https://github.com/robgc/sento-processing

A Natural Language Processing tool designed to perform sentiment analysis on tweets and store the results obtained.
https://github.com/robgc/sento-processing

async asyncpg nlp python sentiment-analysis spacy spacy2

Last synced: 9 months ago
JSON representation

A Natural Language Processing tool designed to perform sentiment analysis on tweets and store the results obtained.

• Host: GitHub
• URL: https://github.com/robgc/sento-processing
• Owner: robgc
• License: agpl-3.0
• Created: 2019-03-16T20:04:06.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2022-12-08T04:53:30.000Z (over 3 years ago)
• Last Synced: 2025-04-12T19:58:27.026Z (about 1 year ago)
• Topics: async, asyncpg, nlp, python, sentiment-analysis, spacy, spacy2
• Language: Python
• Homepage: https://robgc.github.io/sento-blog/
• Size: 98.6 KB
• Stars: 1
• Watchers: 0
• Forks: 1
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Sento Processing

*This project is part of Sento's backend*.

A Natural Language Processing tool designed to perform sentiment analysis

on tweets and store the results obtained.

# Table of contents

# Prerequisites

Sento Processing requires a spaCy model that categorizes text based on the possible sentiment

that it holds. The model must return the probabilities of three categories:

- `P` for positive sentiment.

- `N` for negative sentiment.

- `NEU` for neutral sentiment.

If you wish to create a spaCy model that can perform this task, in the `sento_processing/training`

directory you have some scripts used to develop a model for performing sentiment analysis of

tweets written in spanish. Treat them as an example that may be outdated, as the spaCy team could

have changed their API and those scripts may no longer work. You can store your models in the

`models` directory in the repository root, this directory will be ignored by git.

Processing requires a PostgreSQL database initialised previously using the instructions

available in [Sento API's readme](https://github.com/robgc/sento-api).

# Executing the tool

## Choosing your environment

You have two options:

- Running the Docker container using Docker Compose: `docker-compose up -d`.

  You need the following software installed:

  - _Docker Engine 17.12.0 or greater required_.

  - _Docker Compose 1.18.0 or greater required_.

- Running locally on your machine, requiring:

  - Python 3.7 or greater.

  - Pipenv.

## Configuring the tool

Create a `config.ini` file from a copy of `config.example.ini` and adjust

the configuration according to your needs. If you use the PostgreSQL container

provided in Sento API instructions, let the default value of `sento-db` in the

`[postgres].host` section of your `config.ini`.

## Running the tool

- **With Docker**: run `docker-compose up -d`, this will compile the container image for you if

  you have not done it previously. If you make any changes to your `config.ini` after running

  the container you will need to stop it, remove it and recreate the container's image before

  creating another container instance. This container will try to connect

  to the `sento-net` Docker network, where the PostGIS container is listening for connections.

- **Running locally**:

  - Install the necessary dependencies in a virtual environment with `pipenv sync`.

  - Run the following command `pipenv run sento_processing/main.py`, this will start the tool.

# Acknowledgements

The training and evaluation datasets used for creating models for Sento Processing come from

the Spanish Society for Natural Language Processing (SEPLN, in Spain). They were made for

the Workshop on Semantic Analysis at SEPLN (TASS, in Spain). You can download them

from their [official website](http://www.sepln.org/workshops/tass/).

# License

The source code of this project is licensed under the GNU Affero General Public License v3.0.