Trainable tools in TotalAgility

As the name suggests, trainable or learn-by-example tools support the ability to learn automatically from examples, rather than relying on explicitly defined rules.

TotalAgility provides a mix of rules-based and learn-by-example tools to classify documents and extract data from the documents. These tools are simple to set up and easy to maintain. They only use rules to augment the results if needed.

The trainable tools in TotalAgility include:

• Content and layout classifiers

• Document separation

• Locators for fields and tables

These tools use a variation of standard and proprietary Machine Learning algorithms (Bayesian classifiers, k-Nearest Neighbor classifiers, Conditional Random Field, and more). Learning from examples means that a user or Administrator provides sample documents and teaches TotalAgility what document type they are, or where the fields are located (point and click), rather than creating rules involving keywords, formats, locations, and more. The key advantage of Machine Learning is that the system figures out the preceding details automatically, by learning from the samples.