【文章內(nèi)容簡(jiǎn)介】 of the process model while keeping its content intact. As the validity of a process model is and remains of the utmost importance, 7PMG plements this concern with remendations for how to model. In this way, 7PMG does not help with the pragmatic question of what should be put into a model. This question still has to be assessed based on the purpose of modeling. The second limitation of 7PMG relates to the prioritizing guideline that was described in Section 4. Clearly, the derived ranking has a small empirical ba sis as it relies on the involvement of 21 process modelers only. 這可能是 seen on the one hand as a need for a wider involvement of process modelers39。 experience, but it also raises the question what alternative approaches may be available to arrive at a prioritizing guideline, eg through experimentation. Also, the prioritizing guideline that can be inferred from Table 2 is too coarse to be applicable at the micro level of, for example, deciding between the ap plication of two conflicting guidelines. Still, it seems less attractive to focus at this stage on developing more finegrained guidance for applying 7PMG. 首先 of all, it does not seem realistic that guidance can be developed that is both conclusive and valid for all scenarios. Secondly, we expect the set of 7PMG to be extended as insights into process model quality develop over time. 對(duì)于前 ample, the workshops that we carried out with the professional modelers have already provided us with inspiration for the investigation of further guidelines. 20 第21頁 Potential We continue this section with discussing the potential of 7PMG beyond the ap plication that is hitherto sketched. If we suppose the availability of a function that quantifies the quality of a model (eg in terms of error probability) then this bears the potential of automating the improvement of a model. 幾個(gè) techniques for graph matching and graph edit distance calculation can poten tially be applied when such an available regression function is used as a goal 功能。 An approach like this would require further formal research on how behaviorpreserving sets of edit operations can be defined that are plete. Some work has been done in this area by identifying change operations [50]. Follow up on this type of research might eventually provide mechanisms to automatically enforce 7PMG in a modeling tool and offer intelligent support for 修改。 The seven guidelines also point to some potential for quantifying the pe tence of process modelers by measuring the structure of models they create. This might base on the assumption that good modelers will intuitively put requirements into a model in such a way that the guidelines are followed to a large extent. On the other hand, petent modelers are likely to under stand those models that deviate from the guidelines too. In an experiment that involved students from three European universities we have observed sig nificant differences in performance with respect to reading process models. Although process modeling was part of the curriculum of each of the three groups, the amount of time spent on this subject differed. Students that were trained longer in process modeling and analysis, notably with respect to de tecting deadlocks and recognizing block structures, were also better able to understand the process models that they were faced with [31]. 這樣一來， 7PMG might offer ways to identify the difficulty of a model understanding task, and accordingly a way to assess how petent a person is in understanding the more difficult models. 21 第22頁 Modeling Techniques and Modeling Tools There are additional matters of process modeling that are not directly ad dressed by the seven guidelines, but that are very closely related to it. 我們 single out the subjects of (1) the modeling technique , ie the notation or lan guage that is used to create a process model, and (2) the modeling tool , ie the software package that supports the use of a particular modeling notation. Over the last decades, many different process modeling techniques have been 建議。 Vendors and standardization bodies have a tendency to e up again and again with new diagramming techniques, yet hardly ever providing any empirical or theoretical validation. An approach that builds on empirical insights such as 7PMG would be desirable in this area, too. Some research has been conducted on paring different process modeling languages from an enduser perspective. In particular, languages with an explicit representation of routing elements seem to be easier to understand [35]. These languages also tend to have less elements: In an EPC an ANDsplit requires only one connector while in a Petri net one needs to model a transition with various output places. In this ways, guideline G1 and its supporting research [41,44] have some implications also on the language level: Those languages that re quire less elements representing the same fact than another language might be preferable for munication purposes. In fact, this is backed up by a more general insight that is claimed to relate to all forms of representations, ie that the ones that provide smaller models have the higher efficacy [51]. 在 same time, a notation will never free the designer from the difficult work of actually mapping the realworld process onto the model. The selection of a particular modeling technique determines to some extent what modeling tool to use, and vice versa. Some frameworks, like the cognitive dimension framework, stress that both should be studied jointly to understand how the user interacts with the notation via a tool [52]. An additional selection criterion, based on the guidelines we discussed, would be to favor modeling tools that further support the modeling process. For example, tools could warn 22 第23頁 users if their model bees too l