Lingualytics : Easy codemixed analytics

Lingualytics is a Python library for dealing with code mixed text.
Lingualytics is powered by powerful libraries like Pytorch, Transformers, Texthero, NLTK and Scikit-learn.

🌟 Features

Preprocessing
- Remove stopwords
- Remove punctuations, with an option to add punctuations of your own language
- Remove words less than a character limit
Representation
- Find n-grams from given text
NLP
- Classification using PyTorch
  - Train a classifier on your data to perform tasks like Sentiment Analysis
  - Evaluate the classifier with metrics like accuracy, f1 score, precision and recall
  - Use the trained tokenizer to tokenize text

🧠 Pretrained Models

Checkout some codemix friendly models that we have trained using Lingualytics

💾 Installation

Use the package manager pip to install lingualytics.

pip install lingualytics

🕹️ Usage

Preprocessing

from lingualytics.preprocessing import remove_lessthan, remove_punctuation, remove_stopwords
from lingualytics.stopwords import hi_stopwords,en_stopwords
from texthero.preprocessing import remove_digits
import pandas as pd
df = pd.read_csv(
   "https://github.com/lingualytics/py-lingualytics/raw/master/datasets/SAIL_2017/Processed_Data/Devanagari/validation.txt", header=None, sep='\t', names=['text','label']
)
# pd.set_option('display.max_colwidth', None)
df['clean_text'] = df['text'].pipe(remove_digits) \
                                    .pipe(remove_punctuation) \
                                    .pipe(remove_lessthan,length=3) \
                                    .pipe(remove_stopwords,stopwords=en_stopwords.union(hi_stopwords))
print(df)

Classification

Currently available datasets are

CS-EN-ES-CORPUS Vilares, D., et al.
SAIL-2017 Dipankar Das., et al.
Sub-Word-LSTM Joshi, Aditya, et al.

from lingualytics.learner import Learner

learner = Learner(model_type = 'bert',
                model_name = 'bert-base-multilingual-cased',
                dataset = 'SAIL_2017')
learner.fit()

Custom Dataset

The train data path should have 3 files

train.txt
validation.txt
test.txt

Any file should have the text and label in a line, separated by a tab. Then change the data_dir to the path of your custom dataset.

Find topmost n-grams

from lingualytics.representation import get_ngrams
import pandas as pd
df = pd.read_csv(
   "https://github.com/jbesomi/texthero/raw/master/dataset/bbcsport.csv"
)

ngrams = get_ngrams(df['text'],n=2)

print(ngrams[:10])

Documentation | API Reference

Documentation is a work in progress! Have a look at it here.

👪 Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

⚖️ License

MIT

📚 References

Khanuja, Simran, et al. "GLUECoS: An Evaluation Benchmark for Code-Switched NLP." arXiv preprint arXiv:2004.12376 (2020).

ritwikmishra/py-lingualytics