Orion knowledge network technology can be used in many applications, as shown by the other White Papers on this site. There is an underlying purpose in each of these applications, and it is to provide

A rapid and intelligent response in a dynamic environment

The combination of rapid and intelligent is important – a rapid, preprogrammed but dumb response may get the wrong answer or annoy a customer, an intelligent but slow response – "Please wait while we connect you to our next available operator" – may also annoy the customer or prevent a business interaction or data project from succeeding. Much of computer business analysis can be like deciding what a cloud looked like long after it has blown away - the purpose of analysis should be to decide and predict within an appropriate timescale, not confirm in hindsight.

Orion knowledge network technology combines analysis and experience, and fuses them together at the most basic level. It uses analysis where possible, but there are many areas where analysis won’t reach – the psychology of car buying for example – so experience must be used to fill in the gaps. Experience alone won’t do – it provides no way to exercise corporate control on what happens in the future. The combination of analysis and experience must respond quickly to changes in the environment, so acquisition of experience needs to be both rapid and controllable – the analytic part can control what is learnt, and the experiential part can then control how things are analyzed. Stale experience is worse than useless - there must be means for using experience while it still has value, which may compress the timescale for its introduction into an operational system.

Knowledge Management

The basic application of knowledge networks is knowledge management - the capture, merging, maintenance and deployment of knowledge. The other applications listed below emphasise one particular aspect, particularly capture or deployment. The Orion approach to knowledge management is to first emulate the properties of the human mental apparatus - an active, realised, self-modifiable network. With this technology, knowledge management reduces to determining what are the active structures needed to represent the activity and topology implied in messages about knowledge - the active structures then performing all the necessary operations.

Some applications using knowledge networks in a dynamic environment where the combination of a rapid and intelligent response is important:

Project Management - Development Projects

A development project can be changing its structure from week to week, and with software projects in particular, a good idea can completely reshape the structure of the project. Conventional CPM tools assume no change in structure, and everything needs to be done - there is no way to handle alternatives, contingencies, risk. The CPM plan is usually pasted on the wall at the start, and then ignored, precisely because of its inflexibility. A plan using knowledge networks can be used as a decision system for how to change the project, as well as how to run it.

High Level Planning and Scheduling

Is the system being planned or scheduled capable of changing faster than the planners can update the plan or schedule. Does the scheduling system make heroic and unfounded assumptions about the behaviour of the system so the scheduling method can be used - usually it does, with the result that the efficiency of the overall system becomes hostage to a program. Knowledge networks can change their own topology in response to a change in the behaviour of the system.

Cognitive Core for a web site - a product of Tupai eBusiness Systems

Selling on a web site to discriminating or finicky customers is a highly dynamic task. A customer must be responded to in at most a few seconds, there are only a few minutes to convince them to buy, the virtual sales force needs to be updated about engagements in minutes, sales data needs to be analyzed and implemented operationally in hours, corporate control needs to be exercised in days. Where test marketing is being used and there is little experience to go on, the time scale requires learning on the run. What the various time scales say clearly is that re-programming cannot be involved in any of these feedback and control paths, and systems that do not meet any one of the response times will fail in their purpose.

Bid Generation, Evaluation and Negotiation - a product of Tupai eBusiness Systems

Purchasing components that meet complex requirements is a process that needs to be completed in days. The system may start with little knowledge about what is to be bought, sending out a naïve request and mining the responses until it can build a full specification, or it may already have a detailed specification and has to match this against what is being offered and find the best match. The system needs to change its knowledge rapidly, so re-programming is not an option. Instead it needs to be able to combine pieces of knowledge it is given from many sources – both inside and outside the organization – and respond appropriately as buyer or seller introduce new clusters of features or constraints on how the bids should be assessed.

Data Mining - a product of Tupai eBusiness Systems

Data Mining becomes an essential part of a system using experience to guide its actions, but it cannot be the conventional kind of Data Mining, where mining produces rules which are interpreted by programmers into programs, or neural networks are laboriously trained on static data. Tupai’s Data Mining is an element in a process that requires rapid turnaround, so experience in the database must be able to be implanted into an operational system in minutes. The Tupai Data Miner can continue on a voyage of knowledge discovery, finding increasingly complex relations in the data, but this is not essential to making the experience held in the database operationally available within the organization.

Time Series Analysis

If a system is to provide data mining, it will also need to extract useable experience from data that carries a strong time element. A knowledge network approach to time series analysis gets a global sense of the data, uses other knowledge to control its projections, then plugs those projections back into operational systems to close the loop.

Data Cleaning for Data Warehouses - a product of Tupai eBusiness Systems

A Data Warehouse does not give the impression of a dynamic environment, but the business it mirrors probably is, and is changing its structure faster than the Data Warehouse can keep up. When there are millions of records to transform and validate, a rapid and intelligent response becomes crucial. The usual approach is to bring in highly skilled data analysts, who spend months trying to understand the data, then write rules to filter the data, then test the rules. The process teaches the analysts a lot about the business, but when they walk out, so does their knowledge, and in the meantime new errors are creeping into the flow of new data that weren’t present before. Tupai’s approach is to mine the dirty data, setting up relations that give the system a global sense of what the data means, allowing it to use that sense when it needs to validate a particular field – it follows the dictum "Think global, act local". A rapid response here means no more than a second or so on each record, and an intelligent response means sophisticated error detection and error correction, with the system being malleable enough to adapt as new errors crop up.

Personalized Advertising over Media Channels - Tupai

Adverts need to be placed in a stream of media. The more closely targeted the ads to the individual viewer, the more chance of vendor success. A rapid and intelligent response means picking up quickly on the viewer’s current state – are they absorbed, or flicking between channels – and can an ad be found which suits the content currently being displayed.