karanverma19 commited on
Commit
0ed1faf
·
verified ·
1 Parent(s): 03fe07c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +32 -11
README.md CHANGED
@@ -1,25 +1,45 @@
1
  ---
2
  license: mit
3
  ---
 
4
  # Trilingual Fraud & Consumer Protection Dataset (V2)
5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  ## Overview
7
- This dataset focuses on real-world fraud detection and consumer protection scenarios across English, Hindi, and Punjabi. It is designed to reflect how scams actually appear in everyday communication, often subtle, conversational, and not always obviously malicious.
8
 
9
- Instead of only clear-cut examples, the dataset includes multi-step interactions, edge cases, and realistic safe responses to better capture how intent evolves in real situations.
10
 
11
- ---
 
 
12
 
13
  ## Key Features
 
14
  - Multilingual coverage (English, Hindi, Punjabi)
15
  - Code-mixed conversational data common in Indian contexts
16
  - Multi-step fraud sequences (trust-building → payment → urgency)
17
  - Realistic safe responses (not just advisory, but conversational)
18
  - Edge cases that capture misleading or borderline scenarios
19
 
20
- ---
21
-
22
  ## Structure
 
23
  Each entry includes:
24
  - **text** – message content
25
  - **label** – scam or safe
@@ -29,16 +49,17 @@ Each entry includes:
29
  - **reasoning_tag** – underlying behavioral signal
30
  - **difficulty** – relative complexity
31
 
32
- ---
33
-
34
  ## Use Cases
 
35
  - Fraud and scam detection systems
36
  - LLM safety and alignment evaluation
37
  - Multilingual NLP benchmarking
38
  - Consumer protection and risk modeling
39
 
40
- ---
41
-
42
  ## Notes
43
- The focus of this dataset is on **structure, realism, and variation**, rather than synthetic or templated generation.
44
- Small differences in wording can significantly change intent, and this dataset is designed to capture those nuances.
 
 
 
 
 
1
  ---
2
  license: mit
3
  ---
4
+
5
  # Trilingual Fraud & Consumer Protection Dataset (V2)
6
 
7
+ ## Why this dataset matters
8
+
9
+ Fraud detection in regions like North India is not straightforward.
10
+
11
+ Messages are rarely written in clean English. Instead, they are often a mix of Punjabi, Hindi, and English, and many scam messages look very similar to legitimate communication.
12
+
13
+ In practice, the harder part is not spotting obvious scams, but telling apart messages that look almost the same but mean very different things.
14
+
15
+ Many existing datasets focus on clear scam patterns or keyword-based signals. However, this creates a gap where models learn surface patterns instead of actual intent.
16
+
17
+ I built this dataset to try to address that gap by focusing on:
18
+ - contrastive examples (similar wording, different intent)
19
+ - realistic safe responses (not just warnings, but natural conversation)
20
+ - multi-step scam flows where intent evolves over time
21
+ - code-mixed language that reflects how people actually communicate
22
+
23
+ In practice, this turns fraud detection from a pattern recognition problem into an intent understanding problem.
24
+
25
  ## Overview
 
26
 
27
+ This dataset focuses on capturing subtle differences in intent between scam and legitimate messages in real-world multilingual communication across English, Hindi, and Punjabi.
28
 
29
+ It tries to reflect how scams actually appear in everyday communication, often subtle, conversational, and not always obviously malicious.
30
+
31
+ Instead of only using clear-cut examples, the dataset includes multi-step interactions, edge cases, and realistic safe responses to better capture how intent evolves in real situations.
32
 
33
  ## Key Features
34
+
35
  - Multilingual coverage (English, Hindi, Punjabi)
36
  - Code-mixed conversational data common in Indian contexts
37
  - Multi-step fraud sequences (trust-building → payment → urgency)
38
  - Realistic safe responses (not just advisory, but conversational)
39
  - Edge cases that capture misleading or borderline scenarios
40
 
 
 
41
  ## Structure
42
+
43
  Each entry includes:
44
  - **text** – message content
45
  - **label** – scam or safe
 
49
  - **reasoning_tag** – underlying behavioral signal
50
  - **difficulty** – relative complexity
51
 
 
 
52
  ## Use Cases
53
+
54
  - Fraud and scam detection systems
55
  - LLM safety and alignment evaluation
56
  - Multilingual NLP benchmarking
57
  - Consumer protection and risk modeling
58
 
 
 
59
  ## Notes
60
+
61
+ The focus of this dataset is on structure, realism, and variation, rather than synthetic or templated generation.
62
+
63
+ Small differences in wording can significantly change intent, and this dataset is designed to capture those nuances.
64
+
65
+ This is still an evolving dataset, and I’m continuing to improve edge cases where even humans might hesitate.