Questions for the seminar Paper "Linear Spaces of Meanings: Compositional Structures in Vision-Language Models"
-----------------------------------------------------------------------------------------------
Please send your answers to: schrodi@cs.uni-freiburg.de

1) What do they mean with compositionality?
2) Briefly describe their approach to obtain ideal words.
3) What are their relaxed conditions for linear factorization and what do they imply for linearly factored embeddings?