Text and Program Analysis (TAPAS)
The goal of the project was to develop methods and tools for
automated free text and program text analysis. The text analysis methods were designed from the following interrelated aspects:
- Novel assessment methods were developed for the content and
structural analysis of texts. The critical factors included time efficiency as well as accuracy, i.e., the correlation between the grades given by a human assessor and the grading method.
- Semi-automated assessment features were needed in order to give formative feedback to teachers and students. These features will open an essay in a "white-box", relating the evaluation and grading to the criteria ofassessment. Semi-automatic features such as automated feedback and summary generation can therefore support the learner during the writing process,
not only after it.
- Versatile measures and representations thereof were developed for the quality of a text. Besides the conventional numeric grade for the contents of an essay, the writing style could be evaluated in terms of structure, coherence, and cohesion. Argumentation, conceptual weaknesses, or sudden changes in vocabulary could be represented visually.
To implement a smart environment for automated assessment, three dimensions of an essay were analyzed: the learning dimension to indicate the cognitive level of the learning process, the linguistic dimension characterizing the structure of the essay, and the technical dimension to address available or needed techniques for assessing the essay throughout the writing process. One reason behind the limitations of essay assessment
techniques is the lack of a sufficient model which would cover the three aspects, which co-exist in any essay grading situation, whether manual or automatic. Thus, the three dimensions form a basis for an architecture of a formative semi-automatic essay evaluation and assessment environment.
Academy of Finland, 171 200 euros, 2006-2009.
Tuomo Kakkonen, email@example.com