A newer version of the Streamlit SDK is available:
1.40.1
title: Multi Purpose Text Application
emoji: 😻
colorFrom: indigo
colorTo: purple
sdk: streamlit
sdk_version: 1.31.1
app_file: Home.py
pinned: false
license: apache-2.0
Multi-Purpose Text Application
The application consists of several components that are designed to work with text. It mentions specific functionalities such as analyzing, summarizing, and translating raw text. It can also be handy in finding keywords that provide insights into the topic being discussed.
Components
The Text Analyzer involves a fine-tuned version of distilbert-base-uncased to determine text polarity and the probability of belonging to a class, and uses other packages to extract token, entity, word, and sentence statistics and perform other tasks.
The Text Summarizer utilizes a fine-tuned version of t5-small on the billsum dataset to build an abstractive summarizer that can generate new text that captures the most relevant information. To understand how well it captures all the important information, I used the ROUGE metric.
The Text Translator employs a fine-tuned version of t5-small trained on the English-French subset of the OPUS Books dataset to convert a sequence of text from English to French. To assess the quality of the translation, I utilized the SacreBLEU metric.
The Topic Modeling leverages CountVectorizer to vectorize the input text and utilizes singular value decomposition to decompose the bag-of-words matrix.
Usage
- To use the online version on HuggingFace, just follow services on the sidebar.
- To use it on your local machine, clone the project from this URL GitHub using
git clone git@github.com:amirho3einsedaghati/multi-purpose-text-application.git
then navigate to the multi-purpose-text-application directory containing the Home.py file and run the Streamlit application using the commandstreamlit run Home.py
License
Appendix:
https://github.com/amirho3einsedaghati/multi-purpose-text-application