Semantic Knowledge Acquisition based on Maximum Entropy

摘要

It's necessary to acquire semantic knowledge in Natural Language Processing research. In this paper, we present an approach for acquiring Chinese semantic knowledge based on maximum entropy model. Semantic knowledge units are composed of central word and a group of feature words. Because the maximum entropy to extract features, we utilize it to calculate the semantic distance between the central word and feature words in large-scale network corpus. In the experiment, tests on a number of manual defined data sets show that the Spearman correlation coefficient has been increased 6.2%-20.9%.