File size: 6,554 Bytes
af4ba34 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 |
---
language: en
tags:
- SEGA
- data augmentation
- keywords-to-text generation
- sketch-to-text generation
license: apache-2.0
datasets:
- c4
widget:
- text: "<mask> Conference on Empirical Methods <mask> submission of research papers <mask> Deep Learning <mask>"
example_title: "Example 1"
- text: "<mask> machine learning <mask> my research interest <mask> data science <mask>"
example_title: "Example 2"
- text: "<mask> play basketball <mask> a strong team <mask> Shanghai University of Finance and Economics <mask> last Sunday <mask>"
example_title: "Example 3"
- text: "Good news: <mask> the European Union <mask> month by EU <mask> Farm Commissioner Franz <mask>"
example_title: "Example with a prompt 1"
- text: "Bad news: <mask> the European Union <mask> month by EU <mask> Farm Commissioner Franz <mask>"
example_title: "Example with a prompt 2"
inference:
parameters:
max_length: 200
num_beams: 3
do_sample: True
---
# SEGA-large model
**SEGA: SkEtch-based Generative Augmentation** \
**基于草稿的生成式增强模型**
**SEGA** is a **general text augmentation model** that can be used for data augmentation for **various NLP tasks** (including sentiment analysis, topic classification, NER, and QA). SEGA uses an encoder-decoder structure (based on the BART architecture) and is pre-trained on the `C4-realnewslike` corpus.
- Paper: [coming soon](to_be_added)
- GitHub: [SEGA](https://github.com/beyondguo/SEGA).
**SEGA** is able to write complete paragraphs given a *sketch*, which can be composed of:
- keywords /key-phrases, like "––NLP––AI––computer––science––"
- spans, like "Conference on Empirical Methods––submission of research papers––"
- sentences, like "I really like machine learning––I work at Google since last year––"
- or mixup~
**Model variations:**
| Model | #params | Language | comment|
|------------------------|--------------------------------|-------|---------|
| [`sega-large`](https://huggingface.co/beyond/sega-large) | 406M | English | The version used in paper |
| [`sega-large-k2t`](https://huggingface.co/beyond/sega-large-k2t) | 406M | English | keywords-to-text |
| [`sega-base`](https://huggingface.co/beyond/sega-base) | 139M | English | smaller version |
| [`sega-base-ps`](https://huggingface.co/beyond/sega-base) | 139M | English | pre-trained both in paragraphs and short sentences |
| [`sega-base-chinese`](https://huggingface.co/beyond/sega-base-chinese) | 116M | 中文 | 在一千万纯净中文段落上预训练|
---
### How to use
#### 1. If you want to generate sentences given a **sketch**
```python
from transformers import pipeline
# 1. load the model with the huggingface `pipeline`
sega = pipeline("text2text-generation", model='beyond/sega-large', device=0)
# 2. provide a sketch (joint by <mask> tokens)
sketch = "<mask> Conference on Empirical Methods <mask> submission of research papers <mask> Deep Learning <mask>"
# 3. just do it!
generated_text = sega(sketch, num_beams=3, do_sample=True, max_length=200)[0]['generated_text']
print(generated_text)
```
Output:
```shell
'The Conference on Empirical Methods welcomes the submission of research papers. Abstracts should be in the form of a paper or presentation. Please submit abstracts to the following email address: eemml.stanford.edu. The conference will be held at Stanford University on April 1618, 2019. The theme of the conference is Deep Learning.'
```
#### 2. If you want to do **data augmentation** to generate new training samples
Please Check our Github page: [github.com/beyondguo/SEGA](https://github.com/beyondguo/SEGA), where we provide ready-to-run scripts for data augmentation for text classification/NER/MRC tasks.
---
## SEGA as A Strong Data Augmentation Tool:
- Setting: Low-resource setting, where only n={50,100,200,500,1000} labeled samples are available for training. The below results are the average of all training sizes.
- Datasets: [HuffPost](https://huggingface.co/datasets/khalidalt/HuffPost), [BBC](https://huggingface.co/datasets/SetFit/bbc-news), [SST2](https://huggingface.co/datasets/glue), [IMDB](https://huggingface.co/datasets/imdb), [Yahoo](https://huggingface.co/datasets/yahoo_answers_topics), [20NG](https://huggingface.co/datasets/newsgroup).
- Base classifier: [DistilBERT](https://huggingface.co/distilbert-base-cased)
In-distribution (ID) evaluations:
| Method | Huff | BBC | Yahoo | 20NG | IMDB | SST2 | avg. |
|:----------:|:----------:|:----------:|:----------:|:----------:|:----------:|:----------:|:----------:|
| none | 79.17 | **96.16** | 45.77 | 46.67 | 77.87 | 76.67 | 70.39 |
| EDA | 79.20 | 95.11 | 45.10 | 46.15 | 77.88 | 75.52 | 69.83 |
| BackT | 80.48 | 95.28 | 46.10 | 46.61 | 78.35 | 76.96 | 70.63 |
| MLM | 80.04 | 96.07 | 45.35 | 46.53 | 75.73 | 76.61 | 70.06 |
| C-MLM | 80.60 | 96.13 | 45.40 | 46.36 | 77.31 | 76.91 | 70.45 |
| LAMBADA | 81.46 | 93.74 | 50.49 | 47.72 | 78.22 | 78.31 | 71.66 |
| STA | 80.74 | 95.64 | 46.96 | 47.27 | 77.88 | 77.80 | 71.05 |
| **SEGA** | 81.43 | 95.74 | 49.60 | 50.38 | **80.16** | 78.82 | 72.68 |
| **SEGA-f** | **81.82** | 95.99 | **50.42** | **50.81** | 79.40 | **80.57** | **73.17** |
Out-of-distribution (OOD) evaluations:
| | Huff->BBC | BBC->Huff | IMDB->SST2 | SST2->IMDB | avg. |
|------------|:----------:|:----------:|:----------:|:----------:|:----------:|
| none | 62.32 | 62.00 | 74.37 | 73.11 | 67.95 |
| EDA | 67.48 | 58.92 | 75.83 | 69.42 | 67.91 |
| BackT | 67.75 | 63.10 | 75.91 | 72.19 | 69.74 |
| MLM | 66.80 | 65.39 | 73.66 | 73.06 | 69.73 |
| C-MLM | 64.94 | **67.80** | 74.98 | 71.78 | 69.87 |
| LAMBADA | 68.57 | 52.79 | 75.24 | 76.04 | 68.16 |
| STA | 69.31 | 64.82 | 74.72 | 73.62 | 70.61 |
| **SEGA** | 74.87 | 66.85 | 76.02 | 74.76 | 73.13 |
| **SEGA-f** | **76.18** | 66.89 | **77.45** | **80.36** | **75.22** |
### BibTeX entry and citation info
|