Skip to main content

A Python package to perform unsupervised sentiment analysis using lexicon enhanced Document embeddings

Project description

Lex2Sent - A bagging approach to unsupervised Sentiment Analysis

Lex2Sent is a text classification/clustering model that can be used with minimal a-priori-information to classify texts into two classes. While the original paper used it for sentiment analysis on english documents, it is not limited to that purpose, but can be used for any arbitrary type of classification and language as long as there are lexica that can be used as an information-basis.

Getting Started

You may install this package using either pypi

pip install lex2sent

or GitHub

pip install git+

The following is an example of using the Opinion Lexicon to classify an iMDb movie review data set. You may have to use to download the opinion_lexicon first. First we configure our data set

from datasets import load_dataset
from nltk.corpus import opinion_lexicon
data = load_dataset('imdb')
ratings, reviews = [], []
for stars, text in zip(data["train"]["label"], data["train"]["text"]):
    if text:
        if stars == 0:

And now we can start applying Lex2Sent

from lex2sent.textClass import *
lexicon = ClusterLexicon([opinion_lexicon.positive(), opinion_lexicon.negative()])
rated_texts = RatedTexts(reviews, lexicon, ratings)

#Basic "counting" method of classification:
count_res = rated_texts.lexicon_classification_eval(label_list=["positive", "negative"])
l2s_res = rated_texts.lbte(label_list=["positive", "negative"], workers=4)
print("Counting accuracy: {}%; Lex2Sent accuracy: {}%".format(count_res * 100, l2s_res*100))

yielding the result "Counting accuracy: 73.772%; Lex2Sent accuracy: 78.172%".


Please refer to "Lex2Sent - A bagging approach to unsupervised Sentiment Analysis" when using this package. When you use this package in a publication, please cite it as

  title = {{Lex2Sent}: {A} bagging approach to unsupervised sentiment analysis},
	shorttitle = {{Lex2Sent}},
	publisher = {arXiv},
	author = {Lange, Kai-Robin and Rieger, Jonas and Jentsch, Carsten},
	month = sep,
	year = {2022},
	note = {arXiv:2209.13023 [cs]},
	keywords = {Computer Science - Computation and Language},

Future Features

-Calling from the console

-FastText and SentenceBERT as alternatives to Doc2Vec

-Options to classify into more than two clusters

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

lex2sent-0.0.2.tar.gz (11.1 kB view hashes)

Uploaded Source

Built Distribution

lex2sent-0.0.2-py3-none-any.whl (12.2 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page