Design and Analysis of a POS Tag Multilingual Dictionary for Mongolian

In this paper, we report a multilingual parallel electronic dictionary, named MPEDM. The MPEDM dictionary covers the languages in Chinese and Mongolian, including TM (traditional Mongolian), TODO (Mongolian TODO) used in China, and NM (New Mongolian Cyrillic character) used in Mongolia. The average coverage evaluation of MPEDM among different systems of Mongolian was conducted by applying the Cosine similarity measure.To build MPEDM, 80 thousands of NM words were selected by linguists. Then, a part-of-speech (POS) and Chinese word pairs were added to MPEDM by manually referring to a standard version from Inner Mongolian University and supported by NSFC of China. After that, the traditional and TODO Mongolian words were paralleled automatically based on parsing process of the Mongolian multilingual parallel corpus. It is believed that the MPEDM dictionary can be applied in many tasks, such as multilingual word searching and interpreting a word in its reading and grammatical form on line.The experimental results demonstrate that the average coverage of different texts achieves 85.2%. In addition, the relativity evaluations of MPEDM between different systems show that the similarity of TM and TODO and TODO and NM pairs are 0.72 and 0.86, respectively. Finally the best F-score of 0.67 is attained when converting Mongolian NM texts to TODO ones.