pages/3_Traditional_Techniques_of_NLP.py

import streamlit as st
import string
import numpy as np
import pandas as pd
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.datasets import fetch_20newsgroups
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.decomposition import LatentDirichletAllocation
from sklearn.decomposition import NMF

# Page title
st.title('Traditional NLP Techniques')

# Text Preprocessing
st.header('1. Text Preprocessing')

st.subheader('Definition:')
st.write("""
Text preprocessing is the process of cleaning and preparing raw text for further analysis or modeling. 
This includes tasks such as removing unnecessary punctuation, converting text to lowercase, 
and handling special characters like emojis.
""")

# Interactive example for preprocessing
text_input = st.text_area("Enter text to preprocess", "I love NLP! 😍 This is amazing.")

# Punctuation removal
if st.button('Remove Punctuation'):
    processed_text = ''.join([char for char in text_input if char not in string.punctuation])
    st.write(f"Text without punctuation: {processed_text}")

# Convert to lowercase
if st.button('Convert to Lowercase'):
    lowercase_text = text_input.lower()
    st.write(f"Text in lowercase: {lowercase_text}")

# Handle emojis (replace with a message)
if st.button('Remove Emojis'):
    processed_text_no_emoji = ''.join(char for char in text_input if char.isalnum() or char.isspace())
    st.write(f"Text without emojis: {processed_text_no_emoji}")

# Text Vectorization
st.header('2. Text Vectorization')

st.subheader('Definition:')
st.write("""
Text vectorization converts text into numerical form so that machine learning models can process it. 
Two common techniques are Bag of Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF).
""")

# Interactive example for vectorization
vectorization_choice = st.selectbox('Choose vectorization technique:', ('Bag of Words', 'TF-IDF'))

# Text for vectorization
sample_text = ["I love programming.", "NLP is fun.", "Streamlit makes things easy!"]

if st.button('Apply Vectorization'):
    if vectorization_choice == 'Bag of Words':
        vectorizer = CountVectorizer()
        X = vectorizer.fit_transform(sample_text)
        st.write(f"Bag of Words representation:\n{X.toarray()}")
        st.write(f"Feature names: {vectorizer.get_feature_names_out()}")
    elif vectorization_choice == 'TF-IDF':
        vectorizer = TfidfVectorizer()
        X = vectorizer.fit_transform(sample_text)
        st.write(f"TF-IDF representation:\n{X.toarray()}")
        st.write(f"Feature names: {vectorizer.get_feature_names_out()}")

# Basic Machine Learning
st.header('3. Basic Machine Learning')

st.subheader('Definition:')
st.write("""
Basic machine learning techniques, such as Naive Bayes, Logistic Regression, and Support Vector Machines (SVM), 
are commonly used for text classification tasks.
""")

# Load dataset
newsgroups = fetch_20newsgroups(subset='train')
X_train, X_test, y_train, y_test = train_test_split(newsgroups.data, newsgroups.target, test_size=0.3)

model_choice = st.selectbox('Choose machine learning model for text classification:', 
                            ('Naive Bayes', 'Logistic Regression', 'SVM'))

# Vectorization for classification
vectorizer = TfidfVectorizer()
X_train_vec = vectorizer.fit_transform(X_train)
X_test_vec = vectorizer.transform(X_test)

# Train model based on choice
if st.button('Train Model'):
    if model_choice == 'Naive Bayes':
        model = MultinomialNB()
    elif model_choice == 'Logistic Regression':
        model = LogisticRegression(max_iter=1000)
    elif model_choice == 'SVM':
        model = SVC()

    model.fit(X_train_vec, y_train)
    y_pred = model.predict(X_test_vec)

    accuracy = accuracy_score(y_test, y_pred)
    st.write(f"Model Accuracy: {accuracy * 100:.2f}%")

# Topic Modeling
st.header('4. Topic Modeling')

st.subheader('Definition:')
st.write("""
Topic modeling is a technique used to identify the underlying topics in a collection of text data. 
Latent Dirichlet Allocation (LDA) and Non-negative Matrix Factorization (NMF) are two common techniques for this task.
""")

topic_model_choice = st.selectbox('Choose topic modeling technique:', ('LDA', 'NMF'))

# Apply LDA or NMF for topic modeling
if st.button('Run Topic Modeling'):
    vectorizer = TfidfVectorizer(max_df=0.95, min_df=2)
    X = vectorizer.fit_transform(newsgroups.data)

    if topic_model_choice == 'LDA':
        model = LatentDirichletAllocation(n_components=3, random_state=42)
    elif topic_model_choice == 'NMF':
        model = NMF(n_components=3, random_state=42)

    model.fit(X)
    feature_names = vectorizer.get_feature_names_out()

    # Display top words for each topic
    for topic_idx, topic in enumerate(model.components_):
        st.write(f"Topic {topic_idx + 1}:")
        top_words_idx = topic.argsort()[:-10 - 1:-1]
        top_words = [feature_names[i] for i in top_words_idx]
        st.write(", ".join(top_words))