Problem Description

In order to automatically generate a dictionary from scratch, one sub-task is to generate explanatory sentences for each word, which are called glosses. For example a gloss for the word book could be: “a handwritten or printed work of fiction or nonfiction, usually on sheets of paper fastened or bound together within covers.” Recently there have been some approaches to solve this task (see, e.g.: McMillan-Majo, 2020; Bevilacqua, 2020; Vardhan, 2022). The goal of this thesis is to give an overview on existing methods and re-implement or even improve one of them.


  1. Give an overview on methods to automatically generate glosses for dictionaries.
  2. Re-implement a method or improve an existing one.
  3. Create automatically generated glosses.

Machine Learning, Interest in Natural Language Processing and Transformer Models.

Note that it is required to have passed the Machine Learning exam (i.e. choose Machine Learning as major subject) to choose this topic. We can also find different or similar topics related to Natural Language Processing.


Maike Behrendt
Raum 25.13.O2.36 ·