SheenXO/bayesian_simulator

1import streamlit as st2import numpy as np3import pandas as pd4import matplotlib.pyplot as plt5import seaborn as sns6from nltk.tokenize import word_tokenize7import nltk8 9nltk.download('punkt_tab')10 11st.title("📊 Bayesian Token Co-occurrence Simulator")12 13# User input14user_input = st.text_area("✍️ Enter your training sentences (one per line):", 15"""16fido loves the red ball17timmy and fido go to the park18fido and timmy love to play19the red ball is timmy's favorite toy20""")21 22sentences = user_input.strip().split('\n')23tokenized = [word_tokenize(s.lower()) for s in sentences if s.strip()]24vocab = sorted(set(word for sentence in tokenized for word in sentence))25token2idx = {word: i for i, word in enumerate(vocab)}26idx2token = {i: word for word, i in token2idx.items()}27 28# Co-occurrence matrix29window_size = 230matrix = np.zeros((len(vocab), len(vocab)))31 32for sentence in tokenized:33    for i, word in enumerate(sentence):34        for j in range(max(0, i - window_size), min(len(sentence), i + window_size + 1)):35            if i != j:36                matrix[token2idx[word]][token2idx[sentence[j]]] += 137 38alpha = st.slider("🔧 Set Bayesian Prior (α smoothing)", 0.0, 2.0, 0.1)39posterior = matrix + alpha40 41df = pd.DataFrame(posterior, index=vocab, columns=vocab)42st.subheader("📈 Co-occurrence Heatmap")43fig, ax = plt.subplots(figsize=(10, 8))44sns.heatmap(df, annot=True, cmap="Blues", fmt=".1f", ax=ax)45st.pyplot(fig)46 47# Next-token prediction48selected_word = st.selectbox("🔮 Predict next token after:", vocab)49row = posterior[token2idx[selected_word]]50probs = row / row.sum()51prediction = np.random.choice(vocab, p=probs)52 53st.markdown(f"**Predicted next token:** `{prediction}`")54