Understanding machine learning

Dictionary.com defines “machine learning” as:
machine learning

  1. a branch of artificial intelligence in which a computer generates rules underlying or based on raw data that has been fed into it

The term was coined back in 1959 by a pioneer in artificial intelligence named Arthur Samuel, and evolved from the deep study of pattern recognition and computational learning. Machine learning leverages mathematical optimization and statistics to build algorithms (analytical models) that can learn from data and make predictions. However, it is also important to differentiate between two key types of machine learning: supervised and unsupervised.

Supervised Machine Learning (SML). SML is the most common form of machine learning and is like a student teacher relationship. We provide known samples as input data and know the result or required output. A simple example would be training a system to identify pictures of mammals. We provide input photos of bears, foxes and lions (because we know what they are), and use algorithms that tie the name or label to the input data (photo).

Unsupervised Machine Learning (UML). In UML, often associated with true artificial intelligence, the machine identifies patterns on its own without supervision of guidance. Using the example above, when you feed it 20 pictures of mammals, and it begins to group, or cluster, them into similar buckets. Essentially, we give the machine a data set and let it analyze, infer and find hidden structure.

How does supervised machine learning work with Smart Capture®?

In the capture industry, there is quite a bit of terminology misuse around “training” and “learning.” Legacy capture applications utilize “training” to teach their applications how to interpret, classify and extract data. This legacy “training” allows for an administrator or user to create a template for processing, which has a one-to-one relationship. For example, if you want to process an invoice from vendor A, then the user creates a training template for vendor A, which only applies to vendor A.

Modern, Smart Capture® technology relies on laser-focused, analytics and supervised machine learning algorithms created specifically for unstructured document content (read our patent to learn more: Ephesoft Document Analytics Patent). These algorithms allow the computer to “see” document structure, elements and then interpret them as a human would. The application leverages supervised machine learning, and can be taught by an end user or administrator simply by interacting with or providing samples. Below are two areas of learning:

  1. Document Classification – After configuring document types (labels), a user can then drag samples into a learning interface where documents are consumed as input data. Then they are passed through the Smart Capture® algorithms. Once the model is built, documents can be identified during processing. As each page is analyzed within Ephesoft’s platform, now the system knows when one document ends and another begins, eliminating the need for page separators or barcodes.
machine learning documents
  1. Data Extraction – Using a simple point and click interface, users can teach the system about the data. With each teaching event, the system gains deep understanding of document data and how it exists on a page or within a document.
machine learning ocr

What are the benefits of supervised machine learning in Smart Capture®?

Here are some of the key benefits of machine learning enabled Smart Capture®:

Adaptability – Supervised machine learning guides the system to interpret samples that were not provided in training. Using a limited sample set of documents, the system can understand a broad range of unstructured document inputs. Unlike a legacy system that requires a complete sample set, this adaptability leads to less initial configuration and improves over time.

Reduced Time to Production – Teaching the system is far less time intensive than custom code or writing extraction expressions with Smart Capture®. The faster the automation platform is in place, the quicker your organization starts saving money and time.

Reduced Support Costs – Legacy systems are costly to support and usually require expensive resources and expertise to maintain. Smart Capture® systems are enabled for users to provide teaching versus a costly development or consulting resource.

Teach through Code – Modern capture systems can be trained through the application interface as well as use Web Services APIs to facilitate training from other sources. Your enterprise apps can have access to document supervised machine learning models.

To learn more about our patented, document-focused supervised machine learning platform, contact us at sales@ephesoft.com or schedule a demo.