Infer information from Tweets. Useful for human-centered computing tasks, such as sentiment analysis, location prediction, authorship profiling and more!
When we do an online update of the system using gh-46, it might be nice to display to the user how this misclassification report affected our performance on our test set. We would need to:
Calculate the system performance before the report.
Update the system with the training instance.
Calculate the system performance after the report.
Display the before and after performance to the user.
To reduce processing time, I would also suggest calculating the before-performance previous to receiving the misclassification report. In addition, storing the test dataset in memory with the features for each instance already extracted should save additional time.
In addition, if we are calculating the system performance anyway, it would be insightful to have that data shown on a new system information page.
When we do an online update of the system using gh-46, it might be nice to display to the user how this misclassification report affected our performance on our test set. We would need to:
To reduce processing time, I would also suggest calculating the before-performance previous to receiving the misclassification report. In addition, storing the test dataset in memory with the features for each instance already extracted should save additional time.
In addition, if we are calculating the system performance anyway, it would be insightful to have that data shown on a new system information page.