Remix

Make your own version of Testimony Drift Analyzer

Edit the prompt below and generate a customized version in seconds.

Original by SHANMUGA PRIYA M CSE-2023

Describe your app

Build a **complete, fully functional full-stack web application** for:

# Longitudinal Semantic Drift Detection in Witness Testimony

Do NOT create only a UI prototype. Create and connect the **frontend, backend, database, NLP/AI processing, and all features**.

## Tech Stack

* Frontend: React.js + Vite + Tailwind CSS + Axios + React Router + Recharts
* Backend: Node.js + Express.js
* Database: MongoDB + Mongoose
* Authentication: JWT + bcrypt
* NLP: Python + spaCy + sentence-transformers
* OCR: Tesseract OCR
* Speech-to-Text: suitable open-source/local solution
* Reports: PDF and DOCX generation

## Main Workflow

User Login → Dashboard → Create Case → Add Witness/Victim/Accused → Add Multiple Statements → OCR/Speech-to-Text/Text Input → NLP Processing → Compare Statements Chronologically → Detect Semantic Drift → Generate Findings → View Timeline → Generate Report.

## Required Features

### 1. Authentication

* Register
* Login
* Logout
* JWT authentication
* Protected routes
* Password hashing

### 2. Case Management

Complete CRUD:

* Create
* View
* Edit
* Delete
* Search
* Filter

### 3. Person Management

Support:

* Witness
* Victim
* Accused

### 4. Statement Management

Allow multiple statements for the same person:

* Police statement
* Magistrate statement
* Court statement
* Subsequent statement

Support:

* Direct text input
* PDF
* DOCX
* Image
* Audio

Allow the user to edit extracted OCR/transcribed text before analysis.

### 5. OCR

Implement actual OCR for images/scanned documents.

Workflow:

Upload → OCR → Extract Text → Edit → Save → Analyze.

### 6. Speech-to-Text

Implement:

Audio → Transcript → User Verification → Save → NLP Analysis.

### 7. NLP Processing

For every statement perform:

* Text preprocessing
* Sentence segmentation
* Tokenization
* Lemmatization
* Named Entity Recognition
* Dependency parsing
* Event extraction
* Fact extraction
* Sentence embeddings
* Semantic similarity

Extract:

* Person
* Location
* Date
* Time
* Organization
* Object
* Quantity
* Subject
* Action
* Object

### 8. Core Semantic Drift Detection

This is the **main feature**.

Compare statements chronologically and detect:

* **Addition**
* **Omission**
* **Mutation**
* **Contradiction**

Do NOT depend only on character/word matching. Use semantic similarity and structured entity/event comparison.

Example:

Statement 1:
"I saw Ravi at 8 PM."

Statement 2:
"I saw Ravi at 10 PM carrying a knife."

Detect:

* Time mutation: 8 PM → 10 PM
* Addition: carrying a knife

### 9. Semantic Drift Score

Calculate a transparent **0–100 Semantic Drift Score** using:

* Entity changes
* Event changes
* Time/date changes
* Location changes
* Additions
* Omissions
* Contradictions

Display:

* 0–20 Very Low
* 21–40 Low
* 41–60 Moderate
* 61–80 High
* 81–100 Critical

### 10. Legal Significance Score

Calculate an explainable **0–100 score** based on:

* Identity changes
* Action changes
* Location changes
* Time changes
* Object/weapon changes
* Presence/absence
* Major contradictions
* Minor wording changes

Classify:

* Low
* Medium
* High
* Critical

Do NOT claim that this determines guilt, innocence, truthfulness, or legal liability.

### 11. Memory Contamination Pattern Detection

Implement transparent rule-based detection for:

* Progressive detail addition
* New entity introduction
* Repeated newly introduced details
* Event expansion
* Temporal shifts

Always display:

**"Potential Memory Contamination Pattern"**

Never say that memory contamination is confirmed.

### 12. Comparison Page

Create a side-by-side comparison:

**Earlier Statement | Later Statement**

Highlight:

* Added information
* Removed information
* Changed information
* Contradictions

For every finding display:

* Change type
* Earlier value
* Later value
* Confidence
* Semantic similarity
* Drift score
* Legal significance
* Explanation

### 13. Timeline

Create an interactive timeline:

Police → Magistrate → Court → Subsequent

Show for every statement:

* Date
* Drift score
* Number of changes
* Contradictions
* Legal significance

### 14. Dashboard

Create a real database-connected dashboard showing:

* Total cases
* Total persons
* Total statements
* Analyzed statements
* Total findings
* Contradictions
* High-significance findings
* Average drift score

Add charts for:

* Changes by type
* Semantic drift over time
* Legal significance

Do NOT hardcode dashboard values.

### 15. Reports

Generate professional PDF/DOCX reports containing:

* Case information
* Person information
* Statement timeline
* Executive summary
* Semantic drift summary
* Additions
* Omissions
* Mutations
* Contradictions
* Entity/event changes
* Potential memory contamination patterns
* Legal significance
* Detailed findings
* Methodology
* Limitations
* Legal disclaimer

Reports must contain **actual analysis results**, not dummy data.

### 16. Database

Create MongoDB/Mongoose models for:

* User
* Case
* Person
* Statement
* Analysis
* Finding
* Entity
* Event
* Report

### 17. Backend APIs

Implement real REST APIs for:

* Authentication
* Cases
* Persons
* Statements
* OCR
* Speech-to-Text
* NLP processing
* Comparison
* Findings
* Timeline
* Reports

Connect every API to the React frontend.

### 18. Sample Data

Create fictional sample data containing:

* Multiple cases
* Multiple witnesses
* Police/Magistrate/Court statements
* Additions
* Omissions
* Mutations
* Contradictions
* Time changes
* Location changes
* Action changes

The system must actually analyze this sample data.

### 19. Project Structure

Create:

```text
project/
├── frontend/
├── backend/
├── nlp-service/
├── sample-data/
├── reports/
├── docs/
├── .env.example
└── README.md
```

### 20. Important Requirement

The final application must be **fully connected**:

```text
React Frontend
      ↕
Node.js + Express Backend
      ↕
MongoDB
      ↕
Python NLP Service
      ↕
OCR / Speech-to-Text / NLP
```

Do not use fake buttons, hardcoded analysis results, or placeholder implementations.

Create actual working code for all major features.

After creating the project, verify:

* Frontend builds
* Backend runs
* MongoDB connection works
* APIs work
* Authentication works
* Statements can be added
* NLP analysis works
* Semantic comparison works
* Drift score works
* Legal significance works
* Contamination detection works
* Reports are generated

Finally provide the **complete folder structure, setup commands, environment variables, database setup, and run instructions**.

Free to use. No credit card required.