Watch Language Models _____ Low-context Word Prediction Tests - Einzelansicht

Correctly answering word prediction tests is a ____ task for many language models. It is used in various applications, like our own netspeak.org. This task gets particularly difficult if the context is limited: "is a _____ task". In this project, we want to investigate how several different language models answer word prediction tests under varying given contexts sizes, given genres of text, and phrase structure of the test. We target three milestones for this project: (1) Collect existing word prediction test datasets, (2) construct a new dataset of word prediction tests that covers all the parameters we want to vary, and (3) run different language models on the datasets and evaluate the results.

time and place: t.b.a.

Abschlusspräsentation und Ausarbeitung

B.Sc. Medieninformatik

M.Sc. Computer Science for Digital Media /Computer Science and Media

M.Sc. Human-Computer Interaction