https://sites.google.com/view/frame-blends-entry/home?authuser=1

Thank you Suzie!

sxi@smith.edu

Thank you, Suzie! A few questions: (1) What sort of dataset did you use to test this? (2) Since word embeddings often average different uses of any given word into a single embedding, do you think including other sorts of data in the nominations models would help (e.g., part of speech)? (3) How is the frame embedding different from the word embedding? Are they the same embeddings (using word2vec or GloVE)? Or are they created differently?

Sorry, ignore any of these. A fourth question, do you take word order into account when you sum the embeddings?

Really exciting work! Have you used some of the information in the FrameNet annotations in this process? I'm thinking of some of the syntactic patterns in the annotated sentences.

Thank you, Suzie!

I have one other question.

I wonder if these great questions might also apply to thinking about the possibility of metaphor embeddings …

Great question!

Once you have found a data set of frame blends, could you then cross check that with gestures to see if there are correlations there?

Thank you!

Here’s an interesting paper on Automatic metaphor detection: https://doi.org/10.1075/cf.8.2.06hon

Thanks Tiago!

Thank you!

Thank you!

Thank you!

Thank you!

Thank you!