metadata
language: en
tags:
- question-answering
- bert
- squad
- extractive-qa
datasets:
- squad
metrics:
- exact_match
- f1
model-index:
- name: Question_Answering
results:
- task:
type: question-answering
dataset:
name: SQuAD v1.1
type: squad
metrics:
- type: exact_match
value: 81.0501
- type: f1
value: 88.5526
BERT Fine-Tuned on SQuAD (Extractive Question Answering)
This model extracts answers to questions directly from a provided text passage.
It was fine-tuned from bert-base-cased on the SQuAD v1.1 dataset.
How It Works
Given a context (a paragraph of text) and a question, the model finds and returns the exact span of text in the context that answers the question.
Performance
| Metric | Score |
|---|---|
| Exact Match | 81.05% |
| F1 Score | 88.55% |
(Evaluated on SQuAD v1.1 validation set — 10,570 examples)
How to Use
from transformers import pipeline
qa = pipeline(
"question-answering",
model="samandar1105/Question_Answering"
)
result = qa(
question="Who designed the Eiffel Tower?",
context="The Eiffel Tower was designed by Gustave Eiffel and built between 1887 and 1889 in Paris."
)
print(result)
# {'answer': 'Gustave Eiffel', 'score': 0.99, 'start': 31, 'end': 45}
Training Details
| Parameter | Value |
|---|---|
| Base model | bert-base-cased |
| Dataset | SQuAD v1.1 (87,599 train / 10,570 val) |
| Learning rate | 2e-5 |
| Epochs | 3 |
| Batch size | 16 |
| Max sequence length | 384 |
| Stride | 128 |
| Framework | PyTorch + HuggingFace Transformers |