Machine Learning for Technical Information Quality Assessment

Abstract

This thesis is about assessing the quality of technical texts such as user manuals and product speci cations. This is done by consulting industry standards and guidelines, and implementing an automatic extractor for features describing the texts, based on these guidelines. These features are then put together into models, which are evaluated by using supervised machine learning algorithms on graded job application tests. Our conclusion is that it is probable that we can use this method and some of the features to classify the quality of technical texts. However, we think that it is hard to draw any con dent conclusions using this small data set and suggest as future work to evaluate this on a larger data set.