Tracking temporal evolution in word meaning with distributed word representations

Abstract

Some words change meaning over time and are thus used differently in text. The purpose of this thesis is to create a model able to find these changes in word meaning, by studying lots of data from different time periods. Building on recent advancements in machine learning and semantic modelling the model is successfully able to find and make sense of changes in word meaning over time. The model can automatically find the most changed words during a time span and these words tend to agree with our perception of the words that have changed the most. When measuring changes the model achieves a 0.6 correlation when compared to human raters.