This is a corpus of manually annotated tweets for the emotion detection task (using Plutchik’s eight basic emotions) produced as a result of my master thesis and published as a paper.
Corpus consists of {tweet_id: label} pairs.
Tweet_id is an original id of the tweet from Twitter. We can provide only id-s due to Twitter’s policy, however the way how to get the texts of the tweets will be explained below.
Each label represents the emotions in a tweet in form of 8-dimensional vector for 8 emotion types. Each value in a vector is a float number between 0 and 3 representing the presence and intensity of each emotion, where 0 is «absent» and 3 is a «maximum intensity». The emotions in the vector are located in the following order: anger, disgust, fear, joy, sadness, surprise, anticipation, trust.
The labeled_tweets.txt file contains the corpus in human-readable form.
The labeled_tweets_dump.txt file contains the corpus in computer readable form. In order to access it, run the following python code:
import pickle
labeled_tweets = pickle.load(open('labeled_tweets_dump.txt', 'rb'))
for key in labeled_tweets:
tweet_id = key
tweet_label = labeled_tweets[key]
print(tweet_id, tweet_label)
To get the text of the tweet using its id:
- install tweepy, create an app and get your credentials
- use following code:
#set up tweepy
import tweepy
CONSUMER_KEY = #your credentials
CONSUMER_SECRET = #your credentials
ACCESS_TOKEN = #your credentials
ACCESS_TOKEN_SECRET = #your credentials
auth = tweepy.OAuthHandler(CONSUMER_KEY, CONSUMER_SECRET)
auth.set_access_token(ACCESS_TOKEN, ACCESS_TOKEN_SECRET)
api = tweepy.API(auth)
#get the texts
for key in labeled_tweets:
try:
tweet_id = key
tweet = api.get_status(tweet_id)
print(tweet.text)
except:
print("this tweet does not exist anymore")
Please, feel free to use the following BibTex citation if you shall use the corpus for publication purposes:
@article{rakhmetullina2018distant,
title={Distant Supervision for Emotion Classification Task using emoji2emotion},
author={Rakhmetullina, Aisulu and Trautmann, Dietrich and Groh, Georg},
url={CEUR-WS.org/Vol-2130/paper6.pdf},
year={2018}
}