Developed by and the University of Siena, WebCrow is the first AI-based software that tackles a language game using NLP technology and the richest self-updating repository of human knowledge available: the web. Designed to be multilingual, WebCrow is the world’s first Italian and English crossword puzzle solver.

The idea

Crossword puzzles are an “AI complete” problem by definition. Solving them requires not only human-level knowledge, but knowledge across a wide variety of subjects and domains. This level of complexity is considered extremely challenging for machines.

At the project’s inception in 2004, an initial nucleus of technology was available—search engines, techniques for information retrieval and machine learning—that enabled computers to assign meaning to text.

Since then, advances in AI technology have shown that machines are capable of challenging and even winning against humans in games like checkers, backgammon, chess, and lately, Go. While AI technology has also improved in the field of natural language understanding, clear limitations remain. Language games such as crossword puzzles highlight the strengths and weaknesses of NLP technology.

The goal

The University of Siena and designed WebCrow to address crossword puzzles using NLP approaches and different sources of knowledge used in NLP, such as ontologies and knowledge graphs, and the web, since it is an extremely rich and self-updating repository of human knowledge.

 An important aspect of solving a crossword puzzle is understanding the meaning of the clues given. It starts with understanding the clue, which leads to a sequence of logic and linguistic steps toward the correct answer.

What is WebCrow?

WebCrow is an AI-based software that contains a set of modules, including a rule-based module, a web search module and a crossword DB module. WebCrow can also invoke other expert modules, such as external NLP experts, using Redis messaging channels. WebCrow’s future challenge is to be competitive against human crossword masters.
WebCrow is mostly language-independent and it aspires to be able to solve crossword puzzles in different languages.