smaller min topic size

twitter_topic_model/clustering.py

@@ -10,6 +10,7 @@ def compute_umap(embeddings: np.ndarray) -> np.ndarray:
         min_dist=0.0,
         metric='cosine',
         low_memory=False,
+        random_state=42,
     )
     return umap.fit_transform(embeddings)  
 

twitter_topic_model/model.py

 from dataclasses import dataclass
 from typing import Dict, List, Optional, Set, Tuple, Union
 
-import hdbscan
 import numpy as np
 import torch
 from keybert import KeyBERT

twitter_topic_model/plugin.py

@@ -18,7 +18,7 @@ from .preprocessing import PreprocessedTweet, preprocess_tweets
 from .schema import Cluster, ClusterEntry, TwitterTopicModel
 from .utils import get_tweets
 
-MIN_TOPIC_SIZE = 8
+MIN_TOPIC_SIZE = 6
 NUM_TOPIC_DESCRIPTORS = 3
 DESCRIPTION_DIVERSITY = 0.3