PhaseTransitionsInWordEmbeddings

Bachelor project by Thor Larsen at Institute of Computer Science @ UCPH, supervised by Daniel Hershcovich

Data and related scripts used in paper's results section can be found in the "data" folder. Our hypothesis is tested using a hyper-geometric significance test using SciPy.

Final models are created using Gensim's implementation of the original Word2Vec model by Mikolov et.al. and evaluated using Spearman's $\rho$. The scripts made to generate models and the model files are found in the folder "models". Models made with AWS SageMaker's BlazingText and their evaluation scores can also be found under "models", but cannot be generated without access to AWS SageMaker.

Final version of paper is also included.

This GitHub is only for reference and storage. Comments and requests are not reviewed.