Crosswords are helping artificial intelligence systems better understand language.
Researchers from the University of Cambridge hope that their system, which uses an artificial neural network to answer crossword questions, will help machines understand language more rapidly and in a more complex way.
The system was trained on six dictionaries' worth of definitions to give it a large number of examples. As a result of the experiment the researchers found that definitions contain a "valuable signal" for helping to interpret and represent the meaning of phrases and sentences.
The results, published in Transactions for the Association for Computational Linguistics, found that the system was able to work out short phrases, metaphors and longer, linked phrases. "Despite recent progress in AI, problems involving language understanding are particularly difficult, and our work suggests many possible applications of deep neural networks to language technology," said Felix Hill, who helped develop the system. "One of the biggest challenges in training computers to understand language is recreating the many rich and diverse information sources available to humans when they learn to speak and read."
The team now hope to train the system to understand information contextually. "Our system can’t go too far beyond the dictionary data on which it was trained, but the ways in which it can are interesting, and make it a surprisingly robust question and answer system – and quite good at solving crossword puzzles," said Hill.
The code used to develop the system is available online.
This article was originally published by WIRED UK
