paragraph segmentation of judicial decisions
Are you recently still working on law related NLP research?
I have noticed that you are also a researcher with ACT and CyberJustice Lab, but it seems that this project is coming to an end this year? Will there be any follow up projects?
https://www.ajcact.org/en/repertoire/kosseim-leila/
What the real world application? (i.e. Has any Canadian court implemented this fact extraction method )
My comment:
Representations
Different approaches to address semantic similarity matching generally fall into one of the two categories of interaction-based and representation-based models. While each approach offers its own benefits and can be used in certain scenarios, using a transformer- based model with a completely interaction-based approach may not be practical in many real-life use cases. In this work, we compare the performance and inference time of interaction-based and representation-based models using contextualized representations. We also propose a novel approach which is based on the late interaction of textual representations, thus benefiting from the advantages of both model type
In all of the 3 models, a single feed-forward layer with Sigmoid activation functions is used as the similarity module. However, in a ranking task, one can simply replace the classification layer with cosine similarity or other ranking schemes and use the same strategies for the task.
isn’t Semantic Similarity Matching (SSM) normally a ranking problem not a classification problem?
What’s the purpose of concat? I thought we want to compare the embeddings of Text1 and Text2 instead of combining them???
What kind of methods do FAISS or ANNOY use for Semantic Similarity Matching (SSM), would it be something similar to (b) or (c) ?
?? No matter which a b c you use, you are always comparing two documents
log n * (n-1) / 2 = n^2 —> this is why we need GPU