Transcription of the Bleek and Lloyd Collection using the Bossa Volunteer Thinking Framework

Munyaradzi, Ngoni (2013) Transcription of the Bleek and Lloyd Collection using the Bossa Volunteer Thinking Framework, MSc.

[img] PDF
Ngoni_Munyaradzi_Thesis_2013.pdf

Download (6MB)

Abstract

The digital Bleek and Lloyd Collection is a rare collection that contains artwork, notebooks and dictionaries of the earliest habitants of Southern Africa. Previous attempts have been made to recognize the complex text in the notebooks using machine learning techniques, but due to the complexity of the manuscripts the recognition accuracy was low. In this research, a crowdsourcing based method is proposed to transcribe the historical handwritten manuscripts, where volunteers transcribe the notebooks online. An online crowdsourcing transcription tool was developed and deployed. Experiments were conducted to determine the quality of transcriptions and accuracy of the volunteers compared with a gold standard. The results show that volunteers are able to produce reliable transcriptions of high quality. The inter-transcriber agreement is 80% for |Xam text and 95% for English text. When the |Xam text transcriptions produced by the volunteers are compared with the gold standard, the volunteers achieve an average accuracy of 69.69%. Findings show that there exists a positive linear correlation between the inter-transcriber agreement and the accuracy of transcriptions. The user survey revealed that volunteers found the transcription process enjoyable, though it was difficult. Results indicate that volunteer thinking can be used to crowdsource intellectually-intensive tasks in digital libraries like transcription of handwritten manuscripts. Volunteer thinking outperforms machine learning techniques at the task of transcribing notebooks from the Bleek and Lloyd Collection.

Item Type: Electronic thesis or dissertation (MSc)
Uncontrolled Keywords: crowdsourcing, transcription, cultural heritage
Subjects: Information systems > Information systems applications
Information systems > Data management systems
Information systems > Data management systems > Database design and models
Human-centered computing
Information systems > Information retrieval
Date Deposited: 04 Dec 2013
Last Modified: 10 Oct 2019 15:33
URI: http://pubs.cs.uct.ac.za/id/eprint/913

Actions (login required)

View Item View Item