# resham-patel/hinglish-sentiment-analysis architecture

> Architecture diagram of the GitHub repository resham-patel/hinglish-sentiment-analysis, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-17). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/resham-patel/hinglish-sentiment-analysis
- Repository: https://github.com/resham-patel/hinglish-sentiment-analysis

## Overview

This is a Hinglish three-class sentiment-classification project. Its main workflow prepares labeled CONLL data, removes conflicting training examples, fine-tunes MuRIL with dev-set model selection, then evaluates once on the held-out test set. Separate branches provide interactive text classification, optional YouTube out-of-domain evaluation, and test-error analysis; the README also documents classical TF-IDF baselines. Source excerpts show the preparation, training, evaluation, and demo implementations; the README supplies the baseline and saved-model details.

Main entry points: [`src/train_transformer.py`](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/train_transformer.py) (MuRIL fine-tuning), [`src/evaluate_final.py`](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/evaluate_final.py) (Final evaluation), [`src/prepare_data.py`](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/prepare_data.py) (CONLL preparation), [`src/demo_app.py`](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/demo_app.py) (Gradio sentiment demo)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Data preparation"]
  node_raw["SentiMix raw data"]
  node_prepare["CONLL preparation<br/>[prepare_data.py]"]
  node_clean["Clean datasets"]
  node_conflicts["Conflict removal<br/>[prepare_data.py]"]
  node_finaltrain["Conflict-free training data"]
end

subgraph group_model["Model training"]
  node_baselines["TF-IDF baselines<br/>[final.ipynb]"]
  node_trainer["MuRIL fine-tuning"]
  node_modelartifact["Fine-tuned model"]
end

subgraph group_inference["Inference and evaluation"]
  node_demo["Gradio sentiment demo<br/>[demo_app.py]"]
  node_evaluate["Final evaluation<br/>[evaluate_final.py]"]
  node_test["Test predictions"]
  node_youtube["YouTube comments"]
end

subgraph group_analysis["Error analysis"]
  node_errors["Error analysis<br/>[error_analysis.py]"]
  node_errorfile["Test error report"]
end

node_user(("Text user"))

node_raw -->|"provides input"| node_prepare
node_prepare -->|"writes datasets"| node_clean
node_clean -->|"supplies training CSV"| node_conflicts
node_conflicts -->|"writes filtered data"| node_finaltrain
node_finaltrain -->|"trains on"| node_trainer
node_clean -->|"selects on dev data"| node_trainer
node_trainer -->|"saves model"| node_modelartifact
node_user -->|"enters text"| node_demo
node_demo -->|"loads and predicts"| node_modelartifact
node_clean -->|"provides test data"| node_evaluate
node_modelartifact -->|"loads model"| node_evaluate
node_youtube -.->|"optional input"| node_evaluate
node_evaluate -->|"writes predictions"| node_test
node_test -->|"analyzes mistakes"| node_errors
node_errors -->|"writes errors"| node_errorfile
node_baselines -.->|"compares models"| node_trainer

click node_raw "https://github.com/resham-patel/hinglish-sentiment-analysis/tree/main/data/raw"
click node_prepare "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/prepare_data.py"
click node_conflicts "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/prepare_data.py"
click node_baselines "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/notebook/final.ipynb"
click node_trainer "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/train_transformer.py"
click node_demo "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/demo_app.py"
click node_evaluate "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/evaluate_final.py"
click node_errors "https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/error_analysis.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_raw,node_prepare,node_clean,node_conflicts,node_finaltrain,node_user toneBlue
class node_baselines,node_trainer,node_modelartifact toneAmber
class node_demo,node_evaluate,node_test,node_youtube toneMint
class node_errors,node_errorfile toneRose
```

## Components

### Data preparation

- [SentiMix raw data](https://github.com/resham-patel/hinglish-sentiment-analysis/tree/main/data/raw): `data/raw`
- [CONLL preparation](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/prepare_data.py): `src/prepare_data.py`
- **Clean datasets**
- [Conflict removal](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/prepare_data.py): `src/prepare_data.py`
- **Conflict-free training data**

### Model training

- [TF-IDF baselines](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/notebook/final.ipynb): `notebook/final.ipynb`
- [MuRIL fine-tuning](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/train_transformer.py): `src/train_transformer.py`
- **Fine-tuned model**

### Inference and evaluation

- [Gradio sentiment demo](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/demo_app.py): `src/demo_app.py`
- [Final evaluation](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/evaluate_final.py): `src/evaluate_final.py`
- **Test predictions**
- **YouTube comments**

### Error analysis

- [Error analysis](https://github.com/resham-patel/hinglish-sentiment-analysis/blob/main/src/error_analysis.py): `src/error_analysis.py`
- **Test error report**

### Other components

- **Text user**

## Connections

- SentiMix raw data → CONLL preparation: provides input (evidence: `src/prepare_data.py`)
- CONLL preparation → Clean datasets: writes datasets (evidence: `src/prepare_data.py`)
- Clean datasets → Conflict removal: supplies training CSV (evidence: `src/prepare_data.py`)
- Conflict removal → Conflict-free training data: writes filtered data (evidence: `src/prepare_data.py`)
- Conflict-free training data → MuRIL fine-tuning: trains on (evidence: `src/train_transformer.py`)
- Clean datasets → MuRIL fine-tuning: selects on dev data (evidence: `src/train_transformer.py`)
- MuRIL fine-tuning → Fine-tuned model: saves model (evidence: `README.md`)
- Text user → Gradio sentiment demo: enters text (evidence: `src/demo_app.py`)
- Gradio sentiment demo → Fine-tuned model: loads and predicts (evidence: `src/demo_app.py`)
- Clean datasets → Final evaluation: provides test data (evidence: `src/evaluate_final.py`)
- Fine-tuned model → Final evaluation: loads model (evidence: `src/evaluate_final.py`)
- YouTube comments → Final evaluation: optional input (evidence: `src/evaluate_final.py`)
- Final evaluation → Test predictions: writes predictions (evidence: `src/evaluate_final.py`)
- Test predictions → Error analysis: analyzes mistakes (evidence: `src/error_analysis.py`)
- Error analysis → Test error report: writes errors (evidence: `src/error_analysis.py`)
- TF-IDF baselines → MuRIL fine-tuning: compares models (evidence: `README.md`)
