supertype consultancy
Full Cycle Data Science Series

Full Cycle Data Science is an educational format that explore the application of data science and analytics implementations in an end-to-end fashion. The series often comprises of 4 or more articles, and aims to bring breadth and depth to the discussion on analytics implementation.

Twitter Sentiment Analysis: Model to Deployment

About the author
Timotius Marselo

Timotius Marselo

Timotius is a voracious learner from his days studying and living in China. He returned to Indonesia with the goal of seriously pursuing data science, investing his time across mathematics, NLP (natural language processing) and machine learning along the way.

This article series features an end-to-end machine learning project where our data scientist, Timotius Marselo, walks you through an entire process of training — and then deploying — a deep learning model (LSTM architecture) that automatically labels tweets by their sentiment score. 
The series consists of (4) parts, and by the end of it, we have a live, fully ready web application ready for prime time!
Here’s how the series breaks down:
  1. Data Collection: There is no machine learning model without training data, so this is the phase that takes us through some data collection in Python and then preparing the data for our deep learning LSTM model
  2. LSTM (Long Short-Term Memory) Model Construction: Training a neural network with data collected from (1) to predict the sentiment of tweets after some text preprecossing
  3.  Model Deployment and Web App: We deploy our model to the cloud courtesy of Streamlit, along with a front end interface for general consumption of our sentiment analysis app
  4.  Business value and case study: Using our deployed Twitter Sentiment Analyzer, we’re going to highlight some actionable items for Coursera, as well as for social scientists trying to get a sense of the societal view towards an issue such as the LGBT movement 

Twitter Sentiment Analysis – Introduction and Data Collection (Part 1)

Twitter Sentiment Analysis – Data Preprocessing and Model Building (Part 2)

Twitter Sentiment Analysis – Creating Dashboard and Deploying Model with Streamlit (Part 3)

Twitter Sentiment Analysis – Use Cases (Part 4)

Create end-to-end machine learning projects

Join Supertype Fellowship for a peer-to-peer, elective-based learning program; Ready to contribute? Be a part of Supertype Collective and be connected to a highly curated network of data scientists and full stack product developers.

Open source. community-driven.

peer to peer learning

But with 🥇 badges, timelines, points and elective-based curations.

The Supertype Fellowship platform is developed to be open source from day 1. It is the backbone of the Fellowship program (formerly Supertype Development Program), where participants learn by writing code, creating features, and shipping code to complete one of 10+ electives under mentorship and peer support.

Supertype Fellowship

Open source. community-driven.

Developer Profiles

But with PDF-print, automatic icons, mobile layouts and automatic blog feed. Lightning fast ⚡.

The Supertype Collective platform is an open source platform tool that generates visually stunning Developer Profiles that look great on every medium: mobile, tablets, PCs, and even offer a PDF export option.

It features plug and play React Components and Hooks you can bring in to effortlessly create a Developer Profile that impress, in 20 lines or less (low-code).

Automatic Sentiment Analysis for your Mobile Apps

A report generation pipeline that takes as input an AppStore or Google Play Store URL and outputs a custom 30-page PDF with aggregated summary and text analysis of app user reviews in minutes.

It uses a keyword extractor routine developed in-house to handle much of the language processing tasks related to the identification and grouping of these reviews into topics (“unfriendly paywall”, “long tutorial”, “app crashes”, “poor customer service”, “very fast loading time” etc), a task known as topic detection in the natural language processing (NLP) space.

The program also uses components of Spacy and NLU by John Snow Labs to meaningfully sift through up to tens of thousands of user reviews in determining user sentiment, with full support of 20+ languages.

The PDF that is generated can be customized to include the client’s logo, as well as an ending CTA (call-to-action) slide, making it perfect for mass lead-generation and client outreach.

Read more about our work

Case Studies

Learn more about our Data Science Consultancy services for companies & enterprises


See a gallery of projects built and shipped by Supertype's finest product developers