NTR webinar: Structured Prediction in NLP

On November 17 Egor Gribkov, ML Engineer, TomskSoft, held a technical Zoom webinar on “Structured Prediction in NLP.”

Screenshot from webinar

Extracting structured information from texts is one of the most important topics in NLP. Often, specialized solutions are required to solve a certain class of problems, which are not easy to adapt to the specific task at hand.

The webinar examined a transition-based approach for machine learning model construction that allows many structured prediction problems to be expressed within a single framework.

We analyzed the general principles of building transition-based models, their pros and cons as compared to specialized solutions and examined cases of solving specific structured prediction problems with these models.Presentation in Russian is available here.

Leave a Reply

Your email address will not be published. Required fields are marked *