Expert.ai to Livestream “WebCrow” on June 16th.
Stage Set for Multilingual Showdown Against Human Experts
Starting today, even machines can solve crossword puzzles thanks to WebCrow 2.0, a software developed by the University of Siena in collaboration with expert.ai (EXAI:IM), a leading company in artificial intelligence (AI) for natural language processing (NLP) and understanding (NLU).
For over a century, crossword puzzles have been an intriguing challenge for humans because of the complexity and nuance of the human language. This also happens to be one of the most complex and challenging areas for AI. In fact, the most advanced linguistic technologies must possess a significant breadth and depth of knowledge to identify the correct meaning of words based on context (e.g., trim a tree vs. trim on a house). They must also be able to interpret slang, catch phrases, wordplay and other forms of ambiguity (e.g., a crossword clue: liquid that does not stick, answer: scotch). WebCrow 2.0 does this and more.
“We’re excited to introduce our intelligent machine, WebCrow, and discuss its evolution and ability to create and solve a daily standard of life, the crossword puzzle,” said Marco Gori, Professor, Department of Information Engineering and Mathematical Sciences, University of Siena. “Can machines solve these as well as humans? How do they compare definitions and answer clues with niche or abstract references? Can they pick up on plays on words, linguistic nuances and even humor? We’re ready to demonstrate how leveraging context can enable humans and software to work together and take AI-based cognitive abilities to new levels.”
Understanding, Knowledge graph, Reasoning
WebCrow 2.0 has been empowered with typical human skills to simulate human-like processes for reading, understanding, and reasoning. This allows the software to identify the meaning of words based on definitions and other clues in crossword puzzles. It accomplishes this by:
• Understanding: Utilizing several logical and language analysis steps focused on understanding clues.
• Knowledge: Finding the correct meaning requires inferences and reasoning that makes use of knowledge in various sources. Expert.ai’s knowledge graph enables this with the market’s richest conceptual database for understanding language.
• Reasoning: A combination of advanced AI, NLU with NLP, machine learning and constraint satisfaction technologies identify possible words, evaluate them and enter the answers in the puzzle grid.
“It’s our business to help organizations improve any activity or process based on understanding and managing the immense wealth of information at their disposal,” said Marco Varone, CTO of expert.ai. “It was very gratifying to work with researchers from the University of Siena and support their efforts with our tools for disambiguation, knowledge graph and expertise in applying AI to language. Anyone who has been challenged by a crossword is familiar with nuanced clues, so automated puzzle solving is a great way to illustrate just how far we’ve come in advancing natural language technologies.”
Livestream: Solving Crosswords with WebCrow AI
A movie about WebCrow and its crossword-solving abilities can be viewed on the website. A special LinkedIn NLP stream session, “Solving Crossword Puzzles with WebCrow AI,” is scheduled for June 16 at 11:00 am EDT. Those interested in attending can register here.
The Next Challenge for WebCrow
Next up for WebCrow is to compete against human experts in a multilingual competition. The “WebCrow 2.0 – Human vs. Machine” challenge is organized by expert.ai and the University of Siena, in collaboration with Sudoku Editori (unpublished crosswords for the Italian language) and AVCX “Crosswords for the (not) faint of heart” (unpublished crosswords in English).