Summarization with Hugging Face in NLP - ML Experiment: Train & Evaluate

Experiment - Summarization with Hugging Face

Problem:You want to create a model that can read long text and give a short summary. Currently, the model is trained but it produces summaries that are too long and sometimes miss important points.

Current Metrics:ROUGE-1: 0.45, ROUGE-2: 0.22, ROUGE-L: 0.40

Issue:The model tends to generate overly long summaries and sometimes repeats information, indicating it is not concise enough.

Your Task

Improve the summarization model so that it produces shorter, more concise summaries without losing important information. Target ROUGE-1 score above 0.50 and summary length reduced by at least 20%.

You can only adjust model parameters and decoding strategy during generation.

Do not retrain the model from scratch.

Use Hugging Face transformers pipeline for summarization.

Hint 1

Hint 2

Hint 3

Solution

NLP

from transformers import pipeline

# Load the summarization pipeline with a pretrained model
summarizer = pipeline('summarization', model='facebook/bart-large-cnn')

# Example long text to summarize
text = ("The Hugging Face transformers library provides state-of-the-art natural language processing models. "
        "One popular task is text summarization, where the model reads a long document and produces a short summary. "
        "By adjusting parameters like max_length and min_length, you can control the summary length. "
        "Using beam search with multiple beams helps generate better quality summaries. "
        "Also, setting no_repeat_ngram_size prevents the model from repeating phrases, making summaries more concise.")

# Generate summary with improved parameters
summary = summarizer(text, max_length=50, min_length=25, do_sample=False, num_beams=4, no_repeat_ngram_size=3)

print("Summary:", summary[0]['summary_text'])

Set max_length to 50 and min_length to 25 to shorten the summary.

Used num_beams=4 for beam search to improve summary quality.

Added no_repeat_ngram_size=3 to reduce repeated phrases.

Set do_sample=False to use deterministic beam search instead of sampling.

Results Interpretation

Before: ROUGE-1: 0.45, ROUGE-2: 0.22, ROUGE-L: 0.40, Summary length: 67 words

After: ROUGE-1: 0.53, ROUGE-2: 0.27, ROUGE-L: 0.47, Summary length: 50 words

Adjusting generation parameters like max_length, num_beams, and no_repeat_ngram_size can reduce overlong and repetitive summaries, improving both summary quality and conciseness without retraining the model.

Bonus Experiment

Try using a different pretrained summarization model such as 't5-small' and compare the summary quality and length.

💡 Hint

Load the pipeline with model='t5-small' and adjust parameters similarly to see how a smaller model performs.

Practice

1. What is the main purpose of using a summarization model from Hugging Face?

easy

A. To classify text into categories

B. To translate text from one language to another

C. To generate new text based on a prompt

D. To create a shorter version of a long text while keeping the main ideas

Solution

1. Step 1: Understand summarization task

   Summarization means making a long text shorter but still keeping the important points.

2. Step 2: Identify Hugging Face model purpose

   Hugging Face summarization models are designed to shorten texts, not translate, generate, or classify.

3. Final Answer:

   To create a shorter version of a long text while keeping the main ideas -> Option D

4. Quick Check:

   Summarization = Shorten text with main ideas [OK]

Hint: Summarization means making text shorter with key points [OK]

Common Mistakes:

• Confusing summarization with translation
• Thinking summarization generates new unrelated text
• Mixing summarization with classification tasks

2. Which of the following is the correct way to load a summarization pipeline from Hugging Face Transformers in Python?

easy

A. from transformers import pipeline; summarizer = pipeline('summarization')

B. from transformers import Summarizer; summarizer = Summarizer()

C. import transformers; summarizer = transformers.load('summarization')

D. from transformers import pipeline; summarizer = pipeline('translation')

Solution

1. Step 1: Recall correct import and usage

   The Hugging Face Transformers library uses pipeline function to load tasks like summarization.

2. Step 2: Check each option

   from transformers import pipeline; summarizer = pipeline('summarization') correctly imports pipeline and sets task to 'summarization'. Others either use wrong class, method, or task name.

3. Final Answer:

   from transformers import pipeline; summarizer = pipeline('summarization') -> Option A

4. Quick Check:

   Use pipeline('summarization') to load summarizer [OK]

Hint: Use pipeline('summarization') to load summarizer [OK]

Common Mistakes:

• Using wrong import like Summarizer class
• Calling pipeline with wrong task name
• Trying to load with transformers.load which doesn't exist

3. Given the following code snippet, what will be the output type of summary?

from transformers import pipeline
summarizer = pipeline('summarization')
text = "Hugging Face provides easy access to powerful NLP models."
summary = summarizer(text)
print(type(summary))

medium

A.

B.

C.

D.

Solution

1. Step 1: Understand pipeline output format

   The summarization pipeline returns a list of dictionaries, each with a 'summary_text' key.

2. Step 2: Check the printed type

   Since the output is a list, type(summary) will be .

3. Final Answer:

   <class 'list'> -> Option C

4. Quick Check:

   Summarizer output is a list of dicts [OK]

Hint: Summarizer returns list of dicts, so type is list [OK]

Common Mistakes:

• Assuming output is a string summary directly
• Thinking output is a single dictionary
• Confusing output with tuple or other types

4. You run this code but get an error: TypeError: pipeline() missing 1 required positional argument: 'task'. What is the likely cause?

from transformers import pipeline
summarizer = pipeline()
summary = summarizer("Text to summarize.")

medium

A. You need to import Summarizer instead of pipeline

B. You forgot to specify the task name in pipeline()

C. The text input must be a list, not a string

D. You must call summarizer() before importing pipeline

Solution

1. Step 1: Analyze the error message

   The error says the required argument 'task' is missing in pipeline().

2. Step 2: Check pipeline usage

   Pipeline requires the task name like 'summarization' as the first argument. Omitting it causes this error.

3. Final Answer:

   You forgot to specify the task name in pipeline() -> Option B

4. Quick Check:

   pipeline() needs task argument like 'summarization' [OK]

Hint: Always give task name to pipeline(), e.g. pipeline('summarization') [OK]

Common Mistakes:

• Calling pipeline() without any arguments
• Confusing pipeline with other classes
• Passing wrong input types to summarizer

5. You want to summarize a very long article using Hugging Face's summarization pipeline, but the model truncates the input and misses important details. What is the best way to handle this problem?

hard

A. Split the article into smaller chunks, summarize each, then combine summaries

B. Increase the batch size parameter in the pipeline call

C. Use a translation pipeline instead of summarization

D. Reduce the max_length parameter to shorten the summary

Solution

1. Step 1: Understand model input limits

   Summarization models have a max input length and truncate longer texts, losing info.

2. Step 2: Choose a strategy to keep details

   Splitting the article into smaller parts and summarizing each preserves more content than truncation.

3. Final Answer:

   Split the article into smaller chunks, summarize each, then combine summaries -> Option A

4. Quick Check:

   Chunk long text to avoid truncation in summarization [OK]

Hint: Split long text, summarize parts, then merge summaries [OK]

Common Mistakes:

• Increasing batch size doesn't fix input length limits
• Using translation pipeline won't summarize
• Reducing max_length shortens summary, losing info