TESA
This is a short descritpion of how to use the T-ESA (Twitter-Easy Sentiment Analysis) package for conducting some sentiment analysis using a lexicon based approach.
Per Default the Hu and Liu lexicon is used (Hu & Liu, 2004; Liu et al., 2005) for the opinions. In addition, for handling emojis the lexicon package (in R) based on work of Novak et al. (2015) is utilized. The data can be accessed from this link: https://rdrr.io/cran/lexicon/man/emojis_sentiment.html (see */TESA/lexicon/emojis/ for more details how to data was preprocessed with a R-Script for usage in TESA).
However, users can also provide other lexicons as text-files (one for positive words and one for negative ones) as well as another emoji lexicons (make sure to keep the column names and file structure).
Please see the provided example for more details.
The package works on single tweets that are expected to be provided as string variable. It is capable of:
-
cleaning the tweet(stop word and special character removal using a given list) -
using the
Porter-Stemmeror a lemmanization routine (both taken from the nltk package) to ensure better matching betweens the words in the tweets and the lexicon -
simple
negation handling -
interpretation of
emojisand their sentiments using the work of Novak et al. (2015) and the emoji Python libary
How to install it:
run this statement on your command line:
pip install -i https://test.pypi.org/simple/ TESA
or if you want to build from source:
-
download the project
-
navigate to the folder containing the setup.py
-
run pip3 install . (or pip3 install -e . to give access to all users of your local machine)
-
in case you do not have administration rights use the --user specification for pip
How to use it:
Some example Python code and data can be found in the /Example directory of the Python package that shows the usage of the API.
As almost mentioned you will have to install the nltk and emoji Python package as well.
References
Minqing Hu and Bing Liu. "Mining and Summarizing Customer Reviews." Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD-2004), Aug 22-25, 2004, Seattle, Washington, USA
Bing Liu, Minqing Hu and Junsheng Cheng. "Opinion Observer: Analyzing and Comparing Opinions on the Web." Proceedings of the 14th International World Wide Web conference (WWW-2005), May 10-14, 2005, Chiba, Japan.
Novak, P. K., Smailovic, J., Sluban, B., and Mozetic, I. "Sentiment of emojis." PLoS ONE 10(12). 2015, doi:10.1371/journal.pone.0144296