In this paper we combined keyword search and content analysis to investigate the employment demands of ERD vs. UML in the context of data modeling. We found that data modeling is one of the popular skills required by many of the general business information systems professionals without specific methodology attached to the employment requirements. ERD and UML are required in narrower and often separate fields of building data models for information systems, in which ERD is mentioned in the domain of system analysis and design and database administration, while UML is required mostly in software engineering and application development. The division between ERD and UML in professional data modeling remains, although there is a small number of employment requirements that demand both methodologies and require a mapping capability between the two approaches.

Information and communication technology is one of the fastest changing fields in employment skills. This has resulted in constant revising of the academic curricula and textbooks to best match the education objectives and demands in professional skills (Gorgone, et al. 2002). In IT education, the various correlations of the recent IT employment opportunities and college students entering this field further stress the needs for examining the employment skills. The available employment opportunities published on the Internet have provided not only a job search tool for many job seekers but also a systematic monitoring of changes in employment skill demands in real-time (SkillPROOF Inc. 2004).

In this study, we analyzed IT employment data published by employers and focused on the ERD(Entity Relationship Diagram) vs. UML(Unified Modeling Language) requirements in the field of data modeling. We hope that our findings can supplement education decisions on what we should include in our teaching scopes and identify trends in the required professional skills

2. EMPLOYMENT DEMANDS

The employment data used in this study is published by individual companies on the Internet and collected daily by SkillPROOF Inc. since the beginning of 2004. The data is collected from up to 137 IT-focused companies. Each data sample contains attributes of company industry, posting date, job title, job responsibility, skill and education or training requirements. The general and background information of the data collection and categorization can be found on the website ofSkiIlProof.com (SkillPROOF Inc. 2004).

From the archived data, there are total of 35,932 jobs. The job counts from the top 12 industries (among a total of 46 industries) are plotted in Figure 1 to illustrate the overall distribution of the employment demands. The distribution reflects the post dot.com and post Sept/11 IT employment demands.

3. DATA ANALYSIS

We first applied keyword search to the job description to categorize the relevant ERD vs. UML skill requirements according to industry and then according to job types or job functions. We further sampled the contents of job description to investigate the implications for the job requirements in ERD vs. UML.

3.1 Keyword Categorization According to 'Industry'

We used the keywords of 'data model' or 'data modeling', 'ERD' and a few commonly referenced design tools like ErWin(2006), Visio(2006) and a database tool 'TOAD'(2006) to search for data modeling and database analysis, design and management related employment requirements. Similarly, we used the keyword 'UML' to search through the same data sets. We classified the search results according to the top 12 industries identified in Figure 1 and plotted the job counts in Figure 2(a), (b) and (c) below for comparison. One extra industry 'pharmaceutical' is added because the job counts in that industry are within the range of interest in the new aggregation.

Both ERD and UML keyword searched job counts are notably reduced to about half compared with data modeling searched job counts. Overall, the distribution for data modeling and ERD are similar in that both job spectra are broadly distributed across the industry line. They both share three common demanding industries: defense, high-tech and IT consulting. One exception is that the retailing industry appears to be very significant in requiring data modeling skills whereas the telecommunication is an industry requiring significant employment of ERD skills.

For the skill demand in UML, the outcome is significantly different. The defense industry is very noticeable in its relatively large job number in requiring UML skill. In terms of the job counts, the demands for UML are similar to the demands in ERD in both high-tech and IT consulting. However, they are both less than 25% of those counted in the defense industry.

3.2 Keyword Categorization According to 'Job Type' Another classification use keyword search is to sort all the job requirements according to the job function or job title. Use a standard job classification, we were able to use keywords of 'data modeling', 'ERD' or 'UML' to plot the searched skills vs. the types of jobs defined. The search results are summarized in Figure 3(a), (b) and (c) respectively for 'data modeling', 'ERD' and 'UML'. In this classification, we find that, again, UML has a lone popular job type as 'software development' whereas 'data modeling' or 'ERD' skill demands are more evenly distributed among the various job types, such as business analyst, software developer, IT consultant, technical writer and database administrator. One exception is that the job type of 'project manager' shows a significant number in ERD searched results but is not as popular in the 'data modeling' group.