Automatic Patterns Acquisition and Evaluation for Web-Based Terminology Translation

To find the translation of a given terminology from web without dictionary is an interesting and challengeable work. The existing methods based on pattern mainly consist of two steps: (1) learn patterns from a training set, and (2) find the candidate terms and score them. However, there are two main deficiencies in the existing works: (1) the amount and reliability of patterns are restricted by the training set, and (2) the methods for scoring patterns and candidate terminologies are too simplified. We present a new method that needs only a pair of terminologies for training to acquire the initial patterns. Patterns acquisition and patterns evaluation are performed automatically when an appropriate candidate terminology is selected for a user's query. We also improve the method of scoring the candidate terminologies by applying heuristic rules. The experiment results show that our method is better than the existing technologies.