| Research >> Applications
Machine Translation
Machine translation of natural
languages (MT) is the oldest non-numerical application
of computing. In its 55-year history, MT has developed
a number of different paradigms and approaches. Since
about 1990, MT products have started to appear on the
consumer market. There are, however, still a number of
unsolved problems in MT that contribute to the still
rather unsatisfactory overall quality of MT. Large amounts
of literature on MT are available. Here is one of the
recent surveys. MT is one of ILIT’s major applications.
Machine-Aided Translation
Until fully automated MT starts
producing high-quality output, machine-aided translation
(MAT) will remain the option of choice for those who
want to raise the efficiency and lower the price of publication-quality
translation using some level of automation but not relying
exclusively on automatic systems. (When publication quality
is not required, even the less-than-perfect fully automatic
MT may suffice).
Information Extraction
Information extraction is the
task of analyzing texts and extracting from them information
for filling prefabricated forms, or templates. The fillers
can be snippets of actual text or – to allow the
output not only to be read by people but also to be used
by computer programs as a source of inference-making – snippets
of that text’s meaning representation.
Information Retrieval
Information retrieval is usually
understood as a task of retrieving documents from document
collections or WWW that match the key words and phrases
given in the query – or, better yet, match the
meanings conveyed by these key words and phrases.
Text Summarization
Text summarization takes a texts
(sometimes, more than one text) as input and generates
brief abstracts carrying the main points of the text.
Question Answering
Question answering is a major
application that requires the system to understand the
topic of conversation with the user, the user’s
goals and plans in asking the questions and the history
of a particular interchange as well as to gauge the appropriateness
of the various ways in which a question can be answered
(contentwise and stylistically).
|
