Query About Sentiment Analysis Capabilities for Jordanian Arabic Texts

[LamaAkmeel]

Hello CAMeL Tools Team,
I am new to Python and NLP and am currently working on a research project involving sentiment analysis of interview texts in Arabic, specifically from the Jordanian dialect. I would like to use CAMeL Tools for this purpose and have some questions:

1- Since the Dialect Identification feature is not available on Windows, can I still effectively use the sentiment analysis tool for Jordanian Arabic texts without identifying the dialect first? How critical is dialect identification to the performance of sentiment analysis within CAMeL Tools?

2- Does CAMeL Tools support sentiment analysis for Jordanian Arabic? If so, are there any specific configurations or pre-processing steps recommended for handling this dialect?

3- As I am new to NLP, could you advise whether CAMeL Tools provides pre-trained models for sentiment analysis that are suitable for Jordanian Arabic, or would I need to train a model myself? If training is required, could you provide some guidance or resources to get started?

[owo]

Hi @LamaAkmeel,

1- Dialect ID is not required for sentiment analysis and should work on Windows.

2- Our sentiment analysis models are fine-tuned on a mixture of Modern Standard Arabic, Egyptian, and Levantine (Jordan, Lebanon, Palestine and Syria). While I don't know what the performance of our system would be on Jordanian specifically, you can read more about the data used and evaluation results in our paper.

3- Unfortunately, camel-tools doesn't expose a training pipeline at the moment but we do aim to add that functionality in upcoming releases. As for alternative systems, you could try asking on the SIGARAB mailing group. @nizarhabash1 @balhafni any ideas?

I hope that helps.