Paragram SL999 (300 dimensions)
300-dimensional Paragram word embeddings for 1,703,756 English words, tuned on the SimLex-999 dataset.
This is a gensim model, packaged for fse (Fast Sentence Embeddings), which turns word vectors into sentence vectors with CBOW averaging, SIF and uSIF.
Usage
This repository uses the fse 1.x API and requires fse<2.
For fse 2.0, use fse/paragram-300-sl999-raw for a finished mean encoder.
See the migration guide before upgrading.
pip install "fse<2"
from fse import SIF, IndexedList, Vectors
vectors = Vectors.from_pretrained("paragram-300-sl999")
model = SIF(vectors)
sentences = IndexedList([["hello", "world"], ["fast", "sentence", "embeddings"]])
model.train(sentences)
model.sv[0] # the vector of the first sentence
Vectors.from_pretrained("paragram-300-sl999", mmap="r") memory-maps the vectors instead of loading them into RAM.
Files
| File | Content |
|---|---|
paragram-300-sl999.model |
gensim KeyedVectors (vocabulary and metadata) |
paragram-300-sl999.model.vectors.npy |
1,703,756 × 300 float32 vectors (2.0 GB) |
Source and license
The vectors were published by John Wieting on his page at Carnegie Mellon University. The page states no license, so none is given here. This repository redistributes them in the gensim format.
- Wieting, Bansal, Gimpel, Livescu and Roth (2015), From Paraphrase Database to Compositional Paraphrase Model and Back
About fse
fse is maintained by Dr. Oliver Borchers. Source code, documentation and the STS benchmark of all vector sets in this organization: github.com/oborchers/Fast_Sentence_Embeddings.
- Downloads last month
- 57