Prediction Tool Progress in Interactive Machine Translation Systems - A study on the evaluation and improvement of the auto completion tool in Computer Aided Translation Tools

Tsoukala, C. (2015) Prediction Tool Progress in Interactive Machine Translation Systems - A study on the evaluation and improvement of the auto completion tool in Computer Aided Translation Tools. Master's Thesis / Essay, Artificial Intelligence.

Preview	Text Msc_thesis_Tsoukala_2015.pdf - Published Version Download (4MB) | Preview
	Text toestemming.pdf - Other Restricted to Backend only Download (22kB)

Abstract

Machine Translation systems are still far from generating error-free translations, and the output requires human post-editing(PE) in order to achieve high-quality translations. This study focuses on Interactive Translation Prediction(ITP), a tool that comes to assist, rather than replace, human translators by attempting to autocomplete the text they are going to insert. Completion suggestions are obtained by matching user input to the search graph, which contains all translations of the source. Detection of parts already translated is not trivial as the mechanism needs to work at user-typing speed. Typically, the best match is the path with the smallest edit distance to the user input, and the highest score. Our first goal is to extract more features that can increase accuracy. Using Machine Learning, we explore features such as whether the last token of the user input was matched to the last token of the matched string(lastMatched), levensthein distance between last token, and others. Of these, lastMatched resulted in higher prediction accuracy by 0,5%. To test the usability of the prediction tool, we run a user. Participants were asked to translate newspaper corpus using ITP and PE. The hypothesis was that users would be in favour of ITP, as it can lead to less typing. Survey results show that editors were in favour of ITP, but logs do not show a speed increase; in some cases PE is faster. Nevertheless, given that user satisfaction is high, it is worth further investigating an increase in accuracy, along with the optimal visualization for ITP

Item Type:	Thesis (Master's Thesis / Essay)
Degree programme:	Artificial Intelligence
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	15 Feb 2018 08:03
Last Modified:	15 Feb 2018 08:03
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/12582

Actions (login required)

View Item