Loading...
Please wait, while we are loading the content...
Similar Documents
Interactive debugging of knowledge-bases ( extended abstract )
| Content Provider | Semantic Scholar |
|---|---|
| Author | Friedrich, Gerhard |
| Copyright Year | 2014 |
| Abstract | The identification and repair of faults in descriptions is a highly complex and expensive task. Descriptions can be formulated for many different purposes such as specifying configurations, hardware designs, algorithms, and requirements of systems. In order to develop technical systems by employing description languages, engineering departments require efficient processes for fault avoidance, testing, and debugging. In particular, the main diagnosis problem in the Engineering Center of Electronics at Siemens AG in order to provide high quality electronic systems has been the localization of faults in software programs, requirements, specifications, and design descriptions. This need for supporting the engineering of hardware and software systems increased over the last decades, since the volume and complexity of implemented functions are sharply increasing and additional formal descriptions for supporting the design processes are being employed. Moreover the application of declarative descriptions in various technical and non-technical areas (e.g. biomedical domains) is surging and requires new forms of fault localization techniques. Traditional debugging methods which are based on the control flow are not applicable since the fundamental idea of declarative descriptions is to avoid the formulation of a control flow. First-principles diagnosis [1; 2; 3] (also termed modelbased diagnosis (MBD) [4]) is a general approach for diagnosing systems and therefore can be applied not only to hardware systems but also to various forms of descriptions. Consequently, MBD offers a solution to localize faults in various kinds of logic based descriptions such as constraints, answer set programming, temporal logical specifications, and description logic. For the engineering of software systems tests play a crucial role. The test-driven development approach is based on an iterative cycle of formulating tests, discovering failures, and re-factoring the software. Our goal is to support this development approach also for knowledge engineering. However, for a given set of tests presenting the diagnoses (or just the minimal cardinality diagnoses) is not informative because of the huge number of possibilities to change a knowledge-base. Therefore, we extend static diagnosis to interactive diagnosis (as introduced by [3]) in order to localize precisely the faults in descriptions. In the following, we will briefly review the main contributions for the diagnosis of various descriptions and provide a summary of the latest developments of applying firstprinciple diagnosis to knowledge-bases which are expressed by a set of logical sentences. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://dx-2014.ist.tugraz.at/papers/DX14_Mon_keynote.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
