search

l294265421 / ASOTE

Aspect-Sentiment-Opinion Triplet Extraction (ASOTE)
https://arxiv.org/pdf/2103.15255.pdf
73 stars 11 forks source link

How can be re-used in a totally new and different dataset? #3

Open Jurys22 opened 2 years ago

Jurys22 commented 2 years ago

Do the users need to annotate the dataset for training,testing,evaluation? or do the users can re-use already annotated datasets available?

I would like to test this repo against tweets - would it be work without data annotation?

Thank you for the open source code!

l294265421 commented 2 years ago

Do the users need to annotate the dataset for training,testing,evaluation? or do the users can re-use already annotated datasets available?

I would like to test this repo against tweets - would it be work without data annotation?

Thank you for the open source code!

Thanks for your interest in our work.

  1. If you have an annotated tweet dataset, you can directly use the codes in this repo on the dataset. The simplest way to test this repo on the tweet dataset have three steps: (1) Convert the tweet dataset and make it have the same format as the datasets used in this repo, for example, https://github.com/l294265421/ASOTE/tree/main/ASOTE-data/absa/ASOTE-v2/rest14/asote_gold_standard. (2) Replace the rest14 dataset (https://github.com/l294265421/ASOTE/tree/main/ASOTE-data/absa/ASOTE-v2/rest14/asote_gold_standard) with the tweet dataset. (3) Run the commands related to the rest14 dataset in PBF section of README (https://github.com/l294265421/ASOTE). Then, you can obtatin the model peformance on your tweet dataset.
  2. If your tweet dataset has not been annotated, you can train our model on the datsets provided in this repo, then extract opinion triplets from the tweet dataset using the trained model (cross-domain opinion triplet extraction). To achieve this, you can also follow the three steps above, but only replace the test dataset with your tweet data.