feat: Add concrete clustering technique for visualization

app.py

@@ -4,6 +4,7 @@ import numpy as np
 
 import streamlit as st
 import tweepy
+import hdbscan
 
 from bokeh.models import ColumnDataSource, HoverTool
 from bokeh.palettes import Cividis256 as Pallete
@@ -65,13 +66,15 @@ def draw_interactive_scatter_plot(
 # Up to here
 def generate_plot(
     df: List[str],
-    labels: List[int],
     model: SentenceTransformer,
 ) -> Figure:
     with st.spinner(text="Embedding text..."):
         embeddings = embed_text(df, model)
     # encoded_labels = encode_labels(labels)
-    encoded_labels = labels
+    cluster = hdbscan.HDBSCAN(min_cluster_size=10,
+        metric='euclidean',                      
+        cluster_selection_method='eom').fit(umap_embeddings)
+    encoded_labels = cluster.labels_
     with st.spinner("Reducing dimensionality..."):
         embeddings_2d = get_tsne_embeddings(embeddings)
     plot = draw_interactive_scatter_plot(
@@ -112,8 +115,7 @@ if tw_user:
         if tw_sample > 0:
             tweets_response = client.get_users_tweets(usr.data.id, max_results=tw_sample)
         tweets_objs += tweets_response.data
-    tweets_txt = [tweet.text for tweet in tweets_objs]
-    labels = [0] * len(tweets_txt)
+    tweets_txt = [tweet.text for tweet in tweets_objs]    
     # plot = generate_plot(df, text_column, label_column, sample, dimensionality_reduction_function, model)
-    plot = generate_plot(tweets_txt, labels, model)
+    plot = generate_plot(tweets_txt, model)
     st.bokeh_chart(plot)