Knight Lab, Medill School of Journalism

Description:

1. Text Analysis and Evaluation

This enables businesses to ensure that content meets quality, safety, and compliance standards, essential for public-facing and user-generated content. The core functionality of the platform is to analyze and evaluate the quality of written content using various natural language processing techniques.

• Sentiment Analysis: Detects the emotional tone of the content (positive, negative, neutral).

• Toxicity Detection: Identifies harmful or inappropriate language to ensure safe content.

• PII Detection: Flags personally identifiable information to maintain privacy compliance.

• Readability Metrics: Evaluates the complexity and clarity of the text using metrics like Flesch Reading Ease, SMOG Index, Gunning Fog, and Dale-Chall.

2. Semantic Search and Response Generation

Provides accurate and contextually relevant results for user queries, improving productivity and decision-making based on relevant content insights. This bucket focuses on retrieving or generating relevant content based on user queries through the integration of LLMs and semantic similarity tools.

• LLM Integration (Azure OpenAI GPT-3.5): Generates or retrieves contextually relevant responses based on user queries.

• Embedding Models: Converts text into vector embeddings to enable semantic search.

• Cosine Similarity: Measures the similarity between user responses and keyphrases to match relevant content.

• Vector Storage (PGVectorStore): Stores and retrieves embeddings for efficient query processing.

3. Web Interface and Usability

Makes advanced text analysis accessible to non-technical users, empowering broader teams (e.g., marketing, customer service) to extract insights without technical expertise. This bucket deals with the usability and accessibility of the platform through an intuitive, user-friendly web interface.

• Flask-based Web Interface: Allows users to submit queries, view responses, and download reports.

• CSV Export: Users can download detailed evaluation reports, making data easy to share and analyze offline.

• Form and Query Management: Supports creating and managing different forms or questions for querying the database.

4. Data Handling and Integration

Ensures scalability, security, and compliance while handling large volumes of data and sensitive information, making it suitable for enterprise use. This bucket covers how the application manages data storage and integrates with other systems for scale, privacy, and security.

• PostgreSQL with Vector Store: Enables large-scale text storage and fast retrieval of responses based on embeddings.

• API Integration: Uses Azure OpenAI and other APIs for smooth LLM and embedding integration.

• Environment Variables & Security: Manages API keys, database credentials, and sensitive configurations securely.