Scientists in Germany are trying to teach computers to read and understand texts written in natural language and interpret their meanings.
The research, conducted jointly by Saarbrucken University in Germany and the University of Amsterdam, is backed by Google through its $140,000 (£92,000) Google Focused Research Award.
‘The model that we have developed simulates how humans create texts,” said the research leader Ivan Titov. “In order to understand texts, we get our computers to work through this process but in the reverse direction: given the text the computer will uncover its meaning or even intent of the writer.”
Analysing millions of sentences taken from common sources such as the Wikipedia, the team aims at generating a computer model and a set of rules enabling computers to identify hidden, context-dependent relationships between words and clauses in texts.
The analysis of the data runs simultaneously on about one hundred computers, using specially developed algorithms. The final outcome of the project should be a software tool that could produce meaningful summaries of short texts and to answer questions about the text content. The researchers believe everyone who has to put up with reading and analysing large amounts of text would appreciate such tools – including students and scientists.
Together with Titov, another Saarland University Researcher has managed to capture the attention of Google – Hans Uszkoreit is trying to determine how computers could identify linguistic relationships in large text collections. The computational linguist has received $220,000 (£144,700) from the Internet giant, who considers the computational linguistics as a field of profound interest to the company. The Prize winners have free access to Google tools and technologies for the purposes of their research.