dayangrah.am

serena : a Bible verse recommendation app.

An app that uses NLP to find a Bible verse relevant to anything that you say. Built with React Native and Node.js.

Open the app and say anything on your mind; Serena will find a relevant Bible verse and podcast. The app, written in React Native, creates both iPhone and Android apps from a single codebase. The backend, written in Node.js, uses Word2Vec embeddings to perform a cosine similarity search among 5000 distinct topics, returning a relevant verse over the API. It also learns from past searches to recommend relevant verses and podcasts.

Technical Details

Building Serena involved several key technical components:

  1. React Native Frontend:

    • Utilized React Native for cross-platform development, allowing a single codebase to generate both iOS and Android apps.
    • Implemented a clean, intuitive UI for user input and verse display.
    • Integrated speech-to-text functionality for seamless user interaction.

  2. Node.js Backend:

    • Developed a robust API using Express.js to handle requests from the mobile app.
    • Implemented Natural Language Processing (NLP) techniques:
      • Used Word2Vec embeddings to convert user input and Bible verses into vector representations.
      • Employed cosine similarity calculations to find the most relevant verses based on user input.

  3. Database and Caching:

    • Stored Bible verses and their corresponding vector representations in a database for quick retrieval.
    • Implemented caching mechanisms to improve response times for frequently requested verses.

  4. Word2Vec Implementation:

    • Utilized the Word2Vec model, a two-layer neural network that processes text and outputs vector representations of words.
    • Pre-trained the Word2Vec model on a large corpus of religious texts to capture semantic relationships specific to biblical language.
    • Implemented a custom tokenization process to handle biblical names and phrases effectively.
    • Used the Gensim library in Python to train and manage the Word2Vec model.
    • Optimized the model to balance between accuracy and performance, considering the app’s real-time response requirements.

  5. API Design:

    • Created RESTful endpoints for verse retrieval, podcast recommendations, and user history synchronization.
    • Implemented proper error handling and response formatting for a smooth user experience.

  6. Performance Optimization:

    • Optimized the similarity search algorithm to handle a large dataset of 5000 distinct topics efficiently.
    • Implemented lazy loading and pagination techniques to manage large sets of verses and podcasts.

  7. Security Measures:

    • Implemented secure API authentication to protect user data and prevent unauthorized access.
    • Ensured proper data sanitization to prevent injection attacks.

These technical aspects combined to create a responsive, intelligent, and user-friendly Bible verse recommendation app that leverages advanced NLP techniques to provide relevant suggestions.

---