prepare_emolex.py

## Prepare EmoLex
from os import system
from os.path import isfile
import pandas as pd

if not isfile("NRC - Sentiment Lexicon - Research EULA Sept 2017 .pdf"):
    system("wget http://sentiment.nrc.ca/lexicons-for-research/NRC-Emotion-Lexicon.zip")
    system("unzip NRC-Emotion-Lexicon.zip")
    system("rm NRC-Emotion-Lexicon.zip")

original_df = pd.read_excel(
    "NRC-Emotion-Lexicon-v0.92/NRC-Emotion-Lexicon-v0.92-In105Languages-Nov2017Translations.xlsx"
)  # Open the downloaded Excel file

# For English:
if not isfile("english_emolex.csv"):
    df = original_df[
        [
            "English (en)",
            "Anger",
            "Anticipation",
            "Disgust",
            "Fear",
            "Joy",
            "Sadness",
            "Surprise",
            "Trust",
        ]
    ]  # Keep only the English tokens and the annotated emotions
    df = df[
        df[
            [
                "Anger",
                "Anticipation",
                "Disgust",
                "Fear",
                "Joy",
                "Sadness",
                "Surprise",
                "Trust",
            ]
        ].sum(axis=1)
        > 0
    ]  # Drop words that does not relate to any emotion
    df.to_csv("english_emolex.csv", index=False)  # Save to file

# For Mandarin Chinese:
if not isfile("chinese_emolex.csv"):
    df = original_df[
        [
            "Chinese (Simplified) (zh-CN)",
            "Anger",
            "Anticipation",
            "Disgust",
            "Fear",
            "Joy",
            "Sadness",
            "Surprise",
            "Trust",
        ]
    ]  # Keep only the Chinese tokens and the annotated emotions
    df.drop_duplicates(
        subset=["Chinese (Simplified) (zh-CN)"], inplace=True
    )  # translation inbalances
    df = df[
        df[
            [
                "Anger",
                "Anticipation",
                "Disgust",
                "Fear",
                "Joy",
                "Sadness",
                "Surprise",
                "Trust",
            ]
        ].sum(axis=1)
        > 0
    ]  # Drop words that does not relate to any emotion
    df.to_csv("chinese_emolex.csv", index=False)  # Save to file

if isfile("NRC - Sentiment Lexicon - Research EULA Sept 2017 .pdf"):
    system(
        'rm -r NRC-Emotion-Lexicon-v0.92/ "NRC - Sentiment Lexicon - Research EULA Sept 2017 .pdf"'
    )  # Remove unzipped files