Luther12
Elder Lister
Matt Ginsberg is good at a lot of things - he is an AI scientist, author, playwright, magician and stunt plane pilot. But he isn't very good at crosswords.
In fact, despite writing them for the New York Times, he says that when they are published, he often cannot solve his own. So when he was sitting in a hotel ballroom losing yet again in a major US crossword competition, he decided to do something about it. "I was with 700 people who were really good at solving crossword puzzles and it annoyed me that I was so terrible, so I decided to write a computer program that would get even on my behalf," he told the BBC. And finally he did. After 10 failed attempts, Dr Fill - as the program is known - has just won its first competition. It came first in the American Crossword Puzzle Tournament, the leading crossword competition in the US.
Dr Fill was trained on a mass of data, including a giant database of crossword clues and answers scraped from the web.
Matt Ginsberg was so bad at crosswords he had to build a machine that could make him look better
It was taught to search at speed through possible placements of words in a crossword grid. It was, admits Dr Ginsberg, quite a "primitive" system.
This year, he had some help. "A few weeks before the event, I was contacted by people working at Berkeley who had built a crossword clue answering system. We realised pretty quickly that we could combine the two."
Prof Dan Klein, who heads the Natural Language Processing Group at Berkeley College, University of California, explained to the BBC that he was looking for something to bring the team together during lockdown - and they came up with the idea of building a crossword solver. When he heard about Dr Fill, he thought the two systems would make a good partnership.
Crossword puzzles can fool the smartest of humans, so how would a machine fare?
"Our system brought a broader understanding of language, and Dr Fill was good at how the answers combine with other clues. They are very different techniques but they spoke a common language of probabilities."
www.bbc.com
In fact, despite writing them for the New York Times, he says that when they are published, he often cannot solve his own. So when he was sitting in a hotel ballroom losing yet again in a major US crossword competition, he decided to do something about it. "I was with 700 people who were really good at solving crossword puzzles and it annoyed me that I was so terrible, so I decided to write a computer program that would get even on my behalf," he told the BBC. And finally he did. After 10 failed attempts, Dr Fill - as the program is known - has just won its first competition. It came first in the American Crossword Puzzle Tournament, the leading crossword competition in the US.
Dr Fill was trained on a mass of data, including a giant database of crossword clues and answers scraped from the web.
Matt Ginsberg was so bad at crosswords he had to build a machine that could make him look better
It was taught to search at speed through possible placements of words in a crossword grid. It was, admits Dr Ginsberg, quite a "primitive" system.
This year, he had some help. "A few weeks before the event, I was contacted by people working at Berkeley who had built a crossword clue answering system. We realised pretty quickly that we could combine the two."
Prof Dan Klein, who heads the Natural Language Processing Group at Berkeley College, University of California, explained to the BBC that he was looking for something to bring the team together during lockdown - and they came up with the idea of building a crossword solver. When he heard about Dr Fill, he thought the two systems would make a good partnership.
Crossword puzzles can fool the smartest of humans, so how would a machine fare?
"Our system brought a broader understanding of language, and Dr Fill was good at how the answers combine with other clues. They are very different techniques but they spoke a common language of probabilities."
'I was terrible at crosswords so I built an AI to do them'
Dr Fill , a computer program, has just won one of the most prestigious crossword competitions in the US.
www.bbc.com