Topic label groups that were generated using both chi-squared and mutual information methods are listed in bbc_topic.py.

In addition, two extra topic groups have been added for comparison. The two extra groups were made up and should return high topic coherence. These are a group of colours (pink,purple,blue,yellow,green,white,orange, brown,red) and fruits (apple,orange,pear,banana,plum,apricot,mango,grape,pineapple).

coherence_lch_wup.py uses NLTK implementations of the LCH and WUP algorithms, this should be run in Python3. You will need to download NLTK resources by running nltk.download()

coherence_lesk.py uses an implementation of LESK written in Python2.7, the code for this is included in the pyswd directory.