The humbling power of computers - when correctly programmed - to surpass, in speed at least, human skills, has been demonstrated by a computer program that can learn language rules and compose sentences, without outside help.
Researchers from Cornell University, in the US, and Tel Aviv University have developed a computer program that can scan text in any of a number of languages, including English and Chinese, and without any previous knowledge infer the underlying rules of grammar. The rules can be used to generate new and meaningful sentences.
According to the researchers, the method also works for such data as sheet music or protein sequences.
The development has implication for speech recognition and for other applications in natural language engineering, as well as for genomics. It also offers insights into language acquisition and psycho-linguistics.
No comments:
Post a Comment