metadata
datasets:
- samsum
language:
- en
metrics:
- rouge
library_name: transformers
pipeline_tag: summarization
tags:
- summarization
- conversational
- seq2seq
- bart large
widget:
- text: |
Hannah: Hey, do you have Betty's number?
Amanda: Lemme check
Amanda: Sorry, can't find it.
Amanda: Ask Larry
Amanda: He called her last time we were at the park together
Hannah: I don't know him well
Amanda: Don't be shy, he's very nice
Hannah: If you say so..
Hannah: I'd rather you texted him
Amanda: Just text him 🙂
Hannah: Urgh.. Alright
Hannah: Bye
Amanda: Bye bye
model-index:
- name: bart-large-xsum-samsum-conversational_summarizer
results:
- task:
name: Abstractive Text Summarization
type: abstractive-text-summarization
dataset:
name: >-
SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive
Summarization
type: samsum
metrics:
- name: Validation ROUGE-1
type: rouge-1
value: 54.3921
- name: Validation ROUGE-2
type: rouge-2
value: 29.8078
- name: Validation ROUGE-L
type: rouge-l
value: 45.1543
- name: Test ROUGE-1
type: rouge-1
value: 53.3059
- name: Test ROUGE-2
type: rouge-2
value: 28.355
- name: Test ROUGE-L
type: rouge-l
value: 44.0953
Usage
from transformers import pipeline
summarizer_pipe = pipeline("summarization", model="yashugupta786/bart_large_xsum_samsum_conv_summarizer")
conversation_data = '''Hannah: Hey, do you have Betty's number?
Amanda: Lemme check
Amanda: Sorry, can't find it.
Amanda: Ask Larry
Amanda: He called her last time we were at the park together
Hannah: I don't know him well
Amanda: Don't be shy, he's very nice
Hannah: If you say so..
Hannah: I'd rather you texted him
Amanda: Just text him 🙂
Hannah: Urgh.. Alright
Hannah: Bye
Amanda: Bye bye
'''
summarizer_pipe(conversation_data)
Results
| key | value |
|---|---|
| eval_rouge1 | 54.3921 |
| eval_rouge2 | 29.8078 |
| eval_rougeL | 45.1543 |
| eval_rougeLsum | 49.942 |
| test_rouge1 | 53.3059 |
| test_rouge2 | 28.355 |
| test_rougeL | 44.0953 |
| test_rougeLsum | 48.9246 |
All the metric Rouge1,2,L are computed using precison and recall then computed the F measure for these Rouge recall= no of overlaping words/total no of referenced humman annotated words Rouge precision= no of overlaping words/total no of candidate machine predicted words