@inproceedings{emms-jayapal-2015-unsupervised,
title = "An unsupervised {EM} method to infer time variation in sense probabilities",
author = "Emms, Martin and
Jayapal, Arun",
booktitle = "Proceedings of the 12th International Conference on Natural Language Processing",
month = dec,
year = "2015",
address = "Trivandrum, India",
publisher = "NLP Association of India",
url = "https://www.aclweb.org/anthology/W15-5913",
pages = "89--94",
}
1. What is it?
They create the generative model:
context words from a sense of target word
sense from the time
They adopted EM algorithm to create the model without annotating sense labels.
2. What is amazing compared to previous works?
They use a simple generative model for semantic change detection.
3. Where is the key to technologies and techniques?
They create the bayesian model, time-stamp Y and sentence words W as below:
Using EM algorithm
In E step, estimate the probability P(s|Y, W) to determine a parameter γ(s).
In M step, update the formulate under the parameter γ(s) as below:
a1da4
commented
4 years ago
0. Paper
@inproceedings{emms-jayapal-2015-unsupervised, title = "An unsupervised {EM} method to infer time variation in sense probabilities", author = "Emms, Martin and Jayapal, Arun", booktitle = "Proceedings of the 12th International Conference on Natural Language Processing", month = dec, year = "2015", address = "Trivandrum, India", publisher = "NLP Association of India", url = "https://www.aclweb.org/anthology/W15-5913", pages = "89--94", }
1. What is it?
They create the generative model:
They adopted EM algorithm to create the model without annotating sense labels.
2. What is amazing compared to previous works?
They use a simple generative model for semantic change detection.
3. Where is the key to technologies and techniques?
They create the bayesian model, time-stamp Y and sentence words W as below:
Using EM algorithm
4. How did evaluate it?
5. Is there a discussion?
6. Which paper should read next?