Philip J. Sallis

Metrics for database systems: an empirical study

An important task for any software project manager is to be able to predict and control project size and development effort. Unfortunately, there is comparatively little work, other than function points, that tackles the problem of building prediction systems for software that is dominated by data considerations, in particular systems developed using 4GLs. We describe an empirical investigation of 70 such systems. Various easily obtainable counts were extracted from data models (e.g. number of entities) and from specifications (e.g. number of screens). Using simple regression analysis, a prediction system of implementation size with accuracy of MMRE=21% was constructed. This approach offers several advantages. First there tend to be fewer counting problems than with function points since the metrics we used were based upon simple counts. Second, the prediction systems were calibrated to specific local environments rather than being based upon industry weights. We believe this enhanced their accuracy. Our work shows that it is possible to develop simple and useful local prediction systems based upon metrics easily derived from functional specifications and data models, without recourse to overly complex metrics or analysis techniques. We conclude that this type of use of metrics can provide valuable support for the management and control of 4GL and database projects.

The ecological approach to interface design: applying the abstraction hierarchy to intentional domains

The ecological approach has traditionally been used for analysing process control type, physically coupled causal system domains to identify structural associations between the physical components of a system and the abstract functions of the system. Systems are then designed to represent the systems structural links as optical invariants in configural displays (J. Rasmussen, 1986; K.J. Vicente and J. Rasmussen, 1992; J.P. Hansen, 1995). Systems designed through this approach have been found to improve user diagnostic and response ability, e.g. W.S. Pawlak and K.J. Vicente (1996). Whilst it has been found to be useful in causal systems domains, it is not clear how applicable the ecological approach is to designing interfaces to support decision making in human activity based intentional systems domains. The paper is an exploration of the usefulness of the ecological approach in intentional domains.

IDENTIFIED (Integrated Dictionary-based Extraction of Non-language-dependent Token Information for Forensic Identification, Examination, and Discrimination): a dictionary-based system for extracting source code metrics for software forensics

The frequency and severity of computer-based attacks such as viruses and worms, logic bombs, trojan horses, computer fraud, and plagiarism of software code have all become of increasing concern to many of those involved with information systems. Part of the difficulty experienced in collecting evidence regarding the attack or theft in such situations has been the definition and collection of appropriate measurements to use in models of authorship, With this purpose in mind a system called IDENTIFIED is being developed to assist with the task of software forensics which is the use of software code authorship analysis for legal or official purposes. IDENTIFIED uses combinations of wildcards and special characters to define count-based metrics, allows for hierarchical metametric definitions, automates much of the file handling task, extracts metric values from source code, and assists with the analysis and modelling processes. It is hoped that the availability of such tools will encourage more detailed research into this area of ever-increasing importance.

Eliciting Information Portrayal Requirements: Experiences with the Critical Decision Method

This study is part of research that is investigating the notion that human performance in dynamic and intentional decision making environments, such as ambulance dispatch management, can be improved if information is portrayed in a manner that supports the decision strategies invoked to achieve the goal states of the process being controlled. Hence, in designing interfaces to support real-time dispatch management decisions, it is suggested that it would be necessary to first discover the goal states and the decision strategies invoked during the process, and then portray the required information in a manner that supports such a user group’s decision making goals and strategies.

The Effect of Layout on Dispatch Planning and Decision Making

This paper reports on an experiment conducted to determine whether the manner in which information is portrayed affects ambulance planning and dispatch decision making performance. Based the outcomes of a series of cognitive task analysis, deficiencies in an actual ambulance status display used for dispatch management was identified. The display was then redesigned by applying cognitive engineering principles to achieve task-to-display compatibility. The new display was then evaluated and it was found to improve dispatch decision making performance by 40%.

A goal-oriented approach for designing decision support displays in dynamic environments

This paper reports on how the critical decision method, a cognitive task analysis technique, was employed to identify the goal states of tasks performed by dispatchers in a dynamic environment, the Sydney Ambulance Coordination Centre. The analysis identified five goal states: Notification; Situation awareness; Planning resource to task compatibility; Speedy response; Maintain history of developments. These goals were then used to guide the development of display concepts that support decision strategies invoked by dispatchers in this task environment.

Deriving semantic data models from structured process descriptions of reality

This paper begins by describing a generally accepted view of the contemporary information systems development process. This process assumes the direct derivation of a prototype from data models which have themselves been derived from a functionally decomposed representation of reality. The data models are constructed using entity-relationship modelling (ERM) and data flow diagram (DFD) techniques. The paper also considers the use of Petri net representations as structured process models of reality. These representations are a means of structuring the functional analysis of observed situations for which an information system design is required. The paper then suggests a refinement to the model just described. It examines the relationship between the ERM and DFD techniques and proposes the use of logical access mapping (LAM) as a method of synthesizing the product of these two data modelling activities. An example is used to illustrate the derivation of ERM and DFD from Petri nets. Also an example of the refined process model is given. ‘Weak links’ in this model are identified and further work towards establishing a method for formally proving data models is proposed.

Usenet newsgroups' profile analysis, utilising standard and non-standard statistical methods

The paper explores building profiles of newsgroups from a corpus of Usenet e-mail messages, employing some standard statistical techniques as well as fuzzy clustering methods. A large set of data from a number of newsgroups has been analysed to elicit some text attributes, such as number of words, length of sentences and other stylistic characteristics. Readability scores have also been obtained by using recognised assessment methods. These text attributes were used for building newsgroups’ profiles. Three newsgroups, each with a similar number of messages, were selected from the processed sample for the analysis of two types of one-dimensional profiles: one by length of texts and the second by readability scores. Those profiles are compared with corresponding profiles of the whole sample and also with those of a group of frequent participants in the newsgroups. Fuzzy clustering is used for creating two-dimensional profiles of the same groups. An attempt is made to identify the newsgroups by defining centres of data clusters. It is contended that this approach to newsgroups’ pro-file analysis could facilitate a better understanding of computer-mediated communication on the Usenet, which is a growing medium of informal business and personal correspondence.

Neural networks and ambiguity in natural language

An attempt is made to consider the salient issues facing researchers who are examining the problems surrounding the resolution of ambiguity in natural language where computational analysis is performed for interpretative (understanding) purposes. Ambiguity in natural language has been chosen because it continues to be an elusive problem for computational linguistic researchers. No attempt to conduct a comprehensive and exhaustive literature review has been made. Rather, a selection of recent published papers has provided the basis for the analysis now presented here.<<ETX>>

Information portrayal design: applying the proximity-compatibility principle

The paper briefly reports on the process and design rationale for re-designing the representation of an existing emergency ambulance dispatch management status screen. The re-engineered screen was then subject to experimentation under four different categories of dispatch task difficulty. The results indicated that under the more difficult conditions, dispatch performance on the new screen improved by 40% over the performance observed by users of the old screen.