Goal: text classifier for tweets containing job offers
-
Load dump of tweets (Settings -> Content -> Your Twitter archive -> Request your archive)
-
Manually mark tweets with a job posting as positive samples (True)
-
Sample and format the dataset as a CSV: one tweet per line with balanced classes and shuffled samples
-
Train a classifier to detect job tweets, e.g. fastText, or support vector machine with RBF kernel trained on uni-grams or random forest.
-
Explain the classifier predictions with LIME