yumeng5 / topclus Goto Github PK
View Code? Open in Web Editor NEW[WWW 2022] Topic Discovery via Latent Space Clustering of Pretrained Language Model Representations
License: Apache License 2.0
[WWW 2022] Topic Discovery via Latent Space Clustering of Pretrained Language Model Representations
License: Apache License 2.0
Hello,
Could you please share the script used for evaluation? The obtained results may differ with different evaluation scripts.
Thanks a lot!
Best wishes,
Lei
Dear authors,
Thanks for your excellent work! I'm running the model on my own dataset and trying to understand the code. However, I always get cluster_loss = 0. I am using a Twitter dataset which is noisy and sparse, I have changed the BERT sequense length to 128 acoordingly in the pre-processing step. I have tried several different batch size and lr. (It seems to be working fine on the given yelp examples). Does anyone has the same problem or perhaps know why..? Thanks in advance!
Hi,
I've read your paper and I like this approach. Thank you for sharing the code. I've one question regarding the pretrained language models (PLMs) that you use for getting the contextualized word representations. I saw in the source code that the model you use is fixed, and it's the classical 'bert-base-uncased':
Line 22 in 01e22fb
Suppose I'm interested on using this method on a corpus of italian texts. In that case, would it be possible to change this model and use a bert-base-multilingual-uncased
instead?
If that's possible, can we make pretrained_lm
a parameter of the TopClusTrainer
?
Thank you.
Hi, thank you for sharing the code:)
Since I am a novice here, It may be a trivial question; but it seems that NMI is tested on the training dataset.
Is this a typical setting for measuring the NMI? (measuring on the training set)
Thank you
Line 79 in 01e22fb
Thank you for sharing the code. As far as I understand, this stage only trains autoencoder parameters, so It seems that
'''
Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=self.args.lr)
'''
should be converted to
'''
Adam(filter(lambda p: p.requires_grad, model.ae.parameters()), lr=self.args.lr).
'''
If it is not, please let me know.
I like your paper but I think it's confusing that how to tackle multiple embeddings of the same word/token. I wonder is there any chance that different embeddings of the same word are mapped to different clusters and all of them are quite close to the cluster center in the spherical space. How do you deal with that?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.