Information Retrieval Oriented Mongolian Homograph Pronunciation Recognition,International Journal of Asian Language Processing
Information Retrieval Oriented Mongolian Homograph Pronunciation Recognition,International Journal of Asian Language Processing
复制标题
DOI：
--
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
萨如拉
中科院分区：
文献类型：
--
作者：
斯劳格劳;萨如拉
Some Mongolian characters are spelt in the same way but pronounced differently. If only spelling was factored in word input, one word would have multiple input methods. One of the main challenges in Mongolian information retrieval is how to correct pronunciations and recognize the pronunciations of Homographs. According to statistics that in text whose words have been input using Mongolian International Standard Codes, mispronounced words account for an average of 40% of total words, up to 60% in some instances. Under such circumstances, we would be unable to get the information we need if mispronunciations were not corrected. The most difficult task in pronunciation correction is Homograph pronunciation recognition. This paper solved this problem with a training method based on conversion and collocation. The accuracy rate of pronunciation recognition has now hit more than 84%. More important, we have designed and implemented an error correction software with a human-computer interactive training mode which allows us to continuously improve the recognition skill of the software.
