Skip to content

7compass/sentimental

Repository files navigation

sentimental

Simple sentiment analysis with Ruby

How it works

Sentences are tokenized and tokens are assigned a numerical score for their average sentiment. The total score is then used to determine the overall sentiment in relation to the threshold.

For example, the default threshold is 0.0. If a sentence has a score of 0, it is deemed "neutral". Higher than the thresold is "positive", lower is "negative".

If you set the threshold to a non-zero amount, e.g. 0.25:

  • Positive scores are > 0.25
  • Neutral scores are -0.25 - 0.25
  • Negative scores are < -0.25

Usage

# Create an instance for usage
analyzer = Sentimental.new

# Load the default sentiment dictionaries
analyzer.load_defaults

# And/or load your own dictionaries
analyzer.load_senti_file('path/to/your/file.txt')

# Set a global threshold
analyzer.threshold = 0.1

# Use your analyzer
analyzer.sentiment 'I love ruby'
#=> :positive

analyzer.sentiment 'I like ruby'
#=> :neutral

analyzer.sentiment 'I really like ruby'
#=> :positive

# You can make new analyzers with individual thresholds:
analyzer = Sentimental.new(threshold: 0.9)
analyzer.sentiment 'I love ruby'
#=> :positive

analyzer.sentiment 'I like ruby'
#=> :neutral

analyzer.sentiment 'I really like ruby'
#=> :neutral

# Get the numerical score of a string:
analyzer.score 'I love ruby'
#=> 0.925

Sentiment dictionaries

These are currently plain-text files containing whitespace-separated scores and tokens, e.g.:

1.0 Awesome
0.0 Meh
-1.0 Horrible

N-grams

You can parse n-grams of words by specifying their max size in the initializer:

  Sentimental.new(ngrams: 4)

The dictionary must have this format:

1.0 very happy
-2.0 no
0.0 meh

Installation

gem install sentimental

License

MIT License

Credits

Based largely on Christopher MacLellan's script: https://github.com/cmaclell/Basic-Tweet-Sentiment-Analyzer

Changes

  • 2013-10-13 Adding :-) to slang

About

Simple sentiment analysis with Ruby

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages