Published April 25, 2018
| Version v1
Dataset
Open
An Empirical Study of Word Embedding Dimensionality Reduction
Description
In order to analyze the impact on model quality while reducing the number of dimensions, strictly controlled trainings of word embedding are performed on Wikipedia corpora of 170 languages. The specially designed word embedding training tool makes use of processed corpus and intermediate results to accelerate the training, while keeping the consistency of negative sampling.
Tests of semantic relatedness show that, except for some corpora of poor scale, the margin gain from extra dimensions significantly decreases above 200.
Files
models-100.zip
Files
(12.1 GB)
Name | Size | Download all |
---|---|---|
md5:41939b1442bf98fca7061e5f453aaf1e
|
1.2 GB | Preview Download |
md5:4231678bb87430787b5e03c9cec34185
|
2.3 GB | Preview Download |
md5:499faeb960521b19661e39bac1e09f47
|
3.4 GB | Preview Download |
md5:56e2838f23536bc1e8b72c672668d3cc
|
4.6 GB | Preview Download |
md5:4c75f9c9b07a189132bd0402cf0ddafd
|
584.7 MB | Download |
md5:2e4f0abf594112f9fb35d86457eb74e9
|
19.3 MB | Preview Download |