For the first time, a neural network model solved a crossword puzzle faster than a human

Written By O. Love

 It is 24.8% better than previous systems

The developers introduced the Berkeley Crossword Solver (BCS) model. It became the first computer program to beat all humans in the world’s premier crossword tournament.

BCS combines neural question answering and probabilistic inference to perform near perfect on most crossword puzzles like the one below:

 

For the first time, a neural network model solved a crossword puzzle faster than a human

BCS uses a two-step process to solve crossword puzzles. First, the model generates a probability distribution of possible answers, and second, it uses probabilistic inference in combination with local search and a generative language model.

The developers tested BCS on puzzles from five major crossword publishers, including The New York Times. The system solves 81.7% of puzzles without errors, which is 24.8% better than previous algorithms.

To Read Great Articles, Click Here

Leave a Comment