Grammar Rules Help Algorithms Grasp Language

June 20, 2019

Researchers at several universities have teamed up with IBM to teach algorithms some subtleties of language. VentureBeat reports, “IBM, MIT, and Harvard’s AI Uses Grammar Rules to Catch Linguistic Nuances of U.S. English.” Writer Kyle Wiggers links to the two resulting research papers, noting the research was to be presented at the recent North American Chapter of the Association for Computational Linguistics conference. We learn:

“The IBM team, along with scientists from MIT, Harvard, the University of California, Carnegie Mellon University, and Kyoto University, devised a tool set to suss out grammar-aware AI models’ linguistic prowess. As the coauthors explain, one model in question was trained on a sentence structure called recurrent neural network grammars, or RNNGs, that imbued it with basic grammar knowledge. The RNNG model and similar models with little-to-no grammar training were fed sentences with good, bad, or ambiguous syntax. The AI systems assigned probabilities to each word, such that in grammatically ‘off’ sentences, low-probability words appeared in the place of high-probability words. These were used to measure surprisal[sic]. The coauthors found that the RNNG system consistently performed better than systems trained on little-to-no grammar using a fraction of the data, and that it could comprehend ‘fairly sophisticated’ rules.”

See the write-up for a few details about those rules, or check out the research papers for more information (links above). This is but a start for their model, the team cautions, for the work must be validated on larger data sets. Still, they believe, this project represents a noteworthy milestone.

Cynthia Murrell, June 20, 2019

Comments

Got something to say?





  • Archives

  • Recent Posts

  • Meta