app.py

import streamlit as st
from transformers import pipeline
import plotly.express as px
import pandas as pd


st.set_page_config(layout="wide")

@st.cache(allow_output_mutation = True)
def get_classifier_model():
    return pipeline("zero-shot-classification", model="typeform/distilbert-base-uncased-mnli")
    #return pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
    #return pipeline("zero-shot-classification",model="sentence-transformers/paraphrase-MiniLM-L6-v2")

#st.sidebar.image("Suncorp-Bank-logo.png",width=255)

#st.image("Suncorp-Bank-logo.png",width=255)

st.title("Review Analyzer")
st.markdown("***")

text = st.text_area(label="Paste/Type the review here..")
    
st.markdown("***")

col1, col2, col3 = st.columns((1,1,1))

col1.header("Select Sentiments")
sentiments = col1.multiselect("",["Happy","Sad","Neutral"],["Happy","Sad","Neutral"])
col1.markdown(" \n")
col1.markdown(" \n")

additional_sentiments = col1.text_input("Enter comma separated sentiments.")

if additional_sentiments:
    sentiments = sentiments + additional_sentiments.split(",")

col2.header("Select Topics")
entities = col2.multiselect("",["Bank Account","Credit Card","Home Loan","Motor Loan"],
                            ["Bank Account","Credit Card","Home Loan","Motor Loan"])
additional_entities= col2.text_input("Enter comma separated entities.")

if additional_entities:
    entities = entities + additional_entities.split(",")


col3.header("Select Reasons")

reasons = col3.multiselect("",["Poor Service","No Empathy","Abuse"],
                            ["Poor Service","No Empathy","Abuse"])
additional_reasons= col3.text_input("Enter comma separated reasons.")

if additional_reasons:
    reasons = reasons + additional_reasons.split(",")

is_multi_class =  st.checkbox("Can have more than one classes",value=True)

st.markdown("***")

classify_button_clicked = st.button("Classify")

def get_classification(candidate_labels):
    classification_output = classifier(sequence_to_classify, candidate_labels, multi_class=is_multi_class)
    data = {'Class': classification_output['labels'], 'Scores': classification_output['scores']}
    df = pd.DataFrame(data)
    df = df.sort_values(by='Scores', ascending=False)
    fig = px.bar(df, x='Scores', y='Class', orientation='h', width=400, height=500)
    fig.update_layout(
        yaxis=dict(
            autorange='reversed'
        )
    )
    return fig



if classify_button_clicked:
    if text:
        st.markdown("***")
        with st.spinner("  Please wait while the text is being classified.."):
            classifier = get_classifier_model()
            sequence_to_classify = text
            # candidate_labels = sentiments + entities + reasons

            if sentiments:
                #print(classification_output)
                fig = get_classification(sentiments)
                # col5, col6= st.columns((1, 1))
                col1.markdown(" \n")
                
                col1.write(fig)

            if entities:
                #print(classification_output)
                fig = get_classification(entities)
                # col7, col8= st.columns((1, 1))
                
                col2.write(fig)

            if reasons:
                #print(classification_output)
                fig = get_classification(reasons)
                # col7, col8= st.columns((1, 1))
                col3.write(fig)