| The application of traffic term is of great significance for improving the normalization of information system and its data content. However, due to the characters of Chinese word, different combinations of words result in different representations of the same term. It is not only hard to check them one by one manually but also requires the workers to be highly familiar with the standards of term. For above reasons, this paper presents the compliance detection technology of traffic term and its normalization, which not only facilitates matching traffic terms, but also supports automatically detecting similar expressions based on text semantic, evaluating the non-standard expressions and then giving some modified advice accordingly.The main work and innovation of this paper are following:1 Establish traffic term corpus based on the data elements of traffic information and extend traditional corpus, which provide not only data support for Chinese word segmentation of traffic information, but the basis for the establishment of the detection system of traffic standards.2 Establish traffic term structure library based on the analysis of structure and sememes description of the HowNet and extend the knowledge base of the HowNet, which provide the basis for getting correct results in similarity measurement when using traffic term.3 Optimize current methods of Chinese word segmentation using Cascaded Hidden Markov Model, which shortens the overhead of system. Improve the performance of segmentation by the introduction of traffic term corpus, which leads to get accurate and complete traffic term segmentation of traffic text.4 A similarity measurement based on formal concept analysis is presented, which optimizes current architecture of HowNet and minimizes error due to sememes depth. Compared to the HowNet one, the presented similarity measurement has a better performance Computer simulations have shown that due to the introduction of traffic term corpus and the use of the proposed word segmentation method based on Cascaded Hidden Markov Model, the correct rate of word segmentation has been improved, which is increased from 96.55% to 97.96%, and the accurate rate of the compliance detection system of traffic term standards has been up to about 70%. |
