Demo.py

import streamlit as st
import sparknlp
import os
import pandas as pd

from sparknlp.base import *
from sparknlp.annotator import *
from pyspark.ml import Pipeline
from sparknlp.pretrained import PretrainedPipeline
from annotated_text import annotated_text
from streamlit_tags import st_tags

# Page configuration
st.set_page_config(
    layout="wide", 
    initial_sidebar_state="auto"
)

# CSS for styling
st.markdown("""
    <style>
        .main-title {
            font-size: 36px;
            color: #4A90E2;
            font-weight: bold;
            text-align: center;
        }
        .section {
            background-color: #f9f9f9;
            padding: 10px;
            border-radius: 10px;
            margin-top: 10px;
        }
        .section p, .section ul {
            color: #666666;
        }
    </style>
""", unsafe_allow_html=True)

@st.cache_resource
def init_spark():
    return sparknlp.start()

@st.cache_resource
def create_pipeline(model, task, zeroShotLables=['']):
    document_assembler = DocumentAssembler() \
        .setInputCol('text') \
        .setOutputCol('document')

    sentence_detector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "xx")\
        .setInputCols(["document"])\
        .setOutputCol("sentence")

    tokenizer = Tokenizer() \
        .setInputCols(['sentence']) \
        .setOutputCol('token')

    if task == "Token Classification":
        TCclassifier = DeBertaForTokenClassification \
            .pretrained(model, "en") \
            .setInputCols(["sentence", "token"]) \
            .setOutputCol("ner") \
            .setCaseSensitive(False) \
            .setMaxSentenceLength(512)
        
        ner_converter = NerConverter() \
            .setInputCols(['sentence', 'token', 'ner']) \
            .setOutputCol('ner_chunk')

        TCpipeline = Pipeline(stages=[document_assembler, sentence_detector, tokenizer, TCclassifier, ner_converter])
        return TCpipeline

    elif task == "Zero-Shot Classification":
        ZSCtokenizer = Tokenizer() \
            .setInputCols(['document']) \
            .setOutputCol('token')
        
        zeroShotClassifier = DeBertaForZeroShotClassification \
            .pretrained('deberta_base_zero_shot_classifier_mnli_anli_v3', 'en') \
            .setInputCols(['token', 'document']) \
            .setOutputCol('class') \
            .setCaseSensitive(False) \
            .setMaxSentenceLength(512) \
            .setCandidateLabels(zeroShotLables)

        ZSCpipeline = Pipeline(stages=[document_assembler, ZSCtokenizer, zeroShotClassifier])
        return ZSCpipeline

    elif task == "Sequence Classification":
        SCtokenizer = Tokenizer() \
            .setInputCols(['document']) \
            .setOutputCol('token')

        sequence_classifier = DeBertaForSequenceClassification \
            .pretrained("deberta_v3_base_sequence_classifier_imdb", "en") \
            .setInputCols(["document", "token"]) \
            .setOutputCol("class")
        
        SCpipeline = Pipeline(stages=[document_assembler, SCtokenizer, sequence_classifier])
        return SCpipeline

def fit_data(pipeline, data, task):
    empty_df = spark.createDataFrame([['']]).toDF('text')
    pipeline_model = pipeline.fit(empty_df)
    model = LightPipeline(pipeline_model)
    result = model.fullAnnotate(data)
    return result

def annotate(data):
    document, chunks, labels = data["Document"], data["NER Chunk"], data["NER Label"]
    annotated_words = []
    for chunk, label in zip(chunks, labels):
        parts = document.split(chunk, 1)
        if parts[0]:
            annotated_words.append(parts[0])
        annotated_words.append((chunk, label))
        document = parts[1]
    if document:
        annotated_words.append(document)
    annotated_text(*annotated_words)

tasks_models_descriptions = {
    "Token Classification": {
        "models": ["deberta_v3_small_token_classifier_conll03", "deberta_v3_large_token_classifier_conll03"],
        "description": "The 'deberta_v3_small_token_classifier_conll03' model is adept at token classification tasks, including named entity recognition (NER). It identifies and categorizes tokens in text, such as names, dates, and locations, enhancing the extraction of meaningful information from unstructured data."
    },
    "Zero-Shot Classification": {
        "models": ["deberta_base_zero_shot_classifier_mnli_anli_v3"],
        "description": "The 'deberta_base_zero_shot_classifier_mnli_anli_v3' model provides flexible text classification without needing training data for specific categories. It is ideal for dynamic scenarios where text needs to be categorized into topics like urgent issues, technology, or sports without prior labeling."
    },
    "Sequence Classification": {
        "models": ["deberta_v3_base_sequence_classifier_imdb"],
        "description": "The 'deberta_v3_base_sequence_classifier_imdb' model is proficient in sequence classification tasks, such as sentiment analysis and document categorization. It effectively determines the sentiment of reviews, classifies text, and sorts documents based on their content and context."
    }
}

# Sidebar content
task = st.sidebar.selectbox("Choose the task", list(tasks_models_descriptions.keys()))
model = st.sidebar.selectbox("Choose the pretrained model", tasks_models_descriptions[task]["models"], help="For more info about the models visit: https://sparknlp.org/models")

# Reference notebook link in sidebar
link = """
<a href="https://github.com/JohnSnowLabs/spark-nlp-workshop/blob/357691d18373d6e8f13b5b1015137a398fd0a45f/Spark_NLP_Udemy_MOOC/Open_Source/17.01.Transformers-based_Embeddings.ipynb#L103">
    <img src="https://colab.research.google.com/assets/colab-badge.svg" style="zoom: 1.3" alt="Open In Colab"/>
</a>
"""
st.sidebar.markdown('Reference notebook:')
st.sidebar.markdown(link, unsafe_allow_html=True)

# Page content
title, sub_title = (f'DeBERTa for {task}', tasks_models_descriptions[task]["description"])
st.markdown(f'<div class="main-title">{title}</div>', unsafe_allow_html=True)
container = st.container(border=True)
container.write(sub_title)

# Load examples
examples_mapping = {
    "Token Classification": [
        "William Henry Gates III (born October 28, 1955) is an American business magnate, software developer, investor, and philanthropist. He is best known as the co-founder of Microsoft Corporation. During his career at Microsoft, Gates held the positions of chairman, chief executive officer (CEO), president and chief software architect, while also being the largest individual shareholder until May 2014. He is one of the best-known entrepreneurs and pioneers of the microcomputer revolution of the 1970s and 1980s. Born and raised in Seattle, Washington, Gates co-founded Microsoft with childhood friend Paul Allen in 1975, in Albuquerque, New Mexico; it went on to become the world's largest personal computer software company. Gates led the company as chairman and CEO until stepping down as CEO in January 2000, but he remained chairman and became chief software architect. During the late 1990s, Gates had been criticized for his business tactics, which have been considered anti-competitive. This opinion has been upheld by numerous court rulings. In June 2006, Gates announced that he would be transitioning to a part-time role at Microsoft and full-time work at the Bill & Melinda Gates Foundation, the private charitable foundation that he and his wife, Melinda Gates, established in 2000.[9] He gradually transferred his duties to Ray Ozzie and Craig Mundie. He stepped down as chairman of Microsoft in February 2014 and assumed a new post as technology adviser to support the newly appointed CEO Satya Nadella.",
        "The Mona Lisa is a 16th century oil painting created by Leonardo. It's held at the Louvre in Paris.",
        "When Sebastian Thrun started working on self-driving cars at Google in 2007, few people outside of the company took him seriously. “I can tell you very senior CEOs of major American car companies would shake my hand and turn away because I wasn’t worth talking to,” said Thrun, now the co-founder and CEO of online higher education startup Udacity, in an interview with Recode earlier this week.",
        "Facebook is a social networking service launched as TheFacebook on February 4, 2004. It was founded by Mark Zuckerberg with his college roommates and fellow Harvard University students Eduardo Saverin, Andrew McCollum, Dustin Moskovitz and Chris Hughes. The website's membership was initially limited by the founders to Harvard students, but was expanded to other colleges in the Boston area, the Ivy League, and gradually most universities in the United States and Canada.",
        "The history of natural language processing generally started in the 1950s, although work can be found from earlier periods. In 1950, Alan Turing published an article titled 'Computing Machinery and Intelligence' which proposed what is now called the Turing test as a criterion of intelligence",
        "Geoffrey Everest Hinton is an English Canadian cognitive psychologist and computer scientist, most noted for his work on artificial neural networks. Since 2013 he divides his time working for Google and the University of Toronto. In 2017, he cofounded and became the Chief Scientific Advisor of the Vector Institute in Toronto.",
        "When I told John that I wanted to move to Alaska, he warned me that I'd have trouble finding a Starbucks there.",
        "Steven Paul Jobs was an American business magnate, industrial designer, investor, and media proprietor. He was the chairman, chief executive officer (CEO), and co-founder of Apple Inc., the chairman and majority shareholder of Pixar, a member of The Walt Disney Company's board of directors following its acquisition of Pixar, and the founder, chairman, and CEO of NeXT. Jobs is widely recognized as a pioneer of the personal computer revolution of the 1970s and 1980s, along with Apple co-founder Steve Wozniak. Jobs was born in San Francisco, California, and put up for adoption. He was raised in the San Francisco Bay Area. He attended Reed College in 1972 before dropping out that same year, and traveled through India in 1974 seeking enlightenment and studying Zen Buddhism.",
        "Titanic is a 1997 American epic romance and disaster film directed, written, co-produced, and co-edited by James Cameron. Incorporating both historical and fictionalized aspects, it is based on accounts of the sinking of the RMS Titanic, and stars Leonardo DiCaprio and Kate Winslet as members of different social classes who fall in love aboard the ship during its ill-fated maiden voyage.",
        "Other than being the king of the north, John Snow is a an english physician and a leader in the development of anaesthesia and medical hygiene. He is considered for being the first one using data to cure cholera outbreak in 1834."
    ],
    "Zero-Shot Classification" : [
        "In today’s world, staying updated with urgent information is crucial as events can unfold rapidly and require immediate attention.",  # Urgent
        "Mobile technology has become indispensable, allowing us to access news, updates, and connect with others no matter where we are.",  # Mobile
        "For those who love to travel, the convenience of mobile apps has transformed how we plan and experience trips, providing real-time updates on flights, accommodations, and local attractions.",  # Travel
        "The entertainment industry continually offers new movies that captivate audiences with their storytelling and visuals, providing a wide range of genres to suit every taste.",  # Movie
        "Music is an integral part of modern life, with streaming platforms making it easy to discover new artists and enjoy favorite tunes anytime, anywhere.",  # Music
        "Sports enthusiasts follow games and matches closely, with live updates and detailed statistics available at their fingertips, enhancing the excitement of every game.",  # Sport
        "Weather forecasts play a vital role in daily planning, offering accurate and timely information to help us prepare for various weather conditions and adjust our plans accordingly.",  # Weather
        "Technology continues to evolve rapidly, driving innovation across all sectors and improving our everyday lives through smarter devices, advanced software, and enhanced connectivity."  # Technology
    ],
    "Sequence Classification": [
        "This movie was absolutely fantastic! The storyline was gripping, the characters were well-developed, and the cinematography was stunning. I was on the edge of my seat the entire time.",
        "A heartwarming and beautiful film. The performances were top-notch, and the direction was flawless. This is easily one of the best movies I've seen this year.",
        "What a delightful surprise! The humor was spot on, and the plot was refreshingly original. The cast did an amazing job bringing the characters to life. Highly recommended!",
        "This was one of the worst movies I’ve ever seen. The plot was predictable, the acting was wooden, and the pacing was painfully slow. I couldn’t wait for it to end.",
        "A complete waste of time. The movie lacked any real substance or direction, and the dialogue was cringe-worthy. I wouldn’t recommend this to anyone.",
        "I had high hopes for this film, but it turned out to be a huge disappointment. The story was disjointed, and the special effects were laughably bad. Don’t bother watching this one.",
        "The movie was okay, but nothing special. It had a few good moments, but overall, it felt pretty average. Not something I would watch again, but it wasn’t terrible either.",
        "An average film with a decent plot. The acting was passable, but it didn't leave much of an impression on me. It's a movie you might watch once and forget about.",
        "This movie was neither good nor bad, just kind of there. It had some interesting ideas, but they weren’t executed very well. It’s a film you could take or leave."
    ]
}

examples = examples_mapping[task]
selected_text = st.selectbox("Select an example", examples)
custom_input = st.text_input("Try it with your own Sentence!")

if task == 'Zero-Shot Classification':
    zeroShotLables = ["urgent", "mobile", "travel", "movie", "music", "sport", "weather", "technology"]
    lables = st_tags(
        label='Select labels',
        text='Press enter to add more',
        value=zeroShotLables,
        suggestions=[
          "Positive", "Negative", "Neutral",
          "Urgent", "Mobile", "Travel", "Movie", "Music", "Sport", "Weather", "Technology",
          "Happiness", "Sadness", "Anger", "Fear", "Surprise", "Disgust",
          "Informational", "Navigational", "Transactional", "Commercial Investigation",
          "Politics", "Business", "Sports", "Entertainment", "Health", "Science",
          "Product Quality", "Delivery Experience", "Customer Service", "Pricing", "Return Policy",
          "Education", "Finance", "Lifestyle", "Fashion", "Food", "Art", "History",
          "Culture", "Environment", "Real Estate", "Automotive", "Travel", "Fitness", "Career"],
        maxtags = -1)

try:
  text_to_analyze = custom_input if custom_input else selected_text
  st.subheader('Full example text')
  HTML_WRAPPER = """<div class="scroll entities" style="overflow-x: auto; border: 1px solid #e6e9ef; border-radius: 0.25rem; padding: 1rem; margin-bottom: 2.5rem; white-space:pre-wrap">{}</div>"""
  st.markdown(HTML_WRAPPER.format(text_to_analyze), unsafe_allow_html=True)
except:
  text_to_analyze = selected_text

# Initialize Spark and create pipeline
spark = init_spark()
if task == 'Zero-Shot Classification':
    pipeline = create_pipeline(model, task, zeroShotLables)
else:
    pipeline = create_pipeline(model, task)
output = fit_data(pipeline, text_to_analyze, task)

# Display matched sentence
st.subheader("Prediction:")

if task == 'Token Classification':
    results = {
        'Document': output[0]['document'][0].result,
        'NER Chunk': [n.result for n in output[0]['ner_chunk']],
        'NER Label': [n.metadata['entity'] for n in output[0]['ner_chunk']]
    }
    annotate(results)
    df = pd.DataFrame({'NER Chunk': results['NER Chunk'], 'NER Label': results['NER Label']})
    df.index += 1
    st.dataframe(df)

elif task == 'Zero-Shot Classification':
    st.markdown(f"Document Classified as: **{output[0]['class'][0].result}**")

elif task == 'Sequence Classification':
    st.markdown(f"Classified as : **{output[0]['class'][0].result}**")