LSTM News Categorizer

Centralized news aggregator that classifies news into 10 predefined topics

Learn how the Onix team built the LSTM News Categorizer and language detection model, allowing users to seamlessly access categorized news, boosting engagement within the client's app.

Mobile Communications
Industry
14 specialists
Team size
Kazakhstan
Location
1 year
Project duration

[ Business context ]

Onix's solution to streamlined news consumption

In today's digital age, a continuous stream of information floods our devices, making it challenging for users to sift through the noise and find content that resonates with their interests and needs.

The client has identified this modern problem, but it mirrors the sentiments of many: there"s a pressing demand for a singular, cohesive platform where news is not only accessible but is also categorized for effortless browsing.

The Onix team needed to develop an innovative news aggregator that categorizes news articles from various sources into ten distinct topics, enhancing user experience by allowing efficient news consumption based on their preferences.

The Onix team needed to:

Establish an efficient project management process to meet deadlines and milestones.

Develop and train the model to accurately categorize news articles into predefined ten topics such as World, Health, Sports, and more.

Craft a language detection model based on a perceptron, enabling the system to identify news languages effectively.

Ensure integration with various APIs and services to ensure an extensive news source pool.

Develop a webview-based interface accessible on Android and iOS apps, ensuring a delightful user experience.

Integrate the client's existing services, ensuring a cohesive user journey and enabling user identification within the app.

Deliver a flawless product, ensuring the system's reliability, performance, and accuracy in news categorization.

Have a project in mind?

Solutions we provide

Innovative news categorization with LSTM models

Our team developed LSTM models using Python to categorize news articles effectively. LSTM, a cutting-edge neural network architecture, was chosen for its capacity to grasp long-term relationships in data sequences, making it ideal for understanding and categorizing news content.

The models were trained to classify news into ten predefined topics, optimizing the user experience by organizing information based on themes such as World, Health, Business, and more. The successful integration of LSTM models elevated the platform's accuracy, resulting in precise news categorization and enhanced user engagement.

Language detector

This feature seamlessly recognizes the language of each news article, providing the ability to sort and organize the news according to the detected language.

The language detector significantly enhances the versatility of the news aggregator, ensuring that users can effortlessly access news in their preferred languages. This feature optimizes the user experience by offering language-based sorting options, making news consumption even more personalized and accessible.

Creating a robust database

Building a robust system to collect news articles from diverse sources, process them through the ML models for categorization, and securely store the information in a database. This involved seamless integration with various APIs and services to ensure an extensive news source pool.

Intuitive and visually appealing user interface

The Onix team developed a webview-based interface accessible on Android and iOS apps, ensuring a delightful user experience. The frontend allowed users to view the latest categorized news, customize their preferences, and filter news by category and source.

Integration with client's services

Integration with the client's existing services was a critical component. The team seamlessly integrated the news aggregator with the client's infrastructure, ensuring a cohesive user journey and enabling user identification within the app.

During the cooperation, the Onix team proved to be:

Innovative problem solvers

Collaborative and communicative

Adaptable and flexible

Results-driven

Dedicated to quality

Core technology stack we used

Next JS,

Python,

LSTM models,

Language detector,

TensorFlow,

SpaCy,

NLTK,

Gensim,

Scikit-learn

Results

The collaboration with the Onix team culminated in the successful development and implementation of a cutting-edge news aggregation solution equipped with advanced features and technologies.

The LSTM-based categorizer demonstrated an impressive validation accuracy of over 0.93, ensuring precise news categorization into the predefined topics.

Implementing the LSTM News Categorizer and language detection model addressed the client's objective of creating a centralized news aggregator, enhancing user engagement within the app.

After implementing the LSTM News Categorizer, the client witnessed a substantial boost in user engagement, resulting in a remarkable increase in their user base. Presently, they boast an impressive 40,000 active users.