Yangting Wang Undergraduate Dissertation 2015/16

Is a readable text a good quality text?

Supervised by L.Specia

Abstract

Natural Language Processing is a field of Artificial Intelligence, which focuses on generation and analysis of human language. Machine translation is one application of the Natural Language Processing, where translation quality and text readability are two criteria which are used to evaluate the text.

This study will focus on investigating different metrics of translation quality and text readability. The aim is to provide more reliable machine translation evaluation techniques. In order to use those metrics, a research of different kinds of evaluation metrics will be taken, as well as a conducted study on datasets and toolkits.

Furthermore, the result will be presented in the report, three datasets will be used for examination. Next, the experiment will be compared between two measures of evaluation metrics with Human evaluation scores, also in between themselves. In the end, the results are obtained by using several correlation techniques such as Pearson and Spearman correlation.