Hrm Dsdvsd free essay sample

The process of Case-Based Reasoning (CB) is to find a case that is similar to the rent situation, modify the solution to FLT the current situation and then to store the case in the case-base. These processes can be carried out using the program Caspian. The purpose of this section is to describe the language used in the case file for specifying cases (CALLS), and to describe how Caspian uses this information to create the case-base. You create a case base In CALLS using your favorite editor and then load It Into Caspian. Caspian checks that It Is a legal CALLS program as It loads It In (see the Caspian document for more details of error messages). 2. General Structure of a Case In CALLS, a case is similar too record in a database. The basic unit is a field which may contain a string, a number, an enumeration symbol or a list. A list item may be any of the four basic types. There are a number of differences between a CALLS case and an ordinary database record: 1) There are two groups of fields in a case. The first group describes the situation and the nature of the problem. The second group describes the solution to the problem. ) It is possible for a field to be omitted. This Is only true for those fields which are not used for Indexing (see below). Both field lists must contain at least one field. The problem section must use at least one field as an Index. 3) Each case has a value known as the result. This can be the value SUCCESS, or the value FAILURE, or a value between ;5 and 5. 4) A case has a name associated with it. Enumeration fields defined in the problem section of a case may be used as indexes. General Syntax of the case file The case file consists of a number of blocks of code.The overall syntax is: Introduction Case Definition Index Definition [Modification Definition] [Repair Rule Definition] Case Instance {Case Instance) CALLS description document FL . 3 O UH Abernathy, February 1995 The Introduction block contains introductory text which gets displayed when the program Caspian has finished checking the case file. The Case Definition defines the types and the weights of the fields that may appear in a case. The information in the Case Definition is used for typecasting input cases while the weights are used to aid the case-matching process (described below). The Index Definition defines the fields used as indexes when searching for a matching case. A case base should have at least one field used as an index. The type of an index field must be an enumerated type. The Modification definition defines the modification rules. The purpose of the modification definition is twofold: 1) It provides a means of specifying that certain symbols or numbers are similar, for matching purposes. 2) It provides a means of specifying symbols as abstractions of others. This is useful for making the search more general or for defining generalizes cases. The Repair Rule definition contains the repair rules. The repair rules are used to modify the solution retrieved from the case-base, to make it more suitable for the current tuition. Both the Modification definition and the Repair rule definition may be omitted, but a completed system should contain both, if the system is to be a true CB system. The last set of blocks are the case instances. These are the cases that make up the case base. The case file must contain at least one case instance and will need to be seeded with many cases initially to be of any use. A description of the case matching process is given in section 9. . The Introduction Section The Introduction block defines introductory text which is displayed when Caspian is run. The purpose of the text is to help the user understand the purpose of the casabas. The syntax definition is: Introduction = INTRODUCTION IS intro-text END intro-text = string { , string } Within the strings themselves, is used to represent the inline character and t is used to represent tab. This is not true of strings used elsewhere in the case file. 2 4. The Case Definition This is the definition of the syntax of the case definition block. Keywords are given in case-Definition = CASE DEFINITION IS field-definitions SOLUTION DEFINITION IS field-definitions END lied-definitions = fielded {fielded} fielded = FIELD filename TYPE IS filleted [WEIGHT IS positive] ; filleted STRING I NUMBER I LIST I enumeration enumeration = ( symbol { , symbol } The symbols filename and symbol are identifiers. The symbol positive is a positive number. The purpose of the case definition block is to define the fields contained in a case. It consists of a series of field definitions each of which defines the name, type and optionally the weight of each field.The fields defined in the first field definition list are known as the problem fields. The Leeds defined in the section field definition list are known as the solution fields. In a acquaintance a field may only appear in the section for which it was defined. The weight value is used in matching cases. The larger the weight, the more important the field is. If the weight is not included in the field definition, the default value of one is assumed as the weight for the field. The case definition is used to perform type-checking on the cases and user input to Caspian. 5.The Index Definition The syntax of the index definition block is as follows: Index-Definition = INDEX DEFINITION IS indexes END indexers = indexed {indexed} indexed = INDEX ON filename The filename symbol is an identifier denoting the name of a field which must be an enumerated type. The purpose of the index definition is to define which fields are to be used as indexes. This information is used by Caspian to generate an index structure to improve the search. Further information on indexes is given in the section on the modification rules. A field should be indexed only if the feature it represents is very important.Indexing on an appropriate field helps to ensure that the retrieved cases re in the right ball-park. There should be no need to use a weight on an indexed field. Experience of using indexes shows that they are rather less useful than had been expected. They are intended to prune the matching process so that many cases are rejected early and a lot of work is saved. Unless you have a very large case base, then you might as well match on all fields. Dont have more than one or two indexes. 3 The modification definition consists of a list of definitions known as the modification rules.The term modification rule is used slightly differently here in that the concern s with defining certain values to be similar, thus guiding the matching process, rather than performing modifications to the retrieved solution. The syntax of the Modification Definition is as follows: Modification-Definition = MODIFICATION DEFINITION IS modes END modes mode {mode} mode= modularize I numberless modularize = FIELD filename ABSTRACTION symbol IS enumeration numberless = FIELD filename SIMILAR RANGE number TO number ; There are two types of modification rule: 1) Rules on fields which are enumerated types.These have a dual function. First, the humbly after the keyword ABSTRACTION is considered to be an abstraction of the symbols in the following enumeration list. This serves two purposes: a) When matching cases, if the user enters an abstraction symbol, the case matcher will match with the abstraction symbol or any of the symbols in the following enumeration list. B) In a similar fashion, when examining an enumeration value to activate a repair rule, using an abstraction will cause a match with the abstraction symbol or any of the symbols in the enumeration list.Second, when cases are being matched using the weights (see case matching, below) ND two enumeration values are being compared, if neither symbol is used elsewhere as an abstraction and both symbols are in the the same enumeration list within a modification rule, then the symbols are defined to be similar. The result of comparing two symbols and finding them to be similar will cause the case matcher to return a weight for that field equal to three-quarters of the fields full weight. ) Rules on fields which are numbers. The purpose of these rules is to define ranges over which numbers are considered similar. If two number fields have values which are both contained in a range defined by a edification rule, then the numbers are similar and this will cause the case matcher to return a weight value equal to three-quarters of the fields full weight. 7.