Based on emotion, sentiment, subjectivity, orientation, and color.

• NRC Emotion Lexicon
• Bing Liu's Opinion Lexicon
• MPQA Subjectivity Lexicon
• Harvard General Inquirer
• NRC Word-Colour Association Lexicon

These lexicons were parsed and compiled using the script compile_lexicons.py into file lexicons/lexicons_compiled.csv using these categories with the following counts:

• Total word count: 14,852
• Words with emotion: 4,463 (30.0%)
• Words with sentiment: 10,916 (73.5%)
• Words with subjectivity: 6,886 (46.4%)
• Words with orientation: 2,192 (14.8%)
• Words with color: 5,404 (36.4%)

1. Download text, e.g. texts/moby_dick.txt

2. Parse text using gutenberg_text.py <text file> <output text json file> <output chapter json file>, e.g. gutenberg_text.py ../texts/moby_dick.txt ../output/moby_dick_normal.json ../output/moby_dick_chapters.json. This generates an output that looks like this:

   {
     "title": "Moby-Dick; or, The Whale'",
     "author": "Herman Melville",
     "chapters": [
       {
         "title": "Loomings",
         "text": "discrete words in lowercase separated by spaces with punctuation removed"
       },
       {
         "title": "The Carpet-Bag",
         "text": "discrete words in lowercase separated by spaces with punctuation removed"
       },
       ...
     ]
   }

3. Run get_data.py <json file from previous step> <a path to output csv file>, e.g. get_data.py output/moby_dick_normal.json output/moby_dick_data.csv. This outputs a .csv file in the format:

   emotion,color,orientation,sentiment,subjectivity,chapter
   0,2,1,1,-1,0
   ...
   

   Where each row represents a word, and each column represent the index of each category listed in data/categories.json

4. Run analyze_data.py <csv file from previous step> <a path to output csv file> <word buffer> <word offset>, e.g. python analyze_data.py output/moby_dick_data.csv output/moby_dick_analysis.json 400 200. This outputs a .json file in the format:

   [
    {
      "chapter": 0,
      "emotion": [
        0.500, // anger
        0.250, // fear
        ...
      ],
      "subjectivity": [
        0.600, // weak
        0.150 // strong
      ],
      "sentiment": [
        0.750, // positive
        0.050 // negative
      ],
      "orientation": [
        0.850, // active
        0.450 // passive
      ],
      "color": [
        0.950, // white
        0.001, // black
        ...
      ]
    },
    ...
   ]

   Where each item represents a group of words (with a size of word buffer as configured in the previous step). The numbers are percentages between 0 and 1 that represents the relative weight of that particular category value.

5. Optionally, run python report_data.py <analysis json file> <output dir> to write individual .csv files for each category, e.g. python report_data.py output/moby_dick_analysis.json output/moby_dick/