Auto-recognizing Letter-word Phrases in Chinese Texts

As a group of unknown words of Chinese information processing, the letter-word phrases used in Chinese texts can´t be identified correctly by the existed segmentation software. Here, an auto-tagging system of letter-word phrases based on rules and statistical data is presented. At first, the system scans the sentences to get letter-strings, and then takes every letter string as an anchor and scans its two sides, calls boundary words rules, LWP-component rules, exceptional-correct rules and collocation coefficient matrix in turn, to judge whether the letter string should be bound with a LWP-component, finally tags the Letter-word phrases in the sentence. With the close test and open test, the experiment shows that the tagging system precision reaches above 90%.