4: Workflow Recommendation Tool for your Research Project
If you are currently planning or working on a research project with handwritten sources or older retro-digitised prints, it is essential to think about the possible use of ATR.
This module has so far introduced you to the basic principles of Automated Text Recognition (ATR) and to different aspects of working with ATR such as: What constitutes a good transcription? What is the importance of the amount of text and the heterogeneity of hands (diversity of scripts)? In a later chapter you can also learn how to find or train a good text recognition model.
Our overall goal in this module is to teach you how to use ATR in a time-efficient manner. This will depend heavily on the structure, time and objectives of your project. For this reason, we have established four critical dimensions which consider the nature of your corpus (heterogeneity of hands, amount of text) and how this relates to the selected approach (research question, method). Please refer back to these dimensions and their explanation to think about how your own project places on the established dimension scales.
In this section, you will find an interactive tool to help you with deciding how to approach working with ATR once you know approximately where your project is placed on the previously established scales. Our tips will give you some important key points to consider before starting your research project. If you have already started, they might help to tackle some problems that have arisen. For explanations on how to select and train ATR-models and its related metrics, please refer to the last chapter.
Please select where your project lays on the four scales. For this tool, we have used the binary, considering only the two poles on each scale. If you assume that your project lies in between, read the advice for both poles and try to interpolate in between.