Thanks to optical character recognition, the address on your incoming mail may not be scanned by human eyes until the envelope has nearly reached your mailbox.
Machine learning technologies now enable the United States Postal Service (USPS) to correctly read and sort virtually every piece of mail that passes through the system using handwriting recognition algorithms.
The diversity of handwriting styles combined with millions of possible U.S. addresses makes this seem like an incredible feat. In fact, it’s a very complex process, but a machine takes care of it in a fraction of a second.
Each envelope is scanned and converted to a digital image so it can be electronically searched. Supercomputers then begin identifying address information, starting with numerical characters like street number and zip code. Those numbers are matched against a database of over 154 million addresses, and a list of possible recipients is determined.
The software then goes to work deciphering the handwritten letters of the street name. As characters are identified, they are again matched against the potential candidates until the list is narrowed to one recipient. Once the computer returns a result, the guess is given a confidence rating indicating how certain the computer is that the suggestion reflects the actual address. If the confidence rating is high enough, then the letter is routed to the appropriate local post office.
Extremely illegible cases aside, this whole process takes place without any human intervention approximately 155 billion times per year.
The USPS is considered something of a pioneer in the field of machine learning, and they were one of the first organizations to start making substantial investments in the technology. After researching options for over a decade, they deployed their first computer prototype capable of reading handwritten text in 1997 with the help of the University of Buffalo’s Center for Excellence in Document Analysis and Recognition (CEDAR). The prototype correctly identified the addresses on only 10% of envelopes it read with a 2% error rate.
Today, machines at the USPS process approximately 98% of hand-addressed mail and 99.5% of type-written addresses.
Machine learning means computers can autonomously adapt and learn from previous experiences, making them able to emulate the thought process behind human decision-making. Algorithms quickly process large volumes of complex data to uncover patterns without being actually programmed to do so, and deliver critical insights in real-time.
Machine learning is by no means a new technology, but it’s gained momentum in recent years as businesses search for innovative ways to learn and forecast using Big Data. When used properly, the result is high-value predictive analytics that can improve decision-making and enable businesses to react and compete at new levels.
Enterprises are already using machine learning techniques in a variety of applications to drive operational efficiencies and enhance the customer experience:
- Text classification and rule-based filtering are methods used by webmail providers to identify, block and decrease the frequency of spam email messages.
- Personalized content recommendations based on viewing history are made possible on media sites like Netflix; the same technology is used by online retailers to offer product recommendations based on purchase history.
- Customer sentiment analysis, or opinion mining, uses linguistic rules to improve everything from marketing efforts to customer service.
- Intuitive security analytics help organizations boost data security by protecting interactions among users and applications to help safeguard their most valuable digital assets.
Machine learning is driving a paradigm shift in the areas of business intelligence and enterprise software. As the ability to capitalize on Big Data improves, enterprises will increasingly leverage distributed machine intelligence to address a wide range of social and business needs.
It’s an exciting time for enterprise software, an industry where every emerging technology increases the collaboration between humans and machines in order to deliver critical business benefits.