specified words and does not generate states, which can be transited with tracing space transition following to the generated states as processing when inputting the words into the automaton to generate states of a transition pass and a transition target. In addition, concatenation rules, in which a state that is generated immediately before is set to be a transition origin, are retrieved from concatenation rules related to the words. <P>COPYRIGHT: (C)2005,JPO&NCIPI</p> |
