 |
|
|
 |
| |
Last Issue:
Tuesday,
December
18
2007
|
|
|
|
|
Teaching Computers to Translate
By Desair Brown
Published on 27-Sep-04
Computers can now provide basic translations or "gists," of different documents for you, according to Bonnie Dorr, associate professor of computer science and co-director of Computational Linguistics and Information Processing (CLIP).
Dorr, along with 20 other professors, post-doctoral fellows and students are working on creating software that can summarize, sort and translate documents quickly and cross-lingually.
"We are looking for ways to provide information to users who may speak more than one language or need documents written in languages they don't speak."
Dorr, who has 20 years of experience in computational linguistics, says multilingual processing comes about once a document is summarized and scanned. She is currently working on retrieval and summarization engines, and targeting the differences in dialects.
So far, Dorr has looked at Spanish (in which she has a degree) and Chinese. She is also looking into Arabic, Hindi and Korean. She says each language presents its own set of challenges.
Chinese, for example, does not use white space between words, so text must be segmented into individual words. And Arabic, which reads right to left, is highly ambiguous, using multiple meanings and contexts for one word.
"I am fluent in Spanish and English and have a working knowledge of German. However, computational linguists frequently know the basic structures of many different languages and threads of commonality across seemingly diverse language pairs.
The same concept applies to machine translation software, which requires vocabulary training and grammatical fine-tuning.
Dorr says CLIP provides ideas for development so that government and military sponsors can optimize the prototypes for their use. The research is continuously paralleling evolving technology, hardware and natural language-learning techniques.
CLIP is co-directed by Amy Weinberg of the linguistics department, and
Louiqa Raschid of the information systems decision and information technologies department. For more information, visit the CLIP Web site at www.umiacs.umd.edu/research/CLIP.
|
|
|
|
|
|
|
