NLP

Publish Date: 2021-03-31

Word Count: 952

Read Times: 5 Min

Read Count:

Introduction

Paraphrasing and summarizing are vital so that you essay doesn’t become one long quote of other academics’ work. To paraphrase a piece of text is to write it in your own words. In this article, I will show you how I make an app that will help me rephrase the sentence I need.

Quickstart

Create desktop applications with Flask and flaskwebgui.

pip install flaskwebgui

Load the libraries we need.

import os
import torch
from transformers import PegasusForConditionalGeneration
from transformers import PegasusTokenizer
from flask import Flask, request, render_template
from flaskwebgui import FlaskUI

Paraphrase NLP

I use a pre-trained model called PEGASUS, proposed by Google in 2020. It uses self-supervised objective Gap Sentences Generation (GSG) to train a transformer encoder-decoder model.

Also, I use huggingface library to get the fine-tuned model trained by tuner007.

model_name = 'tuner007/pegasus_paraphrase'
torch_device = 'cuda' if torch.cuda.is_available() else 'cpu'

def load_model():
    global model
    global tokenizer
    tokenizer = PegasusTokenizer.from_pretrained(model_name)
    model = PegasusForConditionalGeneration.from_pretrained(model_name).to(torch_device)

def get_response(input_text, num_return_sequences, num_beams):
    batch = tokenizer([input_text], 
                      truncation=True, 
                      padding='longest', 
                      max_length=60, 
                      return_tensors="pt").to(torch_device)
    translated = model.generate(**batch, 
                                max_length=60, 
                                num_beams=num_beams, 
                                num_return_sequences=num_return_sequences, 
                                temperature=1.5)
    tgt_text = tokenizer.batch_decode(translated, 
                                      skip_special_tokens=True)
    return tgt_text

Python App GUI

Next, create a file named app.py.

touch app.py

Put the following code inside of it.

import os
import torch
from transformers import PegasusForConditionalGeneration
from transformers import PegasusTokenizer
from flask import Flask, request, render_template
from flaskwebgui import FlaskUI


app = Flask(__name__)
ui = FlaskUI(app, width=800, height=1000)
model_name = 'tuner007/pegasus_paraphrase'
torch_device = 'cuda' if torch.cuda.is_available() else 'cpu'


@app.route("/", methods=["GET", "POST"])
def index():
    if request.method == "GET":
        return render_template('index.html')
    elif request.method == "POST":
        context = request.form["paragraph"]
        num_beams = 10
        num_return_sequences = 10
        tgt_text = get_response(context, num_return_sequences, num_beams)
        return render_template('index.html', src_text=context, tgt_text=tgt_text)


@app.before_first_request
def load_model():
    global model
    global tokenizer
    tokenizer = PegasusTokenizer.from_pretrained(model_name)
    model = PegasusForConditionalGeneration.from_pretrained(model_name).to(torch_device)


def get_response(input_text, num_return_sequences, num_beams):
    batch = tokenizer([input_text], 
                      truncation=True, 
                      padding='longest', 
                      max_length=60, 
                      return_tensors="pt").to(torch_device)
    translated = model.generate(**batch, 
                                max_length=60, 
                                num_beams=num_beams, 
                                num_return_sequences=num_return_sequences, 
                                temperature=1.5)
    tgt_text = tokenizer.batch_decode(translated, 
                                      skip_special_tokens=True)
    return tgt_text


if __name__ == '__main__':
    ui.run()

Frontend Interface

Create two folders, static/css and templates, and create style.css in static/css folder, and create html files base.html and index.html in templates folder.

base.html

<!doctype html>
<html lang="en">
  <head>
    <!-- Required meta tags -->
    <meta charset="utf-8">
    <meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">

    <!-- Bootstrap CSS -->
    <link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/css/bootstrap.min.css" integrity="sha384-ggOyR0iXCbMQv3Xipma34MD+dH/1fQ784/j6cY/iJTQUOhcWr7x9JvoRxT2MZw1T" crossorigin="anonymous">

    <!-- Semantic UI -->
    <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/semantic-ui@2.4.2/dist/semantic.min.css">
    <script src="https://cdn.jsdelivr.net/npm/semantic-ui@2.4.2/dist/semantic.min.js"></script>

    <title>{% block title %} {% endblock %}</title>
  </head>
  <body>
    <nav class="navbar navbar-expand-md navbar-light bg-light">
        <a class="navbar-brand" href="{{ url_for('index')}}">Paraphraser</a>
        <button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarNav" aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation">
            <span class="navbar-toggler-icon"></span>
        </button>
        <div class="collapse navbar-collapse" id="navbarNav">
            <ul class="navbar-nav">
            <li class="nav-item active">
                <a class="nav-link" href="#">About</a>
            </li>
            </ul>
        </div>
    </nav>
    <div class="container">
        {% block content %} {% endblock %}
    </div>

    <!-- Optional JavaScript -->
    <!-- jQuery first, then Popper.js, then Bootstrap JS -->
    <script src="https://code.jquery.com/jquery-3.3.1.slim.min.js" integrity="sha384-q8i/X+965DzO0rT7abK41JStQIAqVgRVzpbzo5smXKp4YfRvH+8abtTE1Pi6jizo" crossorigin="anonymous"></script>
    <script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.7/umd/popper.min.js" integrity="sha384-UO2eT0CpHqdSJQ6hJty5KVphtPhzWj9WO1clHTMGa3JDZwrnQq4sF86dIHNDz0W1" crossorigin="anonymous"></script>
    <script src="https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/js/bootstrap.min.js" integrity="sha384-JjSmVgyd0p3pXB1rRibZUAYoIIy6OrQ6VrjIEaFf/nJGzIxFDsf4x0xIM+B07jRM" crossorigin="anonymous"></script>
  </body>
</html>

index.html

{% extends 'base.html' %}

{% block content %}
    <h1>Welcome to Paraphraser!</h1>
    <div class="form-group green-border-focus">
    <form method="POST">
        <div class="form-group">
            <textarea class="form-control" id="exampleFormControlTextarea5" placeholder="Your sentence here..." value=" " rows="10" cols="50" name="paragraph"></textarea>
        </div>
        <div class="form-group">
            <input class="btn btn-primary" type="submit" name="Submit" value="Submit">
        </div>
    </form>
    </div>
    <div>
        {% if src_text is not none %}
            <h3>Original Sentence: </h3>
            <p>{{ src_text }}</p>
            <br>
        {% endif %}
    </div>
    <div>
        {% if tgt_text is not none %}
            <h3>Paraphrased Sentences: </h3>
            <ul>
                {% for text in tgt_text %}
                    <li>{{ text }}</li>
                {% endfor %}
            </ul>
        {% endif %}
    </div>
{% endblock %}

style.css

h1 {
    border: 2px #eee solid;
    color: brown;
    text-align: center;
    padding: 30px;
}

textarea {
    border: 1px solid #ba68c8;
}
.form-control:focus {
    border: 1px solid #ba68c8;
    box-shadow: 0 0 0 0.2rem rgba(186, 104, 200, .25);
}

.green-border-focus .form-control:focus {
    border: 1px solid #8bc34a;
    box-shadow: 0 0 0 0.2rem rgba(139, 195, 74, .25);
}

Run the code!

python app.py

Conclusion

Check your paraphrasing for grammar and plagiarism. It’s fast and easy!

Yang Wang

https://penguinwang96825.github.io/Yang-Tech-Blog/Yang-Tech-Blog/2021/03/31/2021-03-31-dissertation-paraphraser/

All articles in this blog are used except for special statements CC BY 4.0 reprint polocy. If reproduced, please indicate source Yang Wang !

Python NLP Flask GUI Pegasus

Develop Strategy for Tic-Tac-Toe

Tic-tac-toe is a paper-and-pencil game for two players, X and O, who take turns marking the spaces in a 3×3 grid. The player who succeeds in placing three of their marks in a diagonal, horizontal, or vertical row is the winner. It is a solved game with a forced draw assuming best play from both players. In this article, I want to discuss whether there's a must-win strategy for this game.

2021-04-01 Mathematics

JavaScript Mathematics Strategy

Interview Preparation for IBM

Come up with a list of topics that I should research and understand to prepare for the first IBM interview. This interview will likely be more behavioral as well as testing to see how much I know about IBM.

2021-03-26 Growth

Interview IBM

Dissertation Paraphraser

Introduction

Quickstart

Paraphrase NLP

Python App GUI

Frontend Interface

Conclusion