liwca 0.1.0a4

liwca (LOO-kə), or Linguistic Inquiry Word Count Assistant, provides helpers for working with LIWC dictionaries and related text analyses in Python.

• Read, write, and merge dictionary files (.dic, dicx)
• Generate category and word frequencies
• Generate Distributed Dictionary Representation (DDR) scores
• Run the LIWC app from Python (requires LIWC-22-cli access)
• Download publicly accessible dictionaries, corpora, and other relevant datasets

See the online docs for more detail.

Installation

pip install --upgrade liwca

Usage

Input/output

Read and write local LIWC-style dictionary files as Pandas DataFrames with schema validation.

import liwca

# Read a local dic file
dx = liwca.read_dic("my.dic")

# Read a local dicx file
dx = liwca.read_dicx("my.dicx")

# Read a local weighted dicx file
dx = liwca.read_dicx_weighted("myweighted.dicx")

# Create a dictionary from a collection of words
dx = liwca.create_dx(
    {
        "fruit": ["apple*", "pear*"],
        "vegetables": ["broccoli", "carrot*"],
    }
)

# Write the dictionary to a new file of any type
liwca.write_dic(dx, "fruit.dic")
liwca.write_dicx(dx, "fruit.dicx")
liwca.write_dicx_weighted(dx, "fruitweighted.dicx")

Word counting

Get category-level frequency scores for each text in a corpus. This will provide similar results to liwca.Liwc22().wc(), but does not require the LIWC-22 app and has the optional benefit of returning word-level frequencies. This is possible because the CountVectorizer from scikit-learn is used to calculate frequencies instead of the LIWC-22 app.

import liwca

# Generate some sample data
food_texts = [
    "My friends likes apples.",
    "She prefers pears over carrots.",
    "What I would do for a carrot right now!",
    "I am excited for pancakes tomorrow morning.",
]
food_categories = {
    "fruit": ["apple*", "pear*"],
    "vegetables": ["broccoli", "carrot*"],
}
food_dx = liwca.create_dx(food_categories)

# Return results at the category and word levels
cat_scores, word_scores = liwca.count(food_texts, food_dx, return_words=True)

cat_scores
# Category  WC  fruit  vegetables
# text_id
# 0          4   0.25    0.000000
# 1          5   0.20    0.200000
# 2          9   0.00    0.111111
# 3          7   0.00    0.000000

word_scores
#          WC  apples  broccoli    carrot  carrots  pears
# text_id
# 0         4    0.25       0.0       0.0      0.0    0.0
# 1         5     0.0       0.0       0.0      0.2    0.2
# 2         9     0.0       0.0  0.111111      0.0    0.0
# 3         7     0.0       0.0       0.0      0.0    0.0

LIWC-22 CLI wrapper

The primary benefits of the LIWC-22 CLI wrapper are:

• Opens and closes the LIWC-22 app automatically and keeps it in the background (GUI does not open)
• Pure Python implementation (e.g., cleaner scripts if pre/post operations are also in Python)
• Cleaner argument names (e.g., text_columns instead of id_columns)
• Cleaner argument types (e.g., bool True/False instead of yes/no strings)
• Improved formatting for results output file (e.g., keep original row ID column name instead of renaming to id)

import liwca
import pandas as pd

texts = pd.Series(food_texts)
outpath = "./test_results.csv"

with liwca.Liwc22() as liwc:
    liwc.wc(texts, outpath, include_categories=["Tone", "ppron", "food", "death"])

cat_scores = pd.read_csv(outpath, index_col=0)
#          Tone  ppron   food  death
# Row ID
# 1       99.00  25.00   0.00      0
# 2       20.23  20.00   0.00      0
# 3       20.23  11.11   0.00      0
# 4       99.00  14.29  14.29      0

cat_scores.dtypes
# Tone     float64
# ppron    float64
# food     float64
# death      int64
# dtype: object

Fetch public datasets

The liwca.datasets subpackage has multiple modules for fetching publicly available resources. Fetching includes downloading the file (if not already downloaded), caching it in local storage (to prevent re-downloading next time), preprocessing the file to fit standard formats, and also caching the preprocessed version.

from liwca.datasets import corpora, dictionaries

texts = corpora.fetch_hippocorpus()
texts.columns
# Index(['AssignmentId', 'WorkTimeInSeconds', 'WorkerId', 'annotatorAge',
#        'annotatorGender', 'annotatorRace', 'distracted', 'draining',
#        'frequency', 'importance', 'logTimeSinceEvent', 'mainEvent', 'memType',
#        'mostSurprising', 'openness', 'recAgnPairId', 'recImgPairId',
#        'similarity', 'similarityReason', 'story', 'stressful', 'summary',
#        'timeSinceEvent'],
#       dtype='str')

texts["story"].head()
# 0    Concerts are my most favorite thing, and my bo...
# 1    The day started perfectly, with a great drive ...
# 2    It seems just like yesterday but today makes f...
# 3    Five months ago, my niece and nephew were born...
# 4    About a month ago I went to burning man. I was...
# Name: story, dtype: str

dx = dictionaries.fetch_threat()
# Category     threat
# DicTerm
# accidents         1
# accusations       1
# advised           1
# afraid            1
# aftermath         1
# ...             ...
# woes              1
# worries           1
# worry             1
# worse             1
# worst             1

Similar projects

• liwc-python
• lingmatch
• pyliwc
• qdap / qdapDictionaries
• sentibank
• sentidict
• Shifterator