An English POS Tagging Approach Based on Maximum Entropy

This paper adopts the maximum model for English part of speech tagging. It makes pre-tagging for the word that has the only part of speech during the pretreatment of corpus, which adds many context features that can be utilized. We also improve the tagging algorithm, and take into account the whole optimization of POS series without extra computation, and the accuracy of tagging is also improved. The experimental results show the accuracy of open test has much room of improvement. The experimental results show that the combined algorithm achieves 94% accuracy and recall rate, and fully integrates the advantages of the maximum entropy method, which can be compared with the results of the same training and test corpus in ideal state.