It may seem to many that the airy intellectualiasm associated with artificial intelligence, and in particular, expert systems, is best left in the classroom. But one artificial intelligence paradigm, case-based reasoning, is not only being used in a useful commercial context – help-desk software, it is actually making money for the company concerned and enabled […]
It may seem to many that the airy intellectualiasm associated with artificial intelligence, and in particular, expert systems, is best left in the classroom. But one artificial intelligence paradigm, case-based reasoning, is not only being used in a useful commercial context – help-desk software, it is actually making money for the company concerned and enabled it to float successfully on Nasdaq. Inference Corp, based in El Segundo, but shortly on the move to Novato, California, has been using artificial intelligence since 1979, but had difficulty convincing investors that it was worth the risk. So from 1985 the company concentrated on software development tools, before shipping its first case-based reasoning customer service and support software product, CBR Express in May 1991. A new chief executive, Peter Tierney, formerly marketing vice-president at Oracle Corp was appointed at the start of that year. The follow-up, CBR2 was released in March this year.
The company finally managed to get its initial public offering away at the end of July at $11 per share, raising around $17m after expenses. The shares are currently trading on Nasdaq at around $17 each. Inference now concentrates on the CBR2 family of products, having spun off the software developement tools business to Brightware Inc, a wholly-owned subsidiary with its main products, ART, ART-IM AND ART Express in May. Limbex Corp, headed by Inference chairman Alexander Jacobson also has a worldwide licence for Inference products. Case-based reasoning is a fairly recent development in artificial intelligence that attempts to mirror the flexibility and imprecise nature of human knowledge derived from previous experiences. Instead of constructing a set of rigid rules using if-then statements, as is the case with early rule-based expert systems, case-based reasoning maps current problems onto libraries of previous cases to see if a problem has been solved in a similar way before. If it has, then the same solut ion is re-usable. If it has not, then the nearest solution is found, and adapted to fit the new one. Obviously if nothing remotely like the new problem has occurred before, then the problem-solver will have to start from scratch and add the solution to the library. In practice though, companies that use case-based reasoning know most of the problems they are likely to come across in advance, and so can construct the libraries of solutions accordingly. The gaps in the knowledge are filled by asking relevant questions. Inference’s initial foray into case-based reasoning help-desk systems, CBR Express, was an application shell for text input. The text was split into trigrams, and matched against all previous cases, achieving a 20% success rate, according to the company. The system then suggested lists of questions, the answers to which would enable it to home in on matching cases if it did not hit upon one in the first pass. For the first time, according to Chris McKee, vice-president of Northern European operations at Inference, analysts and potential investors understood what the product did, and the commercial viability of case-based reasoning. An updated graphical version of CBR Express is part of the CBR2 range. The authoring facility enables cases to be constructed. A case is made up of a title, a description, guiding questions, and actions that need to be taken to resolve the problem.
By Nick Patience
The cases author can influence the importance of questions by weighting them. A test facility evaluates the quality of the cases and Express Generator creates case bases from documents, such as technical specifications, either in ASCII or Microsoft Corp Word formats. The Express help desk module in CBR2 provides call tracking problem management, system administration and management reporting facilities, from a single Windows-like interface, according to the company. It also includes a training scheduler and an inventory manager. Perhaps suprising to some is that the development work for Inference’s CBR range is u
ndertaken not in California, but in the slightly less glamorous and balmy surroundings of Slough in Berkshire. The employees there are more productive as well as being cheaper, according to McKee. CasePoint 2.0 is the end-user tool to view case repositories created by CBR Express. Running as a client under Windows 3.1, HP-UX, IBM Corp OS/2 or Solaris, it can be used for file system access or via SQL links into relational databases. The company announced a CasePoint WebServer for internet access to CBR2 in April (CI No 2,651). Inference also supplies two pre-packaged internal help-desk knowledge bases for solving problems in 35 widely-used desktop packages. Last in the CBR2 family comes DP Umbrella SQL, under licence from Vycor Corp. It is an internal help-desk problem management tool, with various customisable components and application programming interfaces. Inference has also licensed knowledge gathered by ServiceWare Inc for sorting out the odd problem or two experienced by early experimenters with Windows95 as one of its pre-packaged knowledge bases. A different application for Inference’s technology is the increasing use of call avoidance systems. These came about as software became more complicated, users more demanding and experienced technical support staff that could placate disgruntled users more difficult and expensive to hire. The CasePoint WebServer is one example of this application, and Compaq Computer Corp and is one such manufacturer pre-loading Inference support software on its machines. Microsoft has signed to use CBR2 in future releases, but McKee would not reveal which ones.
IBM used Inference software to produce a CD-ROM to aid developers working under OS/2. Gateway 2000 Inc uses CBR2 on its help-desks worldwide, as do Hewlett-Packard Co and Rank Xerox Ltd for their printers and photocopiers respectively. UK supermarket operator J Sainsbury Plc will soon have a case-based reasoning system based on Inference’s CBR technology in each of its supermarkets in the UK for staff to solve the more straightforward problems that arise with the electronic point of sale system. Approximately half of the company’s 330 stores have the system running at present with the rest up by the year-end. CBR3 is currently in development with the specifications completed. It will be a 32-bit application developed under Windows95, released some time in early 1996. McKee said that all of Inference’s existing applications will run successfully under Windows95. Inference will not be resting on its laurels as far as expansion is concerned, as it has set aside around $20m for acquisitions of help-desk-type comp anies both in the US and elsewhere, according to McKee. There is already a healthy help-desk software market, but McKee reckons that case-based reasoning sets Inference apart from companies such as Software Artistry Inc and Scopus Technology Inc. Clarify Corp produces software to do much the same as Inference’s, but that is not suprising as it is developed using CBR Express. Despite being taught in many law and business schools for some years, case-based reasoning’s forays into expert systems had been relative failures in commercial terms. But with Inference and the companies that are using its products to create their own help-desk systems, it may have finally found its metier.