Imagine being able to summarize a book to get to the crux of it or being able to skim through one-pager summaries of patents or legal contracts. Similarly, it would be useful to be able to pick out the useful information from the heaps of newsletters we’ve subscribed to.

Text summarization is another excellent technique by NLP. It enables us to perform these complex tasks without needing any human experts. While the need for summarization was always there, it wasn’t easy to perform with statistical NLP. Therefore, it had limited accuracy and usage.

Press + to interact

Models for Summarization

Model	Description
`facebook/bart-large-cnn`	Based on `bart` and fine-tuned on the CNN Daily Mail dataset for the English language. Has the ability to summarize, translate, classify, and answer questions.
`sshleifer/distilbart-xsum-12-3`	Based on `ditilbart` and fine-tuned on the Extreme Summarization dataset. Built explicitly for text summarization tasks.
`pszemraj/long-t5-tglobal-base-16384-book-summary`	Based on `google/long-t5-tglobal-base` and fine-tuned on the kmfoda/booksum dataset. Used for text summarization.
`google/pegasus-xsum`	Based on `pegasus` and is fine-tuned on the C4 and Huge News datasets. Built explicitly for text summarization tasks.

Parameter	Type	Category	Description
`inputs`	String	Required	Specifies a string or string [ ] to be summarized
`parameters.min_length`	Integer	Optional	Specifies the minimum length of the output summary tokens
`parameters.max_length`	Integer	Optional	Specifies the maximum length of the output summary tokens
`parameters.top_k`	Integer	Optional	Specifies how many tokens will be considered during the summary creation from the input text
`parameters.top_p`	Float	Optional	Specifies the probability of the token to be added in summary from the most probable to less probable until the sum of the probabilities is higher than `top_p`
`parameters.temperature`	Float	Optional	Specifies a sampling technique, and the value ranges from 1.0 to 100.0. Setting `temperature` to `0` takes tokens with the highest probability. If we set `temperature` to `1`, it does regular sampling. At `100.0` `temperature`, it selects the tokens with uniform probability.
`options.use_cache`	Boolean	Optional	Hugging Face Inference API has a cache mechanism implemented to speed up the requests. Use it for the deterministic models. Default value is `true`.
`options.wait_for_model`	Boolean	Optional	Hugging Face Inference API models takes time to initiate and process the requests. If the value is `true`, it waits for the model to get ready instead of returning an error. Default value is `false`.

Press + to interact

// Endpoint URL
const endpointUrl = "https://api-inference.huggingface.co/models/facebook/bart-large-cnn";
const headerParameters = {
  "Authorization": "Bearer {{ACCESS_TOKEN}}"
};
// Input text to classify
const data = JSON.stringify({
  inputs: "Filby became pensive. “Clearly,” the Time Traveller proceeded, “any \
real body must have extension in four directions: it must have Length, \
Breadth, Thickness, and—Duration. But through a natural infirmity of the \
flesh, which I'll explain to you in a moment, we incline to overlook this \
fact. There are really four dimensions, three which we call the three \
planes of Space, and a fourth, Time. There is, however, a tendency to \
draw an unreal distinction between the former three dimensions and the \
latter, because it happens that our consciousness moves intermittently in \
one direction along the latter from the beginning to the end of our lives.”",
  options: {
    wait_for_model: true
  }
});
const options = {
  method: "POST",
  headers: headerParameters,
  body: data
};
async function summarizeText() {
   try {
    const response = await fetch(endpointUrl, options);
    printResponse(response);
  } catch (error) {
    printError(error);
  }
}
summarizeText();

Let’s have a look at the highlighted lines shown in the code widget above:

Line 2: We specify the endpoint URL with the facebook/bart-large-cnn model for the text summarization.
Lines 9–17: We provide input text for the text summarization.
Lines 29–36: We create a function, summarizeText, to call the API and handle exceptions.
Line 38: We call the summarizeText function to invoke the endpoint.

Response fields

The API call above returns a dictionary object or a list of dictionary objects, depending on the inputs. The response contains the following field.

Press + to interact

// Endpoint URL
const endpointUrl = "https://api-inference.huggingface.co/models/facebook/bart-large-cnn";
const headerParameters = {
  "Authorization": "Bearer {{ACCESS_TOKEN}}"
};
// Input text to classify
const data = JSON.stringify({
  inputs: "Filby became pensive. “Clearly,” the Time Traveller proceeded, “any \
real body must have extension in four directions: it must have Length, \
Breadth, Thickness, and—Duration. But through a natural infirmity of the \
flesh, which I'll explain to you in a moment, we incline to overlook this \
fact. There are really four dimensions, three which we call the three \
planes of Space, and a fourth, Time. There is, however, a tendency to \
draw an unreal distinction between the former three dimensions and the \
latter, because it happens that our consciousness moves intermittently in \
one direction along the latter from the beginning to the end of our lives.”",
  parameters: {
    min_length: 20,
    max_length: 30,
    temperature: 0
  },
  options: {
    wait_for_model: true
  }
});
const options = {
  method: "POST",
  headers: headerParameters,
  body: data
};
async function summarizeText() {
   try {
    const response = await fetch(endpointUrl, options);
    printResponse(response);
  } catch (error) {
    printError(error);
  }
}
summarizeText();

Introduction

NLP

Computer Vision

APIs Integration in React

Conclusion

Text Summarization

Summarization

Models for Summarization

Request parameters

Response fields

Examples