When you have to reclassify a huge dataset of unstructured information (like a list of URLs or financial transactions), witboost Tagging is the perfect tool for the job.
With witboost Tagging you can effectively and efficiently create and manage a complex set of tagging rules, based on a categorization tree, semantic rules based on keywords, phrases, or concatenation of radicals in a transaction.
The tagging rules themselves are classified and can grow in complexity in a process of continuous improvement and can also follow user-based rules creating results personalised results for each individua.
With witboost Tagging you can try different cathegorization rule sets and get a classification score for each, so you can then deploy the one that yields the best results
The rule-based expert approach proved time and again to be far superior to Machine Learning for automatic tagging and classification in many instances. For example in the analysis of a list of transaction a new non easily identifiable merchant might appear (that does not contain the type of merchant in its name) and at first the number of transaction is far too limited for ML to retrain on it and understand what it is. witboost Tagging is able to identify these new instances of a new merchant and single them out so that it can be quickly checked and if needed a new specific rule can be created on the fly.
witboost Tagging seamlessly integrates in the witboost environment, but it can also effectively function as a stand-alone solution.
Would you like to know more?