I've published Neural Machine Translation models for English to Persian and vice versa from OpenSubtitle2016 dataset at this repository. Over time, if any other free Parallel Corpus is found in English/Persian, I'll publish more models here as soon as possible.
For translate using these models, use the translate.lua from OpenNMT.
| Model Name | Source Vocab Size | Target Vocab Size | # of Epoch | Perplexity | Download Link |
|---|---|---|---|---|---|
| OpenSubtitles2016_en_fa | 100k | 100k | 20 | 12.62 | Download |
| OpenSubtitles2016_fa_en | 100k | 100k | 20 | 7.91 | Download |
These models are Trained using OpenNMT, See more details about the models here:
| Parameter | Value |
|---|---|
| -src_word_vec_size | 500 |
| -tgt_word_vec_size | 500 |
| -layers | 2 |
| -encoder_type | brnn |
These trained models are as free as the DataSet, for more information about the license, see OpenSubtitles2016 here : http://opus.lingfil.uu.se/OpenSubtitles2016.php