Spaces:
Running
Running
Jatin Mehra
commited on
Commit
·
2f7e733
1
Parent(s):
0d7f003
Add RAG system performance metrics section to README
Browse files
README.md
CHANGED
@@ -6,6 +6,7 @@ An advanced PDF document analysis tool that combines RAG (Retrieval Augmented Ge
|
|
6 |
|
7 |
- [Overview](#overview)
|
8 |
- [Features](#features)
|
|
|
9 |
- [Architecture](#architecture)
|
10 |
- [Technical Stack](#technical-stack)
|
11 |
- [Installation](#installation)
|
@@ -34,6 +35,39 @@ The application employs an agentic approach that can augment the document's info
|
|
34 |
- **Hugging Face Integration**: Automatic deployment to Hugging Face Spaces
|
35 |
- **Android Application**: Native mobile client
|
36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
## Architecture
|
38 |
|
39 |
The application follows a modular architecture with these main components:
|
|
|
6 |
|
7 |
- [Overview](#overview)
|
8 |
- [Features](#features)
|
9 |
+
- [RAG SYSTEM PERFORMANCE](#rag-system-metrics)
|
10 |
- [Architecture](#architecture)
|
11 |
- [Technical Stack](#technical-stack)
|
12 |
- [Installation](#installation)
|
|
|
35 |
- **Hugging Face Integration**: Automatic deployment to Hugging Face Spaces
|
36 |
- **Android Application**: Native mobile client
|
37 |
|
38 |
+
|
39 |
+
## RAG System Metrics
|
40 |
+
|
41 |
+
1. **Key Metrics Overview**:
|
42 |
+
|
43 |
+
| Metric | Value |
|
44 |
+
| ----------------------------------- | ------- |
|
45 |
+
| **Semantic Similarity (Mean)** | `0.852` |
|
46 |
+
| **ROUGE-L F1 Score (Mean)** | `0.395` |
|
47 |
+
| **Semantic Similarity (Max)** | `1.000` |
|
48 |
+
| **ROUGE-L F1 Score (Max)** | `1.000` |
|
49 |
+
| **Semantic Similarity (Min)** | `0.592` |
|
50 |
+
| **ROUGE-L F1 Score (Min)** | `0.099` |
|
51 |
+
| **Standard Deviation (Similarity)** | `0.089` |
|
52 |
+
| **Standard Deviation (ROUGE-L F1)** | `0.217` |
|
53 |
+
|
54 |
+
|
55 |
+
2. **Quantile Distribution**:
|
56 |
+
|
57 |
+
| Percentile | Semantic Similarity | ROUGE-L F1 Score |
|
58 |
+
| ---------- | ------------------- | ---------------- |
|
59 |
+
| **25%** | `0.7946` | `0.2516` |
|
60 |
+
| **50%** | `0.8732` | `0.3256` |
|
61 |
+
| **75%** | `0.9181` | `0.4951` |
|
62 |
+
|
63 |
+
|
64 |
+
3. **Evaluation Status**:
|
65 |
+
|
66 |
+
| Status | Count | Percentage |
|
67 |
+
| ------ | ----- | ---------- |
|
68 |
+
| PASS | `64` | `85.3%` |
|
69 |
+
| FAIL | `11` | `14.7%` |
|
70 |
+
|
71 |
## Architecture
|
72 |
|
73 |
The application follows a modular architecture with these main components:
|