Prime vision explores the application of artificial intelligence and machine learning in the postal, parcel and logistics industry.
If computers can think what we want them to think, then they must be able to make our lives easier. This was the thesis behind Alan Turing’s paper on computing machinery and intelligence published in 1950. More than 60 years later, artificial intelligence systems based on so-called machine learning techniques are indeed deployed in postal, parcel and logistics operations.
The area of use is however relatively narrow at this stage, with learning systems focusing mostly on the domain of human visual interpretation such as the appearance of optical character and image recognition (also called OCR or vision) systems rather than the solution to a broader range of challenges.
Within artificial intelligence, the application of machine learning technologies is very popular. The success of these techniques results from the fact that explicit modeling is no longer required. Machine learning offers a way to solve complex challenges by means of “training” or “learning” the best fitting model and by using statistical and computational techniques, based on a large set of data examples.
Machines can “learn” by creating neural-like models of algorithms, which are able to effectively facilitate predictions and decision making for humans. As much as this seems abstract, it is the way these techniques can be applied and utilized to benefit day to day decision making, based on the analysis of complex models and algorithms.
The interrogation of historic data, in relation to a defined set of rules and actual, real-live data can provide new information. The success of machine learning lies in its ability to provide hidden insights, by crossing different sources of data and “learning” essentials about the central problem.
The benefits of machine learning are manifold. The postal and logistics industry is a perfect play area for data scientists, and this is the reason why leading postal organizations are in a rush to recruit data experts to support the drive to apply big data solutions.
Machine learning can be applied to not only improve operational processes but also to reduce operating losses, and to maintain a smooth and steady operation day to day. In this way this approach can help the decision making process and provide a smoother customer experience.
At An Post, in Ireland, for example, Prime Vision has developed a system which collects OCR read results. From these results additional alias data and names of organizations can be extracted, which forms the basis of an address learning system.
With advanced clustering and interpretation techniques, An Post is able to obtain clean and statistically significant entries that can be used for database enrichment. These entries are connected to a sequence level delivery point. Enrichment of the database is necessary, because the majority of names is not available in any other resource. By adding the alias data to the address database, An Post has been able to gain more than 10% of read rate at sequencing level.
To read more from Prime Vision’s white paper on machine learning, click here.
July 3, 2017