Added Synonym insertion #160
Conversation
|
This is already implemented and is on the verge of merging: #51 |
|
@JosephSefara please see |
|
@kaustubhdhole This implementation is different from #51 .
|
|
Thank you for the clarification @JosephSefara. I think these changes look good to me. |
|
Please pull main once in your branch. |
|
Okay, I just have one comment: Do you think it might be better to write synonynms besides the original words so that the sentence is still well-formed? |
| https://github.com/dsfsi/textaugment | ||
| """ | ||
| tasks = [TaskType.TEXT_CLASSIFICATION, TaskType.TEXT_TO_TEXT_GENERATION] | ||
| languages = ["en"] |
There was a problem hiding this comment.
Also, please add keywords here.
There was a problem hiding this comment.
Also, I would also recommend adding the robustness evaluation for your PR that can be added to the leaderboard.
There was a problem hiding this comment.
- keywords added
- readme and test.json contains the results of the Robustness Evaluation for
- Text Classification
- Text Generation
|
@vukosim @JosephSefara ping! |
|
Also, one more thing: do you think it would be better to include the synonym within a bracket? |
@kaustubhdhole, I don't understand you question but we are inserting a synonym next to its original word. Augmentation sometimes is about adding noise to the sentence hence the sentence might not be well-formed but still retains original context. E.g. stopwords removal #268 removes stop words, thus the sentence might not be well-formed. |
@kaustubhdhole
Why remove special characters?
|
No description provided.