app.py

import pandas as pd
import torch
import streamlit as st
from transformers import DistilBertForSequenceClassification, DistilBertTokenizerFast, Trainer, TrainingArguments

# Define the health care sentiment classification data
data = [
    {"text": "The health care services were excellent and the staff was very friendly.", "label": 1},
    {"text": "I had a bad experience with the health care services. The doctors were not knowledgeable and the staff was rude.", "label": 0},
    {"text": "The health care services were okay, but the waiting time was too long.", "label": 1},
    {"text": "I was very satisfied with the health care services. The doctors were very professional and the staff was helpful.", "label": 1},
    {"text": "The health care services were average. The doctors were not exceptional and the staff was not very friendly.", "label": 0}
]

# Convert the data to a pandas dataframe
df = pd.DataFrame(data)

# Load the pre-trained model and tokenizer
model = DistilBertForSequenceClassification.from_pretrained('distilbert-base-uncased-finetuned-sst-2-english')
tokenizer = DistilBertTokenizerFast.from_pretrained('distilbert-base-uncased')

# Tokenize the text and encode the labels
tokenized_inputs = tokenizer(list(df.text), padding=True, truncation=True, max_length=512)
tokenized_labels = torch.tensor(df.label)

# Define the training arguments
training_args = TrainingArguments(
    output_dir='./results',
    num_train_epochs=3,
    per_device_train_batch_size=16,
    per_device_eval_batch_size=64,
    warmup_steps=500,
    weight_decay=0.01,
    logging_dir='./logs',
    logging_steps=10,
    load_best_model_at_end=True,
    evaluation_strategy='steps',
    eval_steps=100,
    metric_for_best_model='accuracy'
)

# Define the trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_inputs,
    train_labels=tokenized_labels
)

# Train the model
trainer.train()

# Evaluate the model on a sample text
sample_text = "I had a great experience with the health care services. The doctors were very knowledgeable and the staff was friendly."
encoded_sample_text = tokenizer.encode(sample_text, return_tensors='pt')
logits = model(encoded_sample_text)[0]
probabilities = logits.softmax(dim=1)
sentiment = 'positive' if probabilities[0][1] > probabilities[0][0] else 'negative'

# Create the Streamlit app
st.title("Health Care Sentiment Classifier")
text_input = st.text_input("Enter some text to classify:")
if st.button("Classify"):
    encoded_text = tokenizer.encode(text_input, return_tensors='pt')
    logits = model(encoded_text)[0]
    probabilities = logits.softmax(dim=1)
    sentiment = 'positive' if probabilities[0][1] > probabilities[0][0] else 'negative'
    st.write(f"The sentiment of the text is {sentiment}.")
st.write(f"For example, the sentiment of '{sample_text}' is {sentiment}.")