feat: add readme.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,87 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
license: cc-by-sa-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
language:
|
| 3 |
+
- cs
|
| 4 |
+
- en
|
| 5 |
+
- de
|
| 6 |
+
- fr
|
| 7 |
+
- tu
|
| 8 |
+
- zh
|
| 9 |
+
- es
|
| 10 |
+
- ru
|
| 11 |
+
tags:
|
| 12 |
+
- Summarization
|
| 13 |
+
- abstractive summarization
|
| 14 |
+
- multilingual summarization
|
| 15 |
+
- m2m100_418M
|
| 16 |
+
- Czech
|
| 17 |
+
- text2text generation
|
| 18 |
+
- text generation
|
| 19 |
license: cc-by-sa-4.0
|
| 20 |
+
datasets:
|
| 21 |
+
- Multilingual_large_dataset_(multilarge)
|
| 22 |
+
- cnc/dm
|
| 23 |
+
- xsum
|
| 24 |
+
- mlsum
|
| 25 |
+
- cnewsum
|
| 26 |
+
- cnc
|
| 27 |
+
- sumeczech
|
| 28 |
+
metrics:
|
| 29 |
+
- rouge
|
| 30 |
+
- rougeraw
|
| 31 |
+
- MemesCS
|
| 32 |
---
|
| 33 |
+
# mbart25-multilingual-summarization-multilarge-cs
|
| 34 |
+
This model is a fine-tuned checkpoint of [facebook/m2m100_418M](https://huggingface.co/facebook/m2m100_418M) on the Multilingual large summarization dataset focused on Czech texts to produce multilingual summaries.
|
| 35 |
+
## Task
|
| 36 |
+
The model deals with a multi-sentence summary in eight different languages. With the idea of adding other foreign language documents, and by having a considerable amount of Czech documents, we aimed to improve model summarization in the Czech language. Supported languages: ''cs', 'en', 'de', 'es', 'fr', 'ru', 'tu', 'zh'
|
| 37 |
+
## Dataset
|
| 38 |
+
Multilingual large summarization dataset consists of 10 sub-datasets mainly based on news and daily mails. For the training, it was used the entire training set and 72% of the validation set.
|
| 39 |
+
```
|
| 40 |
+
Train set: 3 464 563 docs
|
| 41 |
+
Validation set: 121 260 docs
|
| 42 |
+
```
|
| 43 |
+
| Stats | fragment | | | avg document length | | avg summary length | | Documents |
|
| 44 |
+
|-------------|----------|---------------------|--------------------|--------|---------|--------|--------|--------|
|
| 45 |
+
| __dataset__ |__compression__ | __density__ | __coverage__ | __nsent__ | __nwords__ | __nsent__ | __nwords__ | __count__ |
|
| 46 |
+
| cnc | 7.388 | 0.303 | 0.088 | 16.121 | 316.912 | 3.272 | 46.805 | 750K |
|
| 47 |
+
| sumeczech | 11.769 | 0.471 | 0.115 | 27.857 | 415.711 | 2.765 | 38.644 | 1M |
|
| 48 |
+
| cnndm | 13.688 | 2.983 | 0.538 | 32.783 | 676.026 | 4.134 | 54.036 | 300K |
|
| 49 |
+
| xsum | 18.378 | 0.479 | 0.194 | 18.607 | 369.134 | 1.000 | 21.127 | 225K|
|
| 50 |
+
| mlsum/tu | 8.666 | 5.418 | 0.461 | 14.271 | 214.496 | 1.793 | 25.675 | 274K |
|
| 51 |
+
| mlsum/de | 24.741 | 8.235 | 0.469 | 32.544 | 539.653 | 1.951 | 23.077 | 243K|
|
| 52 |
+
| mlsum/fr | 24.388 | 2.688 | 0.424 | 24.533 | 612.080 | 1.320 | 26.93 | 425K |
|
| 53 |
+
| mlsum/es | 36.185 | 3.705 | 0.510 | 31.914 | 746.927 | 1.142 | 21.671 | 291K |
|
| 54 |
+
| mlsum/ru | 78.909 | 1.194 | 0.246 | 62.141 | 948.079 | 1.012 | 11.976 | 27K|
|
| 55 |
+
| cnewsum | 20.183 | 0.000 | 0.000 | 16.834 | 438.271 | 1.109 | 21.926 | 304K |
|
| 56 |
+
#### Tokenization
|
| 57 |
+
Truncation and padding were set to 512 tokens for the encoder (input text) and 128 for the decoder (summary).
|
| 58 |
+
## Training
|
| 59 |
+
Trained based on cross-entropy loss.
|
| 60 |
+
```
|
| 61 |
+
Time: 3 days 10 hours
|
| 62 |
+
Epochs: 1072K steps = 10 (from 10)
|
| 63 |
+
GPUs: 4x NVIDIA A100-SXM4-40GB
|
| 64 |
+
eloss: 2.824 - 1.745
|
| 65 |
+
tloss: 4.559 - 1.615
|
| 66 |
+
```
|
| 67 |
+
### ROUGE results per individual dataset test set:
|
| 68 |
+
|
| 69 |
+
| ROUGE | ROUGE-1 | | | ROUGE-2 | | | ROUGE-L | | |
|
| 70 |
+
|------------|---------|---------|-----------|--------|--------|-----------|--------|--------|---------|
|
| 71 |
+
| dataset | Precision | Recall | Fscore | Precision | Recall | Fscore | Precision | Recall | Fscore |
|
| 72 |
+
| cnc | 30.13 | 22.56 | 25.21 | 10.53 | 8.01 | 8.9 | 22.47 | 16.92 | 18.86 |
|
| 73 |
+
| sumeczech- | 26.6 | 19.66 | 22.01 | 8.17 | 6.12 | 6.82 | 19.93 | 14.81 | 16.54 |
|
| 74 |
+
| cnndm | 41.8 | 38.41 | 38.94 | 18.74 | 17.14 | 17.4 | 29.69 | 27.33 | 27.68 |
|
| 75 |
+
| xsum | 38.27 | 33.62 | 35.16 | 14.39 | 12.69 | 13.25 | 30.77 | 27.05 | 28.29 |
|
| 76 |
+
| mlsum-tu | 52.44 | 44.36 | 46.39 | 36.98 | 31.51 | 32.86 | 46.04 | 39.04 | 40.8 |
|
| 77 |
+
| mlsum-de | 42.19 | 40.5 | 40.7 | 28.8 | 28.51 | 28.37 | 38.95 | 37.7 | 37.79 |
|
| 78 |
+
| mlsum-fr | 34.57 | 27.74 | 29.95 | 16.27 | 13.04 | 14.08 | 27.18 | 21.89 | 23.6 |
|
| 79 |
+
| mlsum-es | 30.93 | 26.41 | 27.66 | 11.42 | 9.85 | 10.28 | 25.12 | 21.59 | 22.55 |
|
| 80 |
+
| mlsum-ru | 0.65 | 0.52 | 0.56 | 0.15 | 0.15 | 0.15 | 0.65 | 0.52 | 0.56 |
|
| 81 |
+
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
# USAGE
|
| 85 |
+
```
|
| 86 |
+
soon
|
| 87 |
+
```
|