oduonye commited on
Commit
c18344b
·
verified ·
1 Parent(s): 1acf44b

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +41 -0
README.md ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ pipeline_tag: text-classification
4
+ tags:
5
+ - kubeflow
6
+ - kubeflow-pipelines
7
+ - banking77
8
+ - customer-support
9
+ base_model: distilbert/distilbert-base-uncased
10
+ datasets:
11
+ - PolyAI/banking77
12
+ license: cc-by-4.0
13
+ language:
14
+ - en
15
+ ---
16
+
17
+ # SupportFlow Banking Intent Router
18
+
19
+ ## Intended use
20
+
21
+ SupportFlow predicts the operational intent of an English customer-support message so a queue, help-desk workflow, or human agent can route it faster.
22
+
23
+ It is not a bank, does not move money, and must not make credit, fraud, identity, or eligibility decisions. Low-confidence predictions should go to manual review.
24
+
25
+ ## Training data
26
+
27
+ - Dataset: [PolyAI/banking77](https://huggingface.co/datasets/PolyAI/banking77)
28
+ - Source: [PolyAI task-specific datasets](https://github.com/PolyAI-LDN/task-specific-datasets)
29
+ - License: CC BY 4.0
30
+ - Task: 77-way English banking-support intent classification
31
+
32
+ ## Base model and training
33
+
34
+ The pipeline fine-tunes [`distilbert/distilbert-base-uncased`](https://huggingface.co/distilbert/distilbert-base-uncased) with a deterministic seed, a stratified validation split, and the untouched upstream test split for final evaluation.
35
+
36
+ ## Limitations
37
+
38
+ - The examples are English and banking-specific.
39
+ - The model can be wrong when a message contains several issues or a new intent.
40
+ - Dataset labels and customer language may not match a particular company's support taxonomy.
41
+ - Accuracy and F1 are benchmark measures, not a guarantee of safe production behavior.