| 摘要 |
A method for building a knowledge base containing entailment relations, therefore comprises the steps of:
a) providing at least one input pattern ( p ) with N pattern slots (N>1), said input pattern ( p ) expressing a specific semantic relation between N entities that fill the N pattern slots of the input pattern ( p ) as slot fillers,
b) providing at least one cluster ( c ) of articles, said articles of said cluster ( c ) relating to a common main topic;
c) processing said articles with respect to the input pattern ( p ) and identifying the identities which match the semantic type of the N pattern slots;
d) if said at least one input pattern matches a portion of an article ( a ) of said at least one cluster ( c ):
i) storing the N slot fillers (s 1 , s 2 , ... , s N ), which match the slots of the pattern (p), and a cluster identifier lc of the cluster ( c ) into a first table S , wherein the N-tuple (s 1 , s 2 , ... , s N ) and the cluster identifier l c of the associated cluster ( c ) form one element of said table S ;
ii) for each element of table S, identifying appearances of the slot fillers (s 1 , s 2 , ... , s N ) in a plurality of articles of cluster ( c ) and for each appearance so identified, storing the slot fillers (s 1 , s 2 , ... , s N ) together with the sentence in which they occur into a second table C 0 ;
iii) from the sentences stored in table C 0 , extracting patterns which span over the corresponding N slot fillers (s 1 , s 2 , ... , s N ), said extracted pattern expressing a semantic relation between said N slot fillers; and
iv) storing said extracted pattern together with said input pattern as entailment relation into said knowledge base. |
