/Keyword-Extractor-with-Closeness-Centrality

Primary LanguagePython

Keyword-Extractor-with-Closeness-Centrality

Closeness Centrality is a measure of centrality of individual nodes within a network. This metric explains the role/importance of a node acting as the "middleman" allowing transfer of information between other nodes. Higher closeness centrality = more nodes have to go through this node to reach other nodes.

Experiment

Closeness centrality to be applied on vectors of words to see if keywords can be extracted. Brute force is used as this is a small experiment.

Results

Quite accurate but not very scalable!

Results from short story on Ali Baba and Forty Thieves

A quick peep on the top of keywords returned:

[   ('baba', 0.09611829944547136),
    ('ali', 0.09484106305367379),
    ('gold', 0.03565830721003135),
    ('wife', 0.020097173144876326),
    ('cassim', 0.017294626312538603),
    ('words', 0.016629357211384713),
    ('said', 0.01604726006260195),
    ('rich', 0.015273581738838537),
    ('see', 0.014914365320003276),
    ('door', 0.014876573483733857),
    ('sesame', 0.01447085950544645),
    ('rock', 0.013497478493028774),
    ('shut', 0.012645035781282569),

Improvements

Names should be tokenised together and most importantly, scalability should be improved.