1 / 32

Semantic Structuring in Analyst Acquisition and Representation of Facts in Requirements Analysis

Semantic Structuring in Analyst Acquisition and Representation of Facts in Requirements Analysis. George Marakas and Joyce Elam ISR 1998, vol 9., no.1 Presented for EEL 5881 Virginia Ilie November 5, 2002. Overview. Importance of the RA phase Purpose of the Study Research Questions

verena
Télécharger la présentation

Semantic Structuring in Analyst Acquisition and Representation of Facts in Requirements Analysis

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Semantic Structuring in Analyst Acquisition and Representation of Facts in Requirements Analysis • George Marakas and Joyce Elam • ISR 1998, vol 9., no.1 • Presented for EEL 5881 • Virginia Ilie • November 5, 2002

  2. Overview • Importance of the RA phase • Purpose of the Study • Research Questions • Information Gathering in Requirements Determination • Hypotheses • Methodology • Analysis and Results • Implications • Potential Limitations • Future Directions for Research

  3. “One of the greatest problems the system designer must face is the need to completely define the requirements of the system under development” • (Zmud et al 1993)

  4. Importance of the RA phase • Most IS failures traced back to a lack of specific and correct information requirements (Davis 1982, Kydd 1989) • Many IS successes traced to the proper identification of information requirements early in the design process (Boehm 1976, 1981)

  5. Importance of the RA phase • Over 50% of the systems reviewed had problems necessitating return to the RA phase (Jenkins et al 1984) • Relative cost of correcting an error increases exponentially with the project phase error is detected (Boehm 1973)

  6. Purpose of the Study • Investigate the effects of a semantic structuring process of inquiry (SSPI) on interview-derived information gathering and the subsequent overall correctness of a derived logical model • Specifically, investigate particular question types used by system analysts and the role their semantic construction played in understanding the information flows in a business system

  7. Definitions • Semantic Structuring - the application and use of specific question types for obtaining specific interview-derived information

  8. Research Questions • What specific question types are most effective in gathering direct interview-derived information related to the construction of an accurate model representing the logical data flows within an organizational or business subsystem ? • Can an inquiry process composed of specific question types lead to obtaining greater accuracy in the collection of relevant direct interview-derived information related to business subsystem processes and data flows ?

  9. Premise • The use of a SSPI: • For the analyst: enhance his/her ability to elicit user requirements • For the user: better explicate his/her requirements • For both: more effectively and accurately communicate with each other

  10. Information Gathering in Requirements Determination • Direct interview of the user remains one of the primary means of obtaining information (Holtzblatt&Beyer 1995)

  11. Unstructured vs. Structured Interviews • Without some type of format or structure, the value and accuracy of information is greatly diminished (Dorsey & Koletzke 1997) • Unstructured communications manifested in dysfunctional characteristics and costly errors in design (Salaway 1987)

  12. Unstructured vs. Structured Interviews • Structured interviews: method to improve the quality of information gathering during the questioning • Structured interviews found to be more efficient than unstructured interviews (Agarwal & Tanniru 1990)

  13. The Psychology of Questioning • Loftus & Palmer’s experiment (1974): The film of a traffic accident • The wording of the questions and the modification of a single word within the body of a question altered significantly respondent’s recollection of an event • Very little is known about the role of questionning processes as they relate to information requirements (Lauer et al 1992)

  14. Hypotheses • H1: The use of a semantic structuring model of inquiry •  greater accuracy in the construction of logical representationsover the use of unstructured inquiry

  15. Hypotheses H1a. The use of a semantic structuring model of inquiry greater accuracy in the construction of logical representations by low experience analysts (over the use of unstructured inquiry) H1b. The use of a semantic structuring model of inquiry greater accuracy in the construction of logical representations for high-experience analysts (over the use of unstructured inquiry) H1c. The relative degree of improvement resulting from the use of the semantic structuring model of inquiry in the correctness score of a logical representation will be greater for low experience analysts than for high experience analysts.

  16. Hypotheses • H2a. The number of feasible but nonexistent elements mapped in the construction of a logical representation will be significantly greater for those using no specific inquiry strategy rather than for those making use of the semantic structuring model of inquiry • H2b. The number of feasible but nonexistent elements mapped in the construction of a logical representation will be significantly greater for high experience subjects than for low experience subjects regardless of the inquiry method employed.

  17. Hypotheses • H3. The presence of a question inquiring about a specific data element or flow will be significantly related to the probability of identifying it correctly regardless of the inquiry method employed • H4. Subjects using no specific inquiry strategy that are successful in constructing an accurate logical representation will exhibit semantic questioning patterns similar to other successful subjects • H5. Subjects using no specific inquiry strategy that are successful in constructing an accurate logical representation will exhibit questioning patterns similar to the SSPI

  18. Methodology • Lab experiment • Two control groups (low experience and high experience analysts using an unstructured elicitation approach) - CL and CH • Two experimental groups (low experience and high experience analysts using a semantic model of inquiry) - EL and EH • Low experience subjects recruited from an undergrad MIS program at a southern univ. • The high experience subjects recruited from professional system analysts and software designers from several major corporations

  19. Methodology • Pretest questionnaire • Subjects randomly assigned to one of the two subject groups based on their experience level • Subjects received a copy of an abridged description of an existing business system • Their task was to develop a thorough understanding of the system through interviewing a “key employee” of the organization • Following the interview, each subject was asked to construct a logical DFD from the information obtained

  20. Analysis and Results: H1 • H1: SUPPORTED • Method: ANOVA and post-hoc pairwise comparisons • Comparison of DFD scores from the groups using no specific inquiry model (CL and CH) and the ones using the semantic model of inquiry (EL and EH) •  The use of a specific semantic process model of inquiry has a significant effect on the accuracy of the logical representation (the DFD) over the use of no specific model of question construction

  21. Analysis and Results: H1a • H1a: SUPPORTED • Method: ANOVA • Comparison of the DFD scores obtained from the low experience control and treatment groups only (CL vs EL)  significant differences exists • Low experience analysts benefited from the use of the specific model of inquiry over use of no specific inquiry process

  22. Analysis and Results: H1b • H1b: SUPPORTED • Method: ANOVA • Comparison of DFD scores from the high experience analysts (CH vs EH) •  High experience analysts derived value through the application of the semantic process of inquiry

  23. Analysis and Results: H1c • H1c: NOT SUPPORTED • Method: t-test – ascertain the significance of the differences between groups CL and EL and CH and EH • No support regarding the degree of improvement associated with the use of the specific inquiry process over the unstructured process

  24. Analysis and Results: H2a • H2a: SUPPORTED • Method: t-test – compare the difference between the number of feasible but non-existent data flows mapped by the combined control groups and that of the combined treatment groups • Feasible data flow: it it could theoretically exist within the specified context but but in fact it does not • Any element deemed feasible but non-existent counted and a group mean calculated

  25. Analysis and Results: H2b • H2b: NOT SUPPORTED • Method: t-test – compare the difference between the high experience and low experience subjects regarding feasible but non-existent data flows • The difference is not significant

  26. Analysis and Results: H3 • H3: SUPPORTED • Method: chi-square test for difference in probabilities • The correct identification of a data element in the DFD must be associated with at least one question resulting in a response that either establishes or confirms its existence • The creation of an accurate logical representation of the business process is significantly associated with the inquiry itself (in addition to any experiential or inferential skills)

  27. Analysis and Results: H4 • H4: SUPPORTED • Method: K-means cluster analysis – It produces clusters that best represent the primary similarities among the cluster items and minimize variation within each cluster • 3 clusters emerged – Balanced Pattern, Micro-Inquiry and • Confirmatory Pattern • 7 of 9 high score subjects fell under Balanced Pattern • The majority of high-score control group subjects displayed a similar questioning pattern (even though no specific pattern was imposed

  28. Analysis and Results: H5 • H5: SUPPORTED • Method: K-means cluster analysis • 7 of the 9 control group high-score control subjects - Balanced approach • 13 of the 20 experimental treatment group – Balanced approach High score control group subjects classified into a Balanced approach exhibited a questioning pattern similar in structure to the semantic model

  29. Implications • The SSPI plays an important part in obtaining accurate interview-derived information during the RA phase • There is a definite relationship between certain semantic constructions (the type of question) and the accuracy of logical representations • Application of a SSPI may be independent of the experience level of the analysts • SSPI may serve as a source of reduction of errors associated with guessing or inference based on experience or logic • Application of the SSPI may reduce the number of technically feasible errors associated with the design of a DFD • Accurate representations of business processes may be closely related to the method of inquiry itself than with any given level of experience

  30. Potential Limitations • Lab experiments cannot capture certain realities existent in the real world • Student subjects • Using the DFD as a single method of logical system representation

  31. Future Directions for Research • Examine the applicability of the semantic structuring across various information gathering contexts (RAD, OOD) • Is the semantic structuring of an interview effective in the gathering of physical data elements for the construction of an ERD ? • Can semantic structuring be understood such that once the context of an information gathering activity is established, a model can be designed to maximize interviewer’s efforts ?

  32. Questions ?

More Related