Translating-transliterating named entities for multilingual information access

monolingual named entities. Extending them to multilingual entities is becoming important because a large amount of multilingual materials are generated and disseminated over the Web. The fundamental issues in processing multilingual named entities are recognizing them and finding their correspondence. Embedded technologies include learning formulation and transformation rules for multilingual named entities, and translating–transliterating named entities and content words. The common named entities mentioned at the 7th Message Understanding Conference (MUC, 1998), including date/time expressions and monetary and percentage expressions, have fixed patterns, so their mapping among different languages is easy. This work focuses on more flexible patterns such as person, location, and organization names. Transformation between named entities in different languages is not only translation or transliteration. The mapping may be a combination of meaning translation and/or phoneme transliteration (Chen, Yang, & Lin, 2003). The following five English–Chinese examples clarify this issue, where the symbol A ⇔ B denotes that foreign name A is translated and/or transliterated into a Chinese name B.