I am working on a computer program that understands the Turkish language. The program first reads a written Turkish text and then answers questions about its content. It can also outline a given text paragraph by paragraph.
It may be used to do tasks related to large amounts of text or spoken English provided that they are converted to text.
Evaluation of resumes, production of meeting notes, internet tracking for companies or individuals, contentwise comparison of large texts, translation, tracing of legal recordings of aviation, security etc.
tracing of newspapers and broadcasting on internet.
The architecture of the program in a few words: It is based on JAVACC and its multiple use on byproducts. It imitates exactly the Turkish grammar and produces intermediate level outputs for words, sentences and higher levels.
Currently, a given sentence is analyzed and each word and its extensions are classified as noun, verb, pronoun, adjective usw. The root words are verified with a Turkish dictionary and then extensions are specified.
Consequently, the functions of words in sentences will be specified as subject, object, verb etc.
At this stage, the program will be able to answer simple questions.
If you are interested in this project please contact me.
arsaral ( at ) yahoo.com
More to come.