中文摘要：

词义消歧问题可以形式化为典型的分类问题.通过学习少量带有词义标注的语料构造多个消歧分量分类器,并利用未标语料动态地对这些分类器进行更新,根据最终分量分类器分别对多义词义项的判定结果,组合决策多义词的义项.该方法无需手工构造大规模具有词义标注的语料库,并且具有较高的消歧准确率.

英文摘要：

The problem of word sense disambiguation can be formalized to be a typical classify problem. The committee classifiers are trained by learning a small set of labeled examples, and then these classifiers are updated dynamically by unlabeled examples. The senses of ambiguous words are determined by combining the decision of the final committee classifiers. This approach avoids constructing large-scale sense-tagged corpus, and has higher accurate rate.