I have a sentiment analysis model to classify tweets into three class sentiment labels (positive, negative and neutral).
For training, I have two files. The first one in JSON format which includes all tweets IDs, text, ...etc
The second file in csv format and includes two columns, Tweets IDs and Labels (Negative, Positive and Natural).
For testing, I have test data in csv format which consists of tweet_id and text columns (the file includes Negative, Positive and Natural tweets)
In the training stage, do I need to use both files (JSON and csv) to create a training data with labels?
OR I use JSON data only (without labels)?
I use both files (JSON and csv) to create three datasets: negative_training_data, positive_training_data and natural_training_data?
How the training data should look like (The content text only or text with tweet_id)?