| 摘要 |
The method of indexing multimedia documents comprises the following steps: a) for each document identifying and extracting terms t<SUB>i </SUB>constituted by vectors characterizing properties of the; b) storing terms t<SUB>i </SUB>in a term base comprising P terms; c) determining a maximum number N of desired concepts that group together the most pertinent terms t<SUB>i</SUB>; d) calculating the matrix T of distances between the terms t<SUB>i </SUB>of the term base; e) decomposing the set P of terms t<SUB>i </SUB>of the term base into N portions P<SUB>j </SUB>(1<=j<=N) SUCH THAT P=P<SUB>1</SUB>∪ P<SUB>2 </SUB>. . . ∪ P<SUB>j </SUB>. . . ∪ P<SUB>N</SUB>, each portion P<SUB>j </SUB>comprising a set of terms t<SUB>ij </SUB>and being represented by a concept c<SUB>j</SUB>, the terms t<SUB>i </SUB>being distributed in such a manner that the terms that are farther apart are to be found in distinct portions P<SUB>l</SUB>, P<SUB>m</SUB>, and the terms that are closer together are to be found in the same portion P<SUB>l</SUB>; f) structuring the concept dictionary; and g) constructing a fingerprint base made up of the set of concepts c<SUB>i </SUB>representing the terms t<SUB>i </SUB>of the documents, each document being associated with a fingerprint that is specific thereto.
|
