LabelThis Crowd Sourcing

The Label This application divided the tasks needed for the transcription into a few simple steps. Our volunteers were free to perform as many of these steps as they liked. They could also choose which tasks they enjoyed doing most.

Mark label areas

Before contributors could transcribe the labels, we needed to know which sections needed transcription. Our patrons marked up the labels by drawing boxes around each area of interest on the label, and identified the regions as images or text. Task were performed multiple times by different individuals to improve the accuracy of the results.

Identifying Regions

Transcription and Description

Once all the areas of interest on a label were marked, users used a separate tool to transcribe the text on the label. As label text is often artfully executed, humans are often superior to machines in this task

Transcribing Text

Alternatively, users could describe any image that appeared on the labels. Again, humans can understand the context of an image on a label in ways that machines cannot.

Describing Images

The Application

LabelThis was primarily developed with the open software Scribe API Administration of the transcriptions was managed in an internal database.