eq-layer 0.1.0__py3-none-any.whl

eq_layer/cli.py

@@ -0,0 +1,174 @@
+"""Command-line interface for the EQ-Layer control pass."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from dataclasses import asdict
+
+
+def _load_messages(raw: str) -> list[dict]:
+    payload = json.loads(raw)
+    if isinstance(payload, dict):
+        payload = payload.get("messages")
+    if not isinstance(payload, list) or not payload:
+        raise ValueError("Input must be a non-empty JSON list or {messages:[...]}.")
+    messages: list[dict] = []
+    for index, item in enumerate(payload):
+        if not isinstance(item, dict):
+            raise ValueError(f"message {index} must be an object")
+        role = str(item.get("role") or "")
+        content = str(item.get("content") or "")
+        if role not in {"user", "assistant", "system"}:
+            raise ValueError(f"message {index} has unsupported role {role!r}")
+        if not content.strip():
+            raise ValueError(f"message {index} has empty content")
+        messages.append({"role": role, "content": content})
+    if not any(message["role"] == "user" for message in messages):
+        raise ValueError("Transcript must contain at least one user turn.")
+    return messages
+
+
+def _intent_adapter(mode: str):
+    from .intent import HeuristicIntent
+    from .trained_intent import TrainedIntent
+
+    if mode == "heuristic":
+        return HeuristicIntent(), "heuristic"
+    try:
+        return TrainedIntent.from_bundled(), "trained-tfidf-logreg"
+    except ImportError:
+        if mode == "learned":
+            raise
+        return HeuristicIntent(), "heuristic-fallback-no-ml-extra"
+
+
+def _lightweight_route(messages: list[dict], intent_mode: str) -> dict:
+    from .affect import HeuristicAffect
+    from .dialogue_reference import infer_dialogue_reference
+    from .interaction import infer_interaction_quality, infer_repair_state
+    from .policies import Selector
+    from .steer import Steer
+
+    state = HeuristicAffect().infer(messages, turn_index=len(messages))
+    intent_adapter, intent_backend = _intent_adapter(intent_mode)
+    intent = intent_adapter.infer(messages)
+    repair = infer_repair_state(messages)
+    interaction_quality = infer_interaction_quality(messages)
+    reference = infer_dialogue_reference(messages)
+    selection = Selector().select(state, intent)
+    steer = Steer.build(
+        selection.policy,
+        intent,
+        repair=repair,
+        interaction_quality=interaction_quality,
+        reference=reference,
+    )
+    return {
+        "runtime_mode": "lightweight-local",
+        "intent_backend": intent_backend,
+        "policy": {
+            "name": selection.policy.name,
+            "register": selection.policy.register.value,
+        },
+        "affect": asdict(state),
+        "intent": asdict(intent),
+        "repair": asdict(repair),
+        "interaction_quality": asdict(interaction_quality),
+        "reference": asdict(reference),
+        "factored_action": steer.action.to_dict() if steer.action else None,
+        "system_instruction": steer.system_instruction(),
+        "claim_boundary": (
+            "Lightweight routing is a portable control pass, not the full learned "
+            "research pipeline and not proof of emotional intelligence."
+        ),
+    }
+
+
+def _full_route(messages: list[dict], affect_model: str, subtext_model: str) -> dict:
+    from .response_experiment import FullEQPipeline
+
+    pipeline = FullEQPipeline.load(affect_model, subtext_model)
+    steered, metadata = pipeline.steer_messages(messages)
+    instruction = (
+        steered[0]["content"]
+        if steered and steered[0].get("role") == "system"
+        else ""
+    )
+    return {
+        "runtime_mode": "full-learned",
+        "intent_backend": "trained-tfidf-logreg",
+        "metadata": metadata,
+        "factored_action": metadata.get("factored_action"),
+        "system_instruction": instruction,
+        "claim_boundary": (
+            "Full learned routing uses trained EQ-Layer components; response-quality "
+            "improvement still requires empirical evaluation."
+        ),
+    }
+
+
+def _route(args: argparse.Namespace) -> int:
+    raw = args.messages_json if args.messages_json is not None else sys.stdin.read()
+    if not raw.strip():
+        raise ValueError("Provide transcript JSON with --messages-json or stdin")
+    messages = _load_messages(raw)
+    if bool(args.affect_model) != bool(args.subtext_model):
+        raise ValueError("--affect-model and --subtext-model must be supplied together")
+    if args.affect_model:
+        result = _full_route(messages, args.affect_model, args.subtext_model)
+    else:
+        result = _lightweight_route(messages, args.intent_mode)
+    print(json.dumps(result, ensure_ascii=False, indent=2 if args.pretty else None))
+    return 0
+
+
+def _doctor() -> int:
+    payload = {
+        "package": "eq-layer",
+        "version": "0.1.0",
+        "python": sys.version.split()[0],
+        "ml_extra_available": True,
+    }
+    try:
+        import sklearn  # noqa: F401
+    except ImportError:
+        payload["ml_extra_available"] = False
+    print(json.dumps(payload, indent=2))
+    return 0
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(prog="eq-layer")
+    sub = parser.add_subparsers(dest="command", required=True)
+
+    route = sub.add_parser("route", help="Emit an inspectable EQ-Layer control decision.")
+    route.add_argument("--messages-json")
+    route.add_argument(
+        "--intent-mode",
+        choices=("auto", "learned", "heuristic"),
+        default="auto",
+    )
+    route.add_argument("--affect-model")
+    route.add_argument("--subtext-model")
+    route.add_argument("--pretty", action="store_true")
+    route.set_defaults(handler=_route)
+
+    doctor = sub.add_parser("doctor", help="Report local EQ-Layer runtime availability.")
+    doctor.set_defaults(handler=lambda _args: _doctor())
+    return parser
+
+
+def main(argv: list[str] | None = None) -> int:
+    parser = build_parser()
+    args = parser.parse_args(argv)
+    try:
+        return int(args.handler(args))
+    except (ValueError, json.JSONDecodeError, ImportError) as exc:
+        parser.error(str(exc))
+    return 2
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

eq_layer/coarse_discourse.py

@@ -0,0 +1,107 @@
+from __future__ import annotations
+
+import hashlib
+import io
+import json
+import urllib.request
+import zipfile
+from collections import Counter
+from dataclasses import dataclass
+
+COARSE_DISCOURSE_URL = (
+    "https://zissou.infosci.cornell.edu/convokit/datasets/"
+    "reddit-coarse-discourse-corpus/reddit-coarse-discourse-corpus.zip"
+)
+
+
+@dataclass(frozen=True)
+class CoarseDiscourseExample:
+    utterance_id: str
+    text: str
+    label: str
+    conversation_id: str
+    reply_to: str | None
+    context_text: str = ""
+
+    @property
+    def model_text(self) -> str:
+        if self.context_text:
+            return f"{self.context_text}\n[REPLY]\n{self.text}"
+        return self.text
+
+
+def download_archive() -> bytes:
+    request = urllib.request.Request(
+        COARSE_DISCOURSE_URL,
+        headers={"User-Agent": "eq-layer-coarse-discourse/0.1"},
+    )
+    with urllib.request.urlopen(request, timeout=120) as response:
+        return response.read()
+
+
+def sha256_bytes(data: bytes) -> str:
+    return hashlib.sha256(data).hexdigest()
+
+
+def _find_member(names: list[str], suffix: str) -> str:
+    matches = [name for name in names if name.endswith(suffix)]
+    if len(matches) != 1:
+        raise ValueError(f"Expected one {suffix} member, found {matches}")
+    return matches[0]
+
+
+def load_examples(data: bytes) -> tuple[list[CoarseDiscourseExample], list[str]]:
+    with zipfile.ZipFile(io.BytesIO(data)) as archive:
+        names = archive.namelist()
+        utterances_member = _find_member(names, "utterances.jsonl")
+        examples: list[CoarseDiscourseExample] = []
+
+        rows = []
+        with archive.open(utterances_member) as fh:
+            for raw in fh:
+                rows.append(json.loads(raw.decode("utf-8")))
+
+        text_by_id = {
+            str(row.get("id") or ""): str(row.get("text") or "").strip()
+            for row in rows
+            if row.get("id") is not None
+        }
+
+        for row in rows:
+            text = str(row.get("text") or "").strip()
+            meta = row.get("meta") or {}
+            label = str(meta.get("majority_type") or "").strip().lower()
+            if not text or not label:
+                continue
+
+            utterance_id = str(row.get("id") or "")
+            conversation_id = str(
+                row.get("conversation_id")
+                or row.get("root")
+                or utterance_id
+            )
+            reply_raw = row.get("reply_to")
+            if reply_raw is None:
+                reply_raw = row.get("reply-to")
+            reply_to = str(reply_raw) if reply_raw is not None else None
+
+            examples.append(
+                CoarseDiscourseExample(
+                    utterance_id=utterance_id,
+                    text=text,
+                    label=label,
+                    conversation_id=conversation_id,
+                    reply_to=reply_to,
+                    context_text=text_by_id.get(reply_to or "", ""),
+                )
+            )
+        return examples, names
+
+
+def inventory(examples: list[CoarseDiscourseExample]) -> dict:
+    counts = Counter(example.label for example in examples)
+    return {
+        "examples": len(examples),
+        "conversations": len({x.conversation_id for x in examples if x.conversation_id}),
+        "labels": dict(sorted(counts.items())),
+    }

eq_layer/data/intent_train.jsonl

@@ -0,0 +1,299 @@
+{"text":"Bunu depoya ekle","label":"action_request"}
+{"text":"Dosyayı güncelle","label":"action_request"}
+{"text":"Şu hatayı düzelt","label":"action_request"}
+{"text":"Kodun testlerini çalıştır","label":"action_request"}
+{"text":"Bu metni kısalt","label":"action_request"}
+{"text":"GitHub'a gönder","label":"action_request"}
+{"text":"Devam et ve sonucu yaz","label":"action_request"}
+{"text":"Bu bölümü sil","label":"action_request"}
+{"text":"Raporu oluştur","label":"action_request"}
+{"text":"Kaynakları araştır","label":"action_request"}
+{"text":"Belgeyi incele","label":"action_request"}
+{"text":"Şunu kontrol et","label":"action_request"}
+{"text":"Make this change","label":"action_request"}
+{"text":"Update the file","label":"action_request"}
+{"text":"Run the tests","label":"action_request"}
+{"text":"Add this to the repository","label":"action_request"}
+{"text":"Fix the bug","label":"action_request"}
+{"text":"Create the report","label":"action_request"}
+{"text":"Delete this section","label":"action_request"}
+{"text":"Check the document","label":"action_request"}
+{"text":"Benchmarkı koştur","label":"action_request"}
+{"text":"Testleri koştur","label":"action_request"}
+{"text":"Fonksiyonu yeniden yaz","label":"action_request"}
+{"text":"README'yi değiştir","label":"action_request"}
+{"text":"Bu kısmı yeniden düzenle","label":"action_request"}
+{"text":"Run the benchmark","label":"action_request"}
+{"text":"Rewrite this function","label":"action_request"}
+{"text":"Please run the checks","label":"action_request"}
+{"text":"Edit the README","label":"action_request"}
+{"text":"Depodaki lisans dosyasını yenile","label":"action_request"}
+{"text":"Lisans dosyasını güncelle","label":"action_request"}
+{"text":"Son çıktıyı tabloya ekle","label":"action_request"}
+{"text":"Tabloya yeni satırı ekle","label":"action_request"}
+{"text":"Benchmark koşusunu başlat","label":"action_request"}
+{"text":"Koşuyu başlat","label":"action_request"}
+{"text":"Bu değişiklikleri commit et","label":"action_request"}
+{"text":"Değişiklikleri kaydet","label":"action_request"}
+{"text":"Tekrarlanan bölümü kaldır","label":"action_request"}
+{"text":"Dosyayı yeniden adlandır","label":"action_request"}
+{"text":"Bu paragrafı İngilizceye çevir","label":"action_request"}
+{"text":"Grafiği üret","label":"action_request"}
+{"text":"Sonucu karşılaştır","label":"action_request"}
+{"text":"Bana kısa bir özet hazırla","label":"action_request"}
+{"text":"Bunu şimdi yapabilir misin","label":"action_request"}
+{"text":"Şunu düzeltir misin","label":"action_request"}
+{"text":"Bunu kontrol eder misin","label":"action_request"}
+{"text":"Dosyayı açıp bak","label":"action_request"}
+{"text":"Ana dala gönder","label":"action_request"}
+{"text":"Please remove the duplicate section","label":"action_request"}
+{"text":"Commit these changes","label":"action_request"}
+{"text":"Start the benchmark run","label":"action_request"}
+{"text":"Add the latest output to the table","label":"action_request"}
+{"text":"Rename the file","label":"action_request"}
+{"text":"Translate this paragraph to English","label":"action_request"}
+{"text":"Generate the chart","label":"action_request"}
+{"text":"Compare the results","label":"action_request"}
+{"text":"Prepare a short summary","label":"action_request"}
+{"text":"Could you fix this","label":"action_request"}
+{"text":"Please check this","label":"action_request"}
+{"text":"Open the file and inspect it","label":"action_request"}
+{"text":"Push this to main","label":"action_request"}
+{"text":"Başarısız testi yeniden çalıştır","label":"action_request"}
+{"text":"Başarısız job'u tekrar çalıştır","label":"action_request"}
+{"text":"Please rerun the failed test","label":"action_request"}
+{"text":"Rerun the job","label":"action_request"}
+{"text":"Durum ne","label":"status_check"}
+{"text":"Son durum nedir","label":"status_check"}
+{"text":"Ne kadar kaldı","label":"status_check"}
+{"text":"Bitti mi","label":"status_check"}
+{"text":"Şu an çalışıyor mu","label":"status_check"}
+{"text":"İşlem ilerliyor mu","label":"status_check"}
+{"text":"Koşu tamamlandı mı","label":"status_check"}
+{"text":"Workflow ne durumda","label":"status_check"}
+{"text":"İndirme bitti mi","label":"status_check"}
+{"text":"Şimdi ne oldu","label":"status_check"}
+{"text":"Hesaplama devam ediyor mu","label":"status_check"}
+{"text":"Kaç dosya kaldı","label":"status_check"}
+{"text":"What's the status","label":"status_check"}
+{"text":"Is it finished","label":"status_check"}
+{"text":"How much is left","label":"status_check"}
+{"text":"Is the workflow still running","label":"status_check"}
+{"text":"Did the download finish","label":"status_check"}
+{"text":"Where are we now","label":"status_check"}
+{"text":"CI bitti mi","label":"status_check"}
+{"text":"Kaç görev kaldı","label":"status_check"}
+{"text":"Kaç iş kaldı","label":"status_check"}
+{"text":"Ne aşamada","label":"status_check"}
+{"text":"Workflow tamamlandı mı","label":"status_check"}
+{"text":"Koşu bitti mi","label":"status_check"}
+{"text":"Is the CI finished","label":"status_check"}
+{"text":"How many tasks are left","label":"status_check"}
+{"text":"How many jobs remain","label":"status_check"}
+{"text":"What stage are we at","label":"status_check"}
+{"text":"Is the run done","label":"status_check"}
+{"text":"Runner hâlâ aktif mi","label":"status_check"}
+{"text":"Şimdi kaç tane kaldı","label":"status_check"}
+{"text":"İşlem tamamlandı mı yoksa sürüyor mu","label":"status_check"}
+{"text":"Koşu hâlâ devam ediyor mu","label":"status_check"}
+{"text":"Runner kapandı mı","label":"status_check"}
+{"text":"Şu an hangi aşamada","label":"status_check"}
+{"text":"Geriye ne kaldı","label":"status_check"}
+{"text":"İş bitti mi yoksa devam mı ediyor","label":"status_check"}
+{"text":"Ne kadarı tamamlandı","label":"status_check"}
+{"text":"Yüzde kaç bitti","label":"status_check"}
+{"text":"Workflow sonuçlandı mı","label":"status_check"}
+{"text":"Son job tamamlandı mı","label":"status_check"}
+{"text":"Has the job completed","label":"status_check"}
+{"text":"Is the runner still active","label":"status_check"}
+{"text":"What remains to be done","label":"status_check"}
+{"text":"Is the process still running","label":"status_check"}
+{"text":"How far along is it","label":"status_check"}
+{"text":"What percentage is complete","label":"status_check"}
+{"text":"Has the workflow completed","label":"status_check"}
+{"text":"Is the last job done","label":"status_check"}
+{"text":"Are we finished yet","label":"status_check"}
+{"text":"How many are still pending","label":"status_check"}
+{"text":"Bu neden oluyor","label":"explanation"}
+{"text":"Nasıl çalışıyor","label":"explanation"}
+{"text":"Bunu açıkla","label":"explanation"}
+{"text":"Bunun sebebi ne","label":"explanation"}
+{"text":"Aradaki farkı anlat","label":"explanation"}
+{"text":"Neden bu sonucu verdi","label":"explanation"}
+{"text":"Bu kavram ne demek","label":"explanation"}
+{"text":"Nasıl hesaplanıyor","label":"explanation"}
+{"text":"Bana mantığını anlat","label":"explanation"}
+{"text":"Niçin böyle seçtin","label":"explanation"}
+{"text":"Bu yöntemi açıklar mısın","label":"explanation"}
+{"text":"Neden başarısız oldu","label":"explanation"}
+{"text":"Why does this happen","label":"explanation"}
+{"text":"Explain how this works","label":"explanation"}
+{"text":"What does this mean","label":"explanation"}
+{"text":"Why did it fail","label":"explanation"}
+{"text":"Explain the difference","label":"explanation"}
+{"text":"How is this calculated","label":"explanation"}
+{"text":"Niye böyle davranıyor","label":"explanation"}
+{"text":"Bu neden böyle davranıyor","label":"explanation"}
+{"text":"Bunun mantığı nedir","label":"explanation"}
+{"text":"Bu denklemi anlat","label":"explanation"}
+{"text":"Why is it behaving this way","label":"explanation"}
+{"text":"Why is the model doing this","label":"explanation"}
+{"text":"Explain this equation","label":"explanation"}
+{"text":"Bu seçimin gerekçesi ne","label":"explanation"}
+{"text":"Neye göre bunu seçtin","label":"explanation"}
+{"text":"Bu mekanizmayı bana açıklar mısın","label":"explanation"}
+{"text":"Bu neden gerekli","label":"explanation"}
+{"text":"Bu karar nasıl veriliyor","label":"explanation"}
+{"text":"Bunun arkasındaki mantık ne","label":"explanation"}
+{"text":"Bu hata neden çıkıyor","label":"explanation"}
+{"text":"Bu formül nasıl türetiliyor","label":"explanation"}
+{"text":"Bu sonuç nereden geliyor","label":"explanation"}
+{"text":"Neden bu policy seçildi","label":"explanation"}
+{"text":"Bu iki sonuç neden farklı","label":"explanation"}
+{"text":"Why was this policy selected","label":"explanation"}
+{"text":"Walk me through how this mechanism works","label":"explanation"}
+{"text":"Why is this necessary","label":"explanation"}
+{"text":"How is this decision made","label":"explanation"}
+{"text":"What is the reasoning behind this","label":"explanation"}
+{"text":"Why does this error occur","label":"explanation"}
+{"text":"How is this formula derived","label":"explanation"}
+{"text":"Where does this result come from","label":"explanation"}
+{"text":"Why are these results different","label":"explanation"}
+{"text":"Python 3.12 destekliyor mu","label":"question"}
+{"text":"Bu dosyanın boyutu kaç","label":"question"}
+{"text":"Hangi sürüm daha yeni","label":"question"}
+{"text":"Kaç tane test var","label":"question"}
+{"text":"Modelin bağlam sınırı nedir","label":"question"}
+{"text":"Bu paket ücretsiz mi","label":"question"}
+{"text":"Lisansı MIT mi","label":"question"}
+{"text":"Hangi dosya ana dosya","label":"question"}
+{"text":"Son commit hangi tarihte","label":"question"}
+{"text":"Bu özellik var mı","label":"question"}
+{"text":"İki yöntemden hangisi daha hızlı","label":"question"}
+{"text":"Repo public mi","label":"question"}
+{"text":"Does it support Python 3.12","label":"question"}
+{"text":"How many tests are there","label":"question"}
+{"text":"Which version is newer","label":"question"}
+{"text":"Is this package free","label":"question"}
+{"text":"What is the context limit","label":"question"}
+{"text":"Is the repository public","label":"question"}
+{"text":"Paketin minimum Python sürümü kaç","label":"question"}
+{"text":"Bu dosyada kaç satır var","label":"question"}
+{"text":"Model açık kaynak mı","label":"question"}
+{"text":"Hangi dosya giriş noktası","label":"question"}
+{"text":"GPU gerekiyor mu","label":"question"}
+{"text":"Bu API ücretli mi","label":"question"}
+{"text":"Bu seçenek varsayılan mı","label":"question"}
+{"text":"Hangi lisans kullanılıyor","label":"question"}
+{"text":"Bu dosya JSON mı","label":"question"}
+{"text":"Repo kaç commit içeriyor","label":"question"}
+{"text":"Bu özellik Windows'ta çalışıyor mu","label":"question"}
+{"text":"Sürüm numarası ne","label":"question"}
+{"text":"Which file is the entry point","label":"question"}
+{"text":"Does this require a GPU","label":"question"}
+{"text":"Is the model open source","label":"question"}
+{"text":"What is the minimum Python version","label":"question"}
+{"text":"How many lines are in this file","label":"question"}
+{"text":"Is this API paid","label":"question"}
+{"text":"Is this the default option","label":"question"}
+{"text":"Which license is used","label":"question"}
+{"text":"Is this file JSON","label":"question"}
+{"text":"How many commits are in the repo","label":"question"}
+{"text":"Does this work on Windows","label":"question"}
+{"text":"What is the version number","label":"question"}
+{"text":"Randevumu üç kez değiştirdiler","label":"statement"}
+{"text":"Bugün biraz yorgunum","label":"statement"}
+{"text":"Bu sonuç beklediğimden farklı","label":"statement"}
+{"text":"Dosya artık burada","label":"statement"}
+{"text":"Yeni sürüm daha temiz görünüyor","label":"statement"}
+{"text":"Bu yaklaşım bana mantıklı geliyor","label":"statement"}
+{"text":"Sunumu yarına erteledim","label":"statement"}
+{"text":"Model bu örnekte iyi cevap verdi","label":"statement"}
+{"text":"Kod şu anda main dalında","label":"statement"}
+{"text":"Veri seti küçük","label":"statement"}
+{"text":"Bu fikir ilginç","label":"statement"}
+{"text":"Testlerin çoğu geçti","label":"statement"}
+{"text":"The result is different from what I expected","label":"statement"}
+{"text":"The file is here now","label":"statement"}
+{"text":"The dataset is small","label":"statement"}
+{"text":"This approach looks reasonable","label":"statement"}
+{"text":"The model handled this example well","label":"statement"}
+{"text":"Most tests passed","label":"statement"}
+{"text":"Yeni yapı daha sade","label":"statement"}
+{"text":"Yeni sürüm daha basit","label":"statement"}
+{"text":"Workflow bugün hata verdi","label":"statement"}
+{"text":"CI bugün hata verdi","label":"statement"}
+{"text":"The new structure is simpler","label":"statement"}
+{"text":"The workflow failed yesterday","label":"statement"}
+{"text":"CI failed today","label":"statement"}
+{"text":"The build failed today","label":"statement"}
+{"text":"Bu sonuç önceki koşudan daha iyi","label":"statement"}
+{"text":"Bugün veri indirme yavaşladı","label":"statement"}
+{"text":"Yeni mimari daha okunabilir görünüyor","label":"statement"}
+{"text":"Son test beklediğimden kötü çıktı","label":"statement"}
+{"text":"Repo artık public","label":"statement"}
+{"text":"Bu dosya dün değişti","label":"statement"}
+{"text":"Model Türkçede daha iyi görünüyor","label":"statement"}
+{"text":"Çalışma bugün hata verdi","label":"statement"}
+{"text":"İndirme biraz yavaş","label":"statement"}
+{"text":"Yeni çıktı daha kısa","label":"statement"}
+{"text":"Bu yaklaşım riskli olabilir","label":"statement"}
+{"text":"The latest result looks better","label":"statement"}
+{"text":"The dataset was updated yesterday","label":"statement"}
+{"text":"The new architecture is more readable","label":"statement"}
+{"text":"The last test performed worse than expected","label":"statement"}
+{"text":"The repository is public now","label":"statement"}
+{"text":"The file changed yesterday","label":"statement"}
+{"text":"The model looks better in Turkish","label":"statement"}
+{"text":"The run failed today","label":"statement"}
+{"text":"The download is slow","label":"statement"}
+{"text":"The new output is shorter","label":"statement"}
+{"text":"This approach may be risky","label":"statement"}
+{"text":"Son commit sadece dokümantasyonu değiştirdi","label":"statement"}
+{"text":"Commit sadece belgeleri değiştirdi","label":"statement"}
+{"text":"The latest commit only changes documentation","label":"statement"}
+{"text":"The commit only updates docs","label":"statement"}
+{"text":"Bunu","label":"unknown"}
+{"text":"Şunu","label":"unknown"}
+{"text":"Onu","label":"unknown"}
+{"text":"Bu şey","label":"unknown"}
+{"text":"Şey","label":"unknown"}
+{"text":"Buna mı","label":"unknown"}
+{"text":"Şu mu","label":"unknown"}
+{"text":"Onu da","label":"unknown"}
+{"text":"Bununla","label":"unknown"}
+{"text":"Şundan","label":"unknown"}
+{"text":"Oradaki şey","label":"unknown"}
+{"text":"Hani o","label":"unknown"}
+{"text":"That one","label":"unknown"}
+{"text":"This","label":"unknown"}
+{"text":"It","label":"unknown"}
+{"text":"That thing","label":"unknown"}
+{"text":"The other one","label":"unknown"}
+{"text":"This one","label":"unknown"}
+{"text":"Diğeri","label":"unknown"}
+{"text":"Öteki","label":"unknown"}
+{"text":"O şey","label":"unknown"}
+{"text":"Buna","label":"unknown"}
+{"text":"The other","label":"unknown"}
+{"text":"That other one","label":"unknown"}
+{"text":"Şu","label":"unknown"}
+{"text":"Ondan","label":"unknown"}
+{"text":"Şuradaki","label":"unknown"}
+{"text":"Bununki","label":"unknown"}
+{"text":"Öbürü","label":"unknown"}
+{"text":"Oradaki","label":"unknown"}
+{"text":"O taraftaki","label":"unknown"}
+{"text":"Şu olan","label":"unknown"}
+{"text":"Bununla ilgili olan","label":"unknown"}
+{"text":"Önceki şey","label":"unknown"}
+{"text":"Hani şu","label":"unknown"}
+{"text":"Diğer olan","label":"unknown"}
+{"text":"That one there","label":"unknown"}
+{"text":"The thing over there","label":"unknown"}
+{"text":"The previous one","label":"unknown"}
+{"text":"The other thing","label":"unknown"}
+{"text":"The one over there","label":"unknown"}
+{"text":"This thing","label":"unknown"}
+{"text":"That thing there","label":"unknown"}
+{"text":"The other one there","label":"unknown"}