| CPC G06N 5/04 (2013.01) [G06F 16/2365 (2019.01); G06F 16/3329 (2019.01); G06F 16/3344 (2019.01); G06F 16/355 (2019.01); G06N 5/02 (2013.01); G06N 5/025 (2013.01); G06N 20/00 (2019.01)] | 17 Claims |
|
1. A question-answering learning method, comprising:
creating a classifier module by a classifier generation module according to N1 labeled sentences among N sentences, wherein the N sentences comprises the N1 labeled sentences and N2 unlabeled sentences, and a classifier module comprises a plurality of classifiers each representing different question-answering models, wherein both of the N and the N1 are positive integers;
determining at least one corresponding sentence type of each of the N2 unlabeled sentences among the N sentences by each of the classifiers, wherein the N2 is a positive integer;
selecting, in a consistency evaluation process, N3 sentences from the N2 unlabeled sentences by a consistency evaluation module according to a degree of consistency of determined results of the classifiers, wherein the determined results of the N3 sentences are determined by the classifiers and are inconsistent, and the N3 is a positive integer;
selecting, in a complementarity evaluation process, N4 mutually complementary sentences as to-be-labeled sentences from the N3 sentences by a complementarity evaluation module, wherein the N4 is a positive integer;
re-creating, after the selected sentences are labeled, the classifiers of the classifier module by the classifier generation module according to the N1 labeled sentences and the selected sentences; and
adding at least one of classifiers created prior to the recreation to the classifier module to be members of the classifier module by a classifier evaluation module;
wherein the N4 mutually complementary sentences are unlabeled sentences, and N4 is less than N.
|