How did you determine the cosine similarity threshold between sentence embeddings? #4
Description
Hi, thank you for the work you have done. I have some questions for you.
In the paper, you mentioned that the dynamic citation window involves cosine similarity between sentence embeddings and that only sentences above a certain threshold are selected. So, my question is: what is this threshold? Is it a deterministic or a dynamic value? And how is it determined? Whether the difference in threshold values affects the experimental results?
The sentences in the paper are: The cosine similarity between the two document embeddings determines the threshold for adding other neighbouring sentences. in pp.543