Update README.md
Browse files
README.md
CHANGED
@@ -2,6 +2,9 @@
|
|
2 |
tags:
|
3 |
- text-2-text-generation
|
4 |
- t5
|
|
|
|
|
|
|
5 |
---
|
6 |
|
7 |
# Model Card for qcpg-sentences
|
@@ -11,7 +14,7 @@ tags:
|
|
11 |
# Quality Controlled Paraphrase Generation (ACL 2022)
|
12 |
> Paraphrase generation has been widely used in various downstream tasks. Most tasks benefit mainly from high quality paraphrases, namely those that are semantically similar to, yet linguistically diverse from, the original sentence. Generating high-quality paraphrases is challenging as it becomes increasingly hard to preserve meaning as linguistic diversity increases. Recent works achieve nice results by controlling specific aspects of the paraphrase, such as its syntactic tree. However, they do not allow to directly control the quality of the generated paraphrase, and suffer from low flexibility and scalability.
|
13 |
|
14 |
-
<img src="/assets/images/ilus.jpg" width="40%">
|
15 |
|
16 |
> Here we propose `QCPG`, a quality-guided controlled paraphrase generation model, that allows directly controlling the quality dimensions. Furthermore, we suggest a method that given a sentence, identifies points in the quality control space that are expected to yield optimal generated paraphrases. We show that our method is able to generate paraphrases which maintain the original meaning while achieving higher diversity than the uncontrolled baseline.
|
17 |
|
@@ -20,7 +23,7 @@ The code for training, evaluation and inference for both `QCPG` and `QP` is loca
|
|
20 |
|
21 |
Make sure to run `QCPG/scripts/prepare_data.sh` and set the missing datasets directories accordingly before training!
|
22 |
|
23 |
-
<img src="/assets/images/arch.png" width="90%">
|
24 |
|
25 |
## Trained Models
|
26 |
|
@@ -285,5 +288,4 @@ model = AutoModelForSeq2SeqLM.from_pretrained("ibm/qcpg-sentences")
|
|
285 |
|
286 |
|
287 |
```
|
288 |
-
</details>
|
289 |
-
|
|
|
2 |
tags:
|
3 |
- text-2-text-generation
|
4 |
- t5
|
5 |
+
- augmentation
|
6 |
+
- paraphrase
|
7 |
+
- paraphrasing
|
8 |
---
|
9 |
|
10 |
# Model Card for qcpg-sentences
|
|
|
14 |
# Quality Controlled Paraphrase Generation (ACL 2022)
|
15 |
> Paraphrase generation has been widely used in various downstream tasks. Most tasks benefit mainly from high quality paraphrases, namely those that are semantically similar to, yet linguistically diverse from, the original sentence. Generating high-quality paraphrases is challenging as it becomes increasingly hard to preserve meaning as linguistic diversity increases. Recent works achieve nice results by controlling specific aspects of the paraphrase, such as its syntactic tree. However, they do not allow to directly control the quality of the generated paraphrase, and suffer from low flexibility and scalability.
|
16 |
|
17 |
+
<img src="https://github.com/IBM/quality-controlled-paraphrase-generation/raw/main/assets/images/ilus.jpg" width="40%">
|
18 |
|
19 |
> Here we propose `QCPG`, a quality-guided controlled paraphrase generation model, that allows directly controlling the quality dimensions. Furthermore, we suggest a method that given a sentence, identifies points in the quality control space that are expected to yield optimal generated paraphrases. We show that our method is able to generate paraphrases which maintain the original meaning while achieving higher diversity than the uncontrolled baseline.
|
20 |
|
|
|
23 |
|
24 |
Make sure to run `QCPG/scripts/prepare_data.sh` and set the missing datasets directories accordingly before training!
|
25 |
|
26 |
+
<img src="https://github.com/IBM/quality-controlled-paraphrase-generation/raw/main/assets/images/arch.png" width="90%">
|
27 |
|
28 |
## Trained Models
|
29 |
|
|
|
288 |
|
289 |
|
290 |
```
|
291 |
+
</details>
|
|