Update app.py

app.py

@@ -1,10 +1,9 @@
 # ================================
 # 🪞 MoodMirror+ — Conversational Emotional Self-Care
 # Dataset-only: TF-IDF + OneVsRest Logistic Regression on GoEmotions
-# - Persists model & SQLite DB to /data (enable Persistent storage on HF Spaces)
-# - Always give at least one advice tip; sometimes add a quote
-# - Implicit emotion recognition (emoji/slang/negations hints)
-# - Dynamic threshold + Top-1 fallback
+# - Persists model & SQLite DB to /data (HF Spaces persistent storage)
+# - Always gives an advice tip (plus sometimes a quote)
+# - Detects implicit emotion cues (emoji/slang/negation)
 # ================================
 import os
 import re
@@ -32,13 +31,13 @@ DATA_DIR = os.getenv("MM_DATA_DIR", _pick_data_dir())
 os.makedirs(DATA_DIR, exist_ok=True)
 DB_PATH = os.path.join(DATA_DIR, "moodmirror.db")
 MODEL_PATH = os.path.join(DATA_DIR, "goemo_sklearn.joblib")
-MODEL_VERSION = "v2-tfidf-lr-ovr-implicit"  # bump when training recipe changes
+MODEL_VERSION = "v3-always-tip"
 
-print(f"[MM] Using data dir: {DATA_DIR}")
-print(f"[MM] SQLite path:   {DB_PATH}")
-print(f"[MM] Model path:    {MODEL_PATH}")
+print(f"[MM] Data dir: {DATA_DIR}")
+print(f"[MM] DB path:  {DB_PATH}")
+print(f"[MM] Model:    {MODEL_PATH}")
 
-# ---------------- Crisis & regex ----------------
+# ---------------- Regex & Crisis ----------------
 CRISIS_RE = re.compile(
     r"\b(self[- ]?harm|suicid|kill myself|end my life|overdose|cutting|i don.?t want to live|can.?t go on)\b",
     re.I,
@@ -63,70 +62,49 @@ SUGGESTIONS = {
         "Be gentle with yourself. Cry if you need to — that’s healing, not weakness.",
         "Let yourself rest. You don’t have to be productive while your heart is heavy.",
         "Hold a pillow or blanket close and breathe slowly.",
-        "Talk to someone who feels safe, or write down what you wish someone would say to you.",
         "It’s okay if today you just get through the day.",
-        "Take a warm shower, drink water, and do one small thing that feels kind.",
-        "Remember: sadness is a visitor, not your home.",
+        "Talk to someone who feels safe, or write down what you wish someone would say to you.",
         "Even quiet tears are a form of strength — you’re still here.",
+        "Take a warm shower, drink water, and do one small thing that feels kind.",
     ],
     "fear": [
         "Ground yourself with slow breathing — feel your feet on the floor.",
         "You are safe in this moment. Focus on what is real right now.",
-        "Courage isn’t about not being afraid; it’s choosing to keep going anyway.",
         "Remind yourself: not every thought is a fact.",
-        "Small steps are still progress — you don’t need to have the whole plan.",
-        "Try saying out loud: ‘I can handle this one moment at a time.’",
+        "Courage isn’t the absence of fear, it’s choosing to go on anyway.",
         "Place a hand on your chest and whisper, ‘I am safe. I am present.’",
-        "Fear often shows up before growth — maybe something new is calling.",
+        "Try saying: ‘I can handle this one moment at a time.’",
     ],
     "joy": [
         "Savor this — your body deserves to feel good.",
-        "Let the joy sink in; you don’t have to rush to the next thing.",
         "Share your happiness with someone — even a smile counts.",
         "Write down what’s bringing you joy today. It strengthens the memory.",
-        "Take a deep breath and feel gratitude for this small light moment.",
         "Joy can be quiet — a cup of tea, a laugh, a calm pause.",
-        "Smile for no reason; your brain will thank you later.",
-        "Notice how joy feels in your body, and let yourself stay there for a while.",
+        "Let the joy sink in; you don’t have to rush to the next thing.",
     ],
     "anger": [
         "Pause and breathe — you don’t have to react right away.",
         "Anger can mean something important needs attention — listen kindly.",
         "Take a walk or shake out the tension before speaking.",
-        "Write everything you want to say, then decide what actually needs saying.",
         "It’s okay to be angry — what matters is what you do with it.",
-        "Drink water and give your body a few minutes to settle.",
-        "Sometimes the kindest thing is to wait before responding.",
         "Ask yourself: ‘What boundary of mine needs care right now?’",
     ],
     "boredom": [
+        "Try something small and new: a song, a stretch, a random fact.",
         "Boredom can be a doorway — something in you wants to wake up.",
-        "Try something completely small and new: a song, a stretch, a random fact.",
-        "Put on music and tidy for 5 minutes — movement often sparks motivation.",
-        "Write a list of ‘tiny adventures’ you could do today.",
-        "Sometimes doing nothing is what your mind needs — rest counts.",
-        "Look outside for a moment — notice one color or sound you hadn’t before.",
-        "Ask yourself: ‘What might make me curious right now?’",
-        "Do one creative thing without judging it — draw, hum, doodle, move.",
+        "Write a list of tiny adventures you could do today.",
+        "Look outside and notice one color or sound you hadn’t before.",
     ],
     "grief": [
         "Grief is love that has nowhere to go — let it speak softly through you.",
         "You’re allowed to miss them and still keep living.",
-        "Light a candle, whisper their name, and breathe.",
         "Eat something, drink water — caring for your body honors their memory too.",
-        "It’s okay to cry about the same thing more than once.",
-        "Talk about your loss. The pain lessens when it’s witnessed.",
-        "Hold an item that reminds you of love, not loss.",
         "Healing doesn’t mean forgetting; it means remembering with more peace.",
     ],
     "love": [
         "Tell someone you care about them — even in a small text.",
         "Give love back to yourself: you deserve your own gentleness.",
         "Love is quiet presence — you don’t have to prove it.",
-        "A small act of kindness can brighten two days at once.",
-        "You are loved simply for being, not for doing.",
-        "Love often grows in silence and simple gestures.",
-        "Send out a silent thank-you to those who care for you.",
         "Let yourself receive love — it’s not selfish to need warmth.",
     ],
     "nervousness": [
@@ -134,40 +112,24 @@ SUGGESTIONS = {
         "Relax your jaw, drop your shoulders, exhale longer than you inhale.",
         "You don’t have to fix every thought; let them pass like clouds.",
         "Try naming five things you can see right now.",
-        "Anxiety is a false alarm — you can remind your brain you’re safe.",
-        "Put both feet flat on the floor and press gently to ground yourself.",
-        "Write down what worries you, then cross out what you can’t control.",
-        "You can pause; you’re not behind. Safety starts with one breath.",
     ],
     "curiosity": [
         "Follow the spark — it doesn’t have to make sense yet.",
         "Ask the question that feels exciting, not the one that feels safe.",
-        "Let yourself explore something simply because it’s interesting.",
         "Learning is self-care too — wonder is nourishment.",
-        "Take 10 minutes to google something random you love.",
-        "Curiosity is courage in disguise — let it lead.",
         "Even small discoveries can wake up your energy.",
-        "Be a beginner; it’s a powerful place to start.",
     ],
     "gratitude": [
         "Name three things you’re grateful for — big or tiny.",
-        "Gratitude turns ordinary moments into something sacred.",
-        "Say ‘thank you’ — out loud, even if just to yourself.",
         "Look around and notice something that quietly supports you.",
         "Each breath is a gift you didn’t have to earn.",
         "Gratitude softens fear; it reminds you what’s still good.",
-        "Send a short thank-you message to someone right now.",
-        "Appreciate how far you’ve already come — quietly, just for you.",
     ],
     "neutral": [
         "Take one slow, deep breath. That’s a start.",
         "Not every moment has to be meaningful — existing is enough.",
         "Sometimes calm feels empty because we’re used to noise — rest in it.",
-        "Stretch for 30 seconds and notice your body waking up.",
         "Drink water; your brain loves that.",
-        "Sit still for one minute — that’s all you need.",
-        "Neutral moments are where balance grows.",
-        "Doing nothing for a while is still doing something.",
     ],
 }
 
@@ -175,50 +137,26 @@ QUOTES = {
     "sadness": [
         "“Even the darkest night will end and the sun will rise.” – Victor Hugo",
         "“You have survived every hard day so far.”",
-        "“You don’t have to feel better to start healing.”",
     ],
     "fear": [
         "“Feel the fear and do it anyway.” – Susan Jeffers",
         "“This moment will not last forever.”",
-        "“You’ve faced hard things before — you can again.”",
     ],
     "joy": [
         "“Happiness is not out there, it’s in you.”",
-        "“Let joy be your rebellion.”",
         "“Enjoy the little things — one day you’ll realize they were the big things.”",
     ],
     "anger": [
         "“Peace begins with a pause.”",
         "“Anger is energy — guide it, don’t suppress it.”",
     ],
-    "boredom": [
-        "“Boredom is the beginning of imagination.” – Jules Renard",
-        "“Curiosity is the cure for boredom.” – Dorothy Parker",
-    ],
-    "grief": [
-        "“Grief is love that has nowhere to go.”",
-        "“Love doesn’t end, it changes form.”",
-    ],
-    "love": [
-        "“Where there is love, there is life.” – Mahatma Gandhi",
-        "“You are loved just for being who you are.” – Ram Dass",
-    ],
-    "nervousness": [
-        "“Breathe. You are doing enough.”",
-        "“This worry does not define you.”",
-    ],
-    "curiosity": [
-        "“Stay curious — it’s the mind’s way of loving life.”",
-        "“Every question plants a seed.”",
-    ],
-    "gratitude": [
-        "“Gratitude turns what we have into enough.”",
-        "“Thankfulness unlocks joy.”",
-    ],
-    "neutral": [
-        "“Be present — even a calm moment can be a quiet victory.”",
-        "“Peace is not the absence of chaos, but the presence of inner calm.”",
-    ],
+    "boredom": ["“Curiosity is the cure for boredom.” – Dorothy Parker"],
+    "grief": ["“Grief is love that has nowhere to go.”"],
+    "love": ["“Where there is love, there is life.” – Mahatma Gandhi"],
+    "nervousness": ["“Breathe. You are doing enough.”"],
+    "curiosity": ["“Stay curious — it’s the mind’s way of loving life.”"],
+    "gratitude": ["“Gratitude turns what we have into enough.”"],
+    "neutral": ["“Be present — even a calm moment can be a quiet victory.”"],
 }
 
 COLOR_MAP = {
@@ -240,106 +178,54 @@ GOEMO_TO_APP = {
     "sadness": "sadness", "surprise": "neutral", "neutral": "neutral",
 }
 
-# --- Threshold & implicit-emotion controls ---
-THRESHOLD = 0.30          # standard selection threshold
-MIN_THRESHOLD = 0.12       # floor if model is unsure
-TOP1_FALLBACK = True       # ensure at least one label if nothing passes threshold
-
-# --- Implicit cues (emojis, slang, negations, intensity) ---
-EMOJI_HINTS = {
-    "😢": "sadness", "😭": "sadness", "😞": "sadness", "💔": "grief",
-    "😡": "anger", "🤬": "anger", "😤": "anger",
-    "😱": "fear", "😨": "fear", "😰": "fear",
-    "😌": "relief", "🙂": "joy", "😊": "joy", "😄": "joy", "😍": "love",
-    "🤗": "love", "💖": "love", "🙏": "gratitude",
-    "💤": "boredom", "🥱": "boredom", "🤔": "curiosity",
-    "😐": "neutral", "😶": "neutral"
-}
-SLANG_HINTS = {
-    "idk": "confusion", "meh": "boredom", "ugh": "annoyance", "nah": "disapproval",
-    "wtf": "anger", "omg": "surprise", "lol": "amusement", "lmao": "amusement",
-    "miss you": "grief", "miss her": "grief", "miss him": "grief",
-    "im fine": "sadness", "i'm fine": "sadness"
-}
-NEGATION_PATTERNS = [
-    ("not happy", "sadness"),
-    ("not okay", "sadness"),
-    ("not ok", "sadness"),
-    ("not fine", "sadness"),
-    ("no hope", "sadness"),
-    ("no energy", "sadness"),
-    ("no motivation", "boredom"),
-    ("not safe", "fear"),
-    ("not calm", "nervousness"),
-    ("not sure", "confusion"),
-]
-INTENSIFIERS = ["!!", "!!!", "?!", "?!?", "soooo", "very", "really", "super", "extremely"]
+# --- Emotion detection parameters ---
+THRESHOLD = 0.30
+MIN_THRESHOLD = 0.12
+TOP1_FALLBACK = True
 
-def _lower(s: str) -> str:
-    return s.lower() if isinstance(s, str) else ""
+# --- Implicit cues ---
+EMOJI_HINTS = {"😢": "sadness", "😭": "sadness", "😡": "anger", "😍": "love", "🤔": "curiosity"}
+SLANG_HINTS = {"idk": "confusion", "meh": "boredom", "ugh": "annoyance", "im fine": "sadness"}
+NEGATION_PATTERNS = [("not happy", "sadness"), ("not ok", "sadness"), ("not calm", "nervousness")]
+INTENSIFIERS = ["!!", "!!!", "really", "so", "very"]
 
 def augment_text_for_classifier(text: str) -> str:
-    """
-    Inject 'hint tokens' based on implicit cues (emojis/slang/negations/intensity)
-    to help TF-IDF pick up emotions not explicitly named.
-    """
-    if not text:
-        return text
-    t = _lower(text)
+    t = text.lower()
     hints = []
-
-    # emojis
     for ch in text:
         if ch in EMOJI_HINTS:
             hints.append(EMOJI_HINTS[ch])
-
-    # slang
     for k, v in SLANG_HINTS.items():
         if k in t:
             hints.append(v)
-
-    # common negations
     for pat, lab in NEGATION_PATTERNS:
         if pat in t:
             hints.append(lab)
-
-    # intensity: duplicate hints to weigh more
     if hints and any(x in t for x in INTENSIFIERS):
-        hints = hints + hints
-
+        hints += hints
     if hints:
-        hint_tokens = " ".join([f"emo_{h}" for h in hints])
-        return text + " " + hint_tokens
+        return text + " " + " ".join([f"emo_{h}" for h in hints])
     return text
 
-# ---------------- SQLite helpers ----------------
+# ---------------- DB helpers ----------------
 def get_conn():
     return sqlite3.connect(DB_PATH, check_same_thread=False, timeout=10)
 
 def init_db():
     conn = get_conn()
     c = conn.cursor()
-    c.execute("""
-    CREATE TABLE IF NOT EXISTS sessions(
+    c.execute("""CREATE TABLE IF NOT EXISTS sessions(
         id INTEGER PRIMARY KEY AUTOINCREMENT,
-        ts TEXT,
-        country TEXT,
-        user_text TEXT,
-        main_emotion TEXT
-    )
-    """)
-    conn.commit()
-    conn.close()
+        ts TEXT, country TEXT, user_text TEXT, main_emotion TEXT)""")
+    conn.commit(); conn.close()
 
 def log_session(country, msg, emotion):
-    conn = get_conn()
-    c = conn.cursor()
-    c.execute("INSERT INTO sessions(ts, country, user_text, main_emotion) VALUES(?,?,?,?)",
+    conn = get_conn(); c = conn.cursor()
+    c.execute("INSERT INTO sessions(ts,country,user_text,main_emotion)VALUES(?,?,?,?)",
               (datetime.utcnow().isoformat(timespec="seconds"), country, msg[:500], emotion))
-    conn.commit()
-    conn.close()
+    conn.commit(); conn.close()
 
-# ---------------- Train / Load model ----------------
+# ---------------- Train / Load ----------------
 def load_goemotions_dataset():
     ds = load_dataset("google-research-datasets/go_emotions", "simplified")
     label_names = ds["train"].features["labels"].feature.names
@@ -347,195 +233,13 @@ def load_goemotions_dataset():
 
 def train_or_load_model():
     if os.path.isfile(MODEL_PATH):
-        print("[MM] Loading cached classifier...")
         bundle = joblib.load(MODEL_PATH)
         if bundle.get("version") == MODEL_VERSION:
             return bundle["pipeline"], bundle["mlb"], bundle["label_names"]
-
-    print("[MM] Loading GoEmotions dataset...")
-    ds, label_names = load_goemotions_dataset()
-
-    X_train = ds["train"]["text"]; y_train = ds["train"]["labels"]
-    X_val = ds["validation"]["text"]; y_val = ds["validation"]["labels"]
-
-    mlb = MultiLabelBinarizer(classes=list(range(len(label_names))))
-    Y_train = mlb.fit_transform(y_train)
-    Y_val = mlb.transform(y_val)
-
+    ds, names = load_goemotions_dataset()
+    X_train, y_train = ds["train"]["text"], ds["train"]["labels"]
+    X_val, y_val = ds["validation"]["text"], ds["validation"]["labels"]
+    mlb = MultiLabelBinarizer(classes=list(range(len(names))))
+    Y_train, Y_val = mlb.fit_transform(y_train), mlb.transform(y_val)
     clf = Pipeline([
-        ("tfidf", TfidfVectorizer(
-            lowercase=True, ngram_range=(1, 2), min_df=2, max_df=0.9, strip_accents="unicode"
-        )),
-        ("ovr", OneVsRestClassifier(
-            LogisticRegression(solver="saga", max_iter=1000, n_jobs=-1, class_weight="balanced"),
-            n_jobs=-1
-        ))
-    ])
-
-    print("[MM] Training classifier...")
-    clf.fit(X_train, Y_train)
-    macro_f1 = f1_score(Y_val, clf.predict(X_val), average="macro", zero_division=0)
-    print(f"[MM] Validation macro F1: {macro_f1:.3f}")
-
-    joblib.dump({"version": MODEL_VERSION, "pipeline": clf, "mlb": mlb, "label_names": label_names}, MODEL_PATH)
-    print(f"[MM] Saved model -> {MODEL_PATH}")
-    return clf, mlb, label_names
-
-try:
-    CLASSIFIER, MLB, LABEL_NAMES = train_or_load_model()
-except Exception as e:
-    print(f"[WARN] Failed to train/load classifier: {e}")
-    CLASSIFIER, MLB, LABEL_NAMES = None, None, None
-
-# ---------------- Emotion detection ----------------
-def classify_text(text: str):
-    """
-    Returns [(label_name, prob), ...] sorted desc.
-    - Augments input with implicit-cue tokens
-    - Uses dynamic threshold with a safety floor
-    - Top-1 fallback ensures at least one label
-    """
-    if not CLASSIFIER or not LABEL_NAMES:
-        return []
-
-    augmented = augment_text_for_classifier(text)
-
-    try:
-        proba = CLASSIFIER.predict_proba([augmented])[0]
-    except AttributeError:
-        from scipy.special import expit
-        scores = CLASSIFIER.decision_function([augmented])[0]
-        proba = expit(scores)
-
-    # Dynamic threshold
-    maxp = float(max(proba)) if len(proba) else 0.0
-    thr = THRESHOLD if maxp >= THRESHOLD else max(MIN_THRESHOLD, maxp * 0.8)
-
-    idxs = [i for i, p in enumerate(proba) if p >= thr]
-    idxs.sort(key=lambda i: proba[i], reverse=True)
-
-    # Top-1 fallback
-    if not idxs and TOP1_FALLBACK and len(proba):
-        top1 = int(max(range(len(proba)), key=lambda i: proba[i]))
-        idxs = [top1]
-
-    return [(LABEL_NAMES[i], float(proba[i])) for i in idxs]
-
-def detect_emotions(text: str):
-    chosen = classify_text(text)
-    if not chosen:
-        return "neutral"
-    bucket = {}
-    for label, p in chosen:
-        app = GOEMO_TO_APP.get(label.lower(), "neutral")
-        bucket[app] = max(bucket.get(app, 0.0), p)
-    return max(bucket, key=bucket.get) if bucket else "neutral"
-
-# ---------------- Reply composer ----------------
-def compose_support(main_emotion: str, is_first_msg: bool) -> str:
-    # Always include an advice tip; 50% chance to add a quote
-    tip = random.choice(SUGGESTIONS.get(
-        main_emotion,
-        ["Take a slow breath. One small act of kindness can shift your day."]
-    ))
-    quote = random.choice(QUOTES.get(
-        main_emotion,
-        ["“No matter what you feel right now, this moment will pass.”"]
-    ))
-    include_quote = random.random() < 0.5
-
-    reply = tip
-    if include_quote:
-        reply += f"\n\n💬 {quote}"
-
-    if is_first_msg:
-        reply += "\n\n*Can you tell me a bit more about what’s behind that feeling?*"
-
-    return reply
-
-# ---------------- Chat logic ----------------
-def crisis_block(country):
-    msg = CRISIS_NUMBERS.get(country, CRISIS_NUMBERS["Other / Not listed"])
-    return ("💛 I'm really sorry you're feeling like this. You matter.\n\n"
-            f"**If you might be in danger or thinking about harming yourself:**\n{msg}\n\n"
-            "Please reach out to someone now. You are not alone.")
-
-def chat_step(message, history, country, save_session):
-    if CRISIS_RE.search(message):
-        return crisis_block(country), "#FFD6E7"
-    if CLOSING_RE.search(message):
-        return ("Thank you 💛 Take care of yourself. Small steps matter. 🌿", "#FFFFFF")
-
-    recent = " ".join(message.split()[-100:])
-    emotion = detect_emotions(recent)
-    color = COLOR_MAP.get(emotion, "#FFFFFF")
-
-    if save_session:
-        log_session(country, message, emotion)
-
-    reply = compose_support(emotion, is_first_msg=not bool(history))
-    return reply, color
-
-# ---------------- Gradio UI ----------------
-init_db()
-
-custom_css = """
-:root, body, .gradio-container { transition: background-color 0.8s ease !important; }
-.typing { font-style: italic; opacity: 0.8; animation: blink 1s infinite; }
-@keyframes blink { 50% {opacity: 0.4;} }
-"""
-
-with gr.Blocks(css=custom_css, title="🪞 MoodMirror+ (Dataset-only Edition)") as demo:
-    style_injector = gr.HTML("")
-    gr.Markdown(
-        "### 🪞 MoodMirror+ — Emotional Support & Inspiration 🌸\n"
-        "Powered only by the **GoEmotions dataset** (trained locally on startup).\n\n"
-        "_Not medical advice. If you feel unsafe, please reach out for help immediately._"
-    )
-
-    with gr.Row():
-        country = gr.Dropdown(choices=list(CRISIS_NUMBERS.keys()),
-                              value="Other / Not listed", label="Country")
-        save_ok = gr.Checkbox(value=False, label="Save anonymized session (no personal data)")
-
-    chat = gr.Chatbot(height=360)
-    msg = gr.Textbox(placeholder="Type how you feel...", label="Your message")
-    send = gr.Button("Send")
-    typing = gr.Markdown("", elem_classes="typing")
-
-    # Optional: dataset preview
-    with gr.Accordion("🔎 Preview GoEmotions samples", open=False):
-        with gr.Row():
-            n_examples = gr.Slider(1, 10, value=5, step=1, label="Number of examples")
-            split = gr.Dropdown(["train", "validation", "test"], value="train", label="Split")
-            refresh = gr.Button("Show samples")
-        table = gr.Dataframe(headers=["text", "labels"], row_count=5, wrap=True)
-
-        def refresh_samples(n, split_name):
-            try:
-                ds = load_dataset("google-research-datasets/go_emotions", "simplified")
-                names = ds["train"].features["labels"].feature.names
-                rows = ds[split_name].shuffle(seed=42).select(range(min(int(n), len(ds[split_name]))))
-                return [[t, ", ".join([names[i] for i in labs])] for t, labs in zip(rows["text"], rows["labels"])]
-            except Exception as e:
-                return [[f"Dataset load error: {e}", ""]]
-
-        refresh.click(refresh_samples, inputs=[n_examples, split], outputs=[table])
-
-    def respond(user_msg, chat_hist, country_choice, save_flag):
-        if not user_msg or not user_msg.strip():
-            yield chat_hist + [[user_msg, "Please share a short sentence about how you feel 🙂"]], "", "", ""
-            return
-        yield chat_hist, "💭 MoodMirror is thinking...", "", ""
-        reply, color = chat_step(user_msg, chat_hist, country_choice, bool(save_flag))
-        style_tag = f"<style>:root,body,.gradio-container{{background:{color}!important;}}</style>"
-        yield chat_hist + [[user_msg, reply]], "", style_tag, ""
-
-    send.click(respond, inputs=[msg, chat, country, save_ok],
-               outputs=[chat, typing, style_injector, msg], queue=True)
-    msg.submit(respond, inputs=[msg, chat, country, save_ok],
-               outputs=[chat, typing, style_injector, msg], queue=True)
-
-if __name__ == "__main__":
-    demo.queue()
-    demo.launch()
+        ("tfidf", TfidfVectorizer(lowercase=True, ngram_range=(1,2), min_df=2, max_df=0.9)),