@inproceedings{brandl-lassner-2019-times,
title = "Times Are Changing: Investigating the Pace of Language Change in Diachronic Word Embeddings",
author = "Brandl, Stephanie and
Lassner, David",
booktitle = "Proceedings of the 1st International Workshop on Computational Approaches to Historical Language Change",
month = aug,
year = "2019",
address = "Florence, Italy",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/W19-4718",
doi = "10.18653/v1/W19-4718",
pages = "146--150",
}
1. What is it?
The authors proposed Word Embedding Networks (WEN) to chase the semantic change.
2. What is amazing compared to previous studies?
They proposed the new method, WEN.
3. Where is the key to technologies and techniques?
WEN starts with assuming an equal distance between all embeddings and then, over time, shapes the relations by moving certain embeddings closer and others farther apart.
As above, this method is an evolution of Dynamic word embeddings.
The objective function is shown below:
They update the third term (considering similarity over time).
To update the weight w_{t, t'}, they use normalized similarity:
4. How did validate it?
They draw the 2-dimensional map by using Laplacian eigenmaps to visualize.
This graph shows that the time when the semantic changes faster, and the time when semantic changes slower.
5. Is there a discussion?
distributional changes within the data set can have a huge influence on the perceived pace of semantic change.
0. Paper
@inproceedings{brandl-lassner-2019-times, title = "Times Are Changing: Investigating the Pace of Language Change in Diachronic Word Embeddings", author = "Brandl, Stephanie and Lassner, David", booktitle = "Proceedings of the 1st International Workshop on Computational Approaches to Historical Language Change", month = aug, year = "2019", address = "Florence, Italy", publisher = "Association for Computational Linguistics", url = "https://www.aclweb.org/anthology/W19-4718", doi = "10.18653/v1/W19-4718", pages = "146--150", }
1. What is it?
The authors proposed Word Embedding Networks (WEN) to chase the semantic change.
2. What is amazing compared to previous studies?
They proposed the new method, WEN.
3. Where is the key to technologies and techniques?
As above, this method is an evolution of Dynamic word embeddings. The objective function is shown below:
They update the third term (considering similarity over time). To update the weight w_{t, t'}, they use normalized similarity:
4. How did validate it?
They draw the 2-dimensional map by using Laplacian eigenmaps to visualize.
This graph shows that the time when the semantic changes faster, and the time when semantic changes slower.
5. Is there a discussion?