Closeness Based New Word Detection Method for Mechanical Design and Manufacturing Area

10.3966/199115992017102805019

Qiuyuan Chen；Guang Cheng；Di Li；Jian Zhang

information entropy ； left(right) entropy ； logistic regression ； mechanical design ； named entity recognition ； new word detection

電腦學刊

28卷5期（2017 / 10 / 01）

210 - 219

英文

Named entity recognition has been widely used in the area of information retrieval, but the common methods cannot accurately identify the proper nouns from a particular domain. In order to solve the named entity recognition for mechanical design and manufacturing area, this paper proposes a closeness based method, in order to identify the proper nouns. First, we calculate the entropy about each character, and then define the closeness between two adjacent words based on the lexical features and statistic features. Finally we use the logistic regression algorithm to determine the weights in the closeness definition. The proposed method can recognize proper nouns more accurately and efficiently for mechanical design and manufacturing area.