Updated BERTopic quality gates

bertopic/nb_bertopic_lowprep.py

@@ -7,7 +7,7 @@
 #       format_version: '1.3'
 #       jupytext_version: 1.18.0
 #   kernelspec:
-#     display_name: .venv
+#     display_name: .venv (3.12.3)
 #     language: python
 #     name: python3
 # ---
@@ -56,7 +56,7 @@ nltk.download("wordnet")
 # %%
 RECREATE_MODEL = True
 RECREATE_REDUCED_MODEL = True
-PROCESS_DATA = False
+PROCESS_DATA = True
 REDUCE_OUTLIERS = False
 
 # Data Sample Size, -1 for all data
@@ -107,6 +107,12 @@ else:
         .review.to_list()
     )
 
+# Remove all duplicate reviews
+reviews = list(set(reviews))
+
+# Remove reviews that contain less than x words
+reviews = [review for review in reviews if len(review.split()) >= 9]
+
 print("Loaded {} reviews".format(len(reviews)))
 
 # %%
@@ -115,9 +121,6 @@ rep = {
     r"\n": " ",
     r'\\"': "",
     r'"': "",
-    "mongkey": "monkey",
-    "monky": "monkey",
-    "verry": "very",
     "bali": "",
     r"\s+": " ",
 }
@@ -302,75 +305,72 @@ if REDUCE_OUTLIERS:
 #
 
 # %%
-import random
+CLASSIFICATION = False
-from pathlib import Path
+if CLASSIFICATION:
+    import random
+    from pathlib import Path
 
-# --- config ---
+    # --- config ---
-topics_to_keep = {2, 4, 5, 9, 22, 26}
+    topics_to_keep = {2, 4, 5, 9, 22, 26}
-INPUT_PATH = "../data/original/reviews.tab"  # TSV with a 'review' column
+    INPUT_PATH = "../data/original/reviews.tab"  # TSV with a 'review' column
-OUTPUT_CSV = "../data/intermediate/selected_topics_documents.csv"
+    OUTPUT_CSV = "../data/intermediate/selected_topics_documents.csv"
-OUTPUT_DIR = Path("../raft/corpus")
+    OUTPUT_DIR = Path("../raft/corpus")
-OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
 
-BATCH_SIZE = 60
+    BATCH_SIZE = 60
-MIN_CHARS = 40
+    MIN_CHARS = 40
-SEED = 42
+    SEED = 42
 
-# --- load data ---
+    # Topic model document info
-data = pd.read_csv(INPUT_PATH, sep="\t")
+    df = topic_model.get_document_info(reviews)  # assumes your model is already fitted
+    df["Original"] = reviews.values
 
-# If you already have `reviews` elsewhere, replace the next line with that variable
+    # --- filter by topics and length ---
-reviews = data["review"].astype(str).fillna("")
+    filtered = df[df["Topic"].isin(topics_to_keep)].copy()
+    filtered["Original"] = filtered["Original"].str.strip()
+    filtered = filtered[filtered["Original"].str.len() >= MIN_CHARS]
 
-# Topic model document info
+    # Save an audit CSV
-df = topic_model.get_document_info(reviews)  # assumes your model is already fitted
+    filtered[["Original", "Topic"]].to_csv(OUTPUT_CSV, index=False)
-df["Original"] = reviews.values
 
-# --- filter by topics and length ---
+    # --- deterministic shuffle + write batched corpus files ---
-filtered = df[df["Topic"].isin(topics_to_keep)].copy()
+    total_files = 0
-filtered["Original"] = filtered["Original"].str.strip()
+    total_reviews = 0
-filtered = filtered[filtered["Original"].str.len() >= MIN_CHARS]
+    rng = random.Random(SEED)
 
-# Save an audit CSV
+    for topic_val, g in filtered.groupby("Topic", sort=True):
-filtered[["Original", "Topic"]].to_csv(OUTPUT_CSV, index=False)
+        reviews_list = g["Original"].tolist()
 
-# --- deterministic shuffle + write batched corpus files ---
+        # deterministic shuffle within topic
-total_files = 0
+        rng.shuffle(reviews_list)
-total_reviews = 0
-rng = random.Random(SEED)
 
-for topic_val, g in filtered.groupby("Topic", sort=True):
+        # chunk into batches of up to 60
-    reviews_list = g["Original"].tolist()
+        for start in range(0, len(reviews_list), BATCH_SIZE):
+            chunk = reviews_list[start : start + BATCH_SIZE]
+            if not chunk:
+                continue
 
-    # deterministic shuffle within topic
+            # simple header for traceability
-    rng.shuffle(reviews_list)
+            header = (
+                f"[TOPIC] {topic_val}\n"
+                + f"[Stats] N={len(chunk)} | Source={INPUT_PATH}\n"
+            )
 
-    # chunk into batches of up to 60
+            lines = [header, ""]
-    for start in range(0, len(reviews_list), BATCH_SIZE):
+            for i, txt in enumerate(chunk, 1):
-        chunk = reviews_list[start : start + BATCH_SIZE]
+                lines.append(f"({i}) {txt}")
-        if not chunk:
-            continue
 
-        # simple header for traceability
+            part_idx = start // BATCH_SIZE + 1
-        header = (
+            fname = f"topic={topic_val}__part={part_idx:03d}__n={len(chunk)}.txt"
-            f"[TOPIC] {topic_val}\n" + f"[Stats] N={len(chunk)} | Source={INPUT_PATH}\n"
+            (OUTPUT_DIR / fname).write_text("\n".join(lines), encoding="utf-8")
-        )
 
-        lines = [header, ""]
+            total_files += 1
-        for i, txt in enumerate(chunk, 1):
+            total_reviews += len(chunk)
-            lines.append(f"({i}) {txt}")
 
-        part_idx = start // BATCH_SIZE + 1
+    print(
-        fname = f"topic={topic_val}__part={part_idx:03d}__n={len(chunk)}.txt"
+        f"[green]Wrote {total_files} docs with {total_reviews} reviews to {OUTPUT_DIR}[/green]"
-        (OUTPUT_DIR / fname).write_text("\n".join(lines), encoding="utf-8")
+    )
-
+    print(f"[green]Filtered CSV saved to {OUTPUT_CSV}[/green]")
-        total_files += 1
-        total_reviews += len(chunk)
-
-print(
-    f"[green]Wrote {total_files} docs with {total_reviews} reviews to {OUTPUT_DIR}[/green]"
-)
-print(f"[green]Filtered CSV saved to {OUTPUT_CSV}[/green]")
 
 # %%
 doc_topic_matrix = probs