app.py

import streamlit as st
import nltk
from transformers import pipeline
from sentence_transformers import SentenceTransformer
from scipy.spatial.distance import cosine
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
import tensorflow as tf
import tensorflow_hub as hub


def cluster_examples(messages, embed, nc=3):
    km = KMeans(
    n_clusters=nc, init='random',
    n_init=10, max_iter=300, 
    tol=1e-04, random_state=0
    )
    km = km.fit_predict(embed)
    for n in range(nc):
        idxs = [i for i in range(len(km)) if km[i] == n]
        ms = [messages[i] for i in idxs]
        st.markdown ("CLUSTER : %d"%n)
        for m in ms:
            st.markdown (m)


def plot_heatmap(labels, heatmap, rotation=90):
  sns.set(font_scale=1.2)
  fig, ax = plt.subplots()
  g = sns.heatmap(
      heatmap,
      xticklabels=labels,
      yticklabels=labels,
      vmin=-1,
      vmax=1,
      cmap="coolwarm")
  g.set_xticklabels(labels, rotation=rotation)
  g.set_title("Textual Similarity")
  st.pyplot(fig)
  
# Streamlit text boxes
text = st.text_area('Enter sentences:', value="Behavior right this is a kind of Heisenberg uncertainty principle situation if I told you, then you behave differently. What would be the impressive thing is you have talked about winning a nobel prize in a system winning a nobel prize. Adjusting it and then making your own.  That is when I fell in love with computers.  I realized that they were a very magical device.  Can go to sleep come back the next day and it is solved.  You know that feels magical to me.")

nc = st.slider('Select a number of clusters:', min_value=1, max_value=15, value=3)

model_type = st.radio("Choose model:", ('Sentence Transformer', 'Universal Sentence Encoder'), index=0)

# Model setup
if model_type == "Sentence Transformer":
    model = SentenceTransformer('paraphrase-distilroberta-base-v1')
elif model_type == "Universal Sentence Encoder":
    model_url = "https://tfhub.dev/google/universal-sentence-encoder-large/5"
    model = hub.load(model_url)

nltk.download('punkt')

# Run model
if text:
    sentences = nltk.tokenize.sent_tokenize(text)
    if model_type == "Sentence Transformer":
        embed = model.encode(sentences)
    elif model_type == "Universal Sentence Encoder":
        embed = model(sentences).numpy()
    sim = np.zeros([len(embed), len(embed)])
    for i,em in enumerate(embed):
        for j,ea in enumerate(embed):
            sim[i][j] = 1.0-cosine(em,ea)
    st.subheader("Similarity Heatmap")
    plot_heatmap(sentences, sim)
    st.subheader("Results from K-Means Clustering")
    cluster_examples(sentences, embed, nc)