gargantext-0.0.4.9.5: Search, map, share
Copyright(c) CNRS 2017
LicenseAGPL + CECILL v3
Maintainerteam@gargantext.org
Stabilityexperimental
PortabilityPOSIX
Safe HaskellNone
LanguageHaskell2010

Gargantext.Core.Text.Terms.Multi.RAKE

Description

Personal notes for the integration of RAKE in Gargantext.

RAKE algorithm is a simple, rapid and effective algorithm to extract keywords that is very sensitive to the quality of the stop word list.

Indeed, the very first step starts from the stop words list to cut the text towards keywords extraction. The conTexT is the sentence level to compute the coccurrences and occurrences which are divided to compute the metric of one word. Multi-words metrics is equal to the sum of the metrics of each word.

Finally The metrics highlight longer keywords which highly depends of quality of the cut which depends on the quality of the stop word list.

As a consequence, to improve the effectiveness of RAKE algorithm, I am wondering if some bayesian features could be added to increase stop word list quality in time.

Synopsis

Documentation

select :: Double -> [a] -> [a] Source #