Clear instructions/information panel on using Document Classifier trained with custom dictionaries

Document classifiers are extremely valuable in programmatically sorting document types based on its content. The document classifier in WDS can be trained solely on parts-of-speech or in combination with additional assets such as custom dictionaries. Upon deployment of the document classification models trained on both Part-of-Speech as well as custom dictionaries, it turns out that the labels disappear, while the document classification models trained without the custom dictionaries still yield labels. This puzzled us a bit and it took some experimentation to sort out the cause of this. We’ve managed to get the labels up and running with a doc classifier that was trained using POS combined with a custom dictionary. To make a long story short, it turns out that if you train document classification models using custom dictionaries, you need to activate the dictionaries that you’ve used to train the document classifier as enrichments to your analysis pipeline. Otherwise your doc classifier won’t yield labels. There is no warning or error message displayed in case you don’t add the dictionary enrichment on which the doc classifier is dependent on. Perhaps a suggestion to put this more explicitly into the documentation.

Needed By

Yesterday (Let's go already!)

Post comment

By clicking the "Post Comment" or "Submit Idea" button, you are agreeing to the IBM Ideas Portal Terms of Use.
Do not place IBM confidential, company confidential, or personal information into any field.

Please enter your email address

RELATED IDEAS

Clear instructions/information panel on using Document Classifier trained with custom dictionaries