Progress in Language Translation via Computer?

This was a promising post that I made to ingehygd.blogspot.com in 2007:

News and Semantic Technologies

Unfortunately, this approach has to my knowledge not progressed. It was a good idea, however, accurate disambiguation remains an obstacle. The research was from student seminars held at The Data Center Program.

Recently, we (Schuster, Lee, Ehsani, Allen & Rogers) wrote the following in a manuscript:

"... for many years researchers have employed various computer-based techniques such as Artificial Intelligence (AI) to infer the contextual meaning of words associated with a physical or abstract object, thus enabling machine understandable semantics. These approaches have largely failed in practice because the meaning of a word, a phrase, or a sentence depends on the semantics of each word, the syntax or order of usage, and the context in which the word(s) appear. The unique properties of the human mind can determine contextual relevance, and then figure out how several relevant variables are associated (Deacon, 1997). However, to date, Natural Language Processing has not duplicated this mental property.

In addition, most AI techniques rely upon deduction where general concepts are applied to determine meaning. This becomes difficult to do in practice, because most word meanings in every-day situations are inductive with a strong dependence on specific context. Hence AI fails to live up to early promise (Chandler, 2009)."

Deacon, T.W., 1998. The Symbolic Species. 1st Ed. W.W. Norton & Company.

Chandler, D.T., 2009. Rethinking artificial intelligence. MIT News Office, Dec. 8.