vrr/WFGY
2
0
Fork 0
mirror of https://github.com/onestardao/WFGY.git synced 2026-04-28 19:50:17 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4
WFGY/ProblemMap/GlobalFixMap/Language/stopword_and_morphology_controls.md

12 KiB
Raw Blame History

Stopword and Morphology Controls · Global Fix Map

🧭 Quick Return to Map

You are in a sub-page of Language.
To reorient, go back here:

Think of this page as a desk within a ward.
If you need the full triage and all prescriptions, return to the Emergency Room lobby.

Lock language specific stopwords and morphology so retrieval remains stable across scripts and locales. Protect entities from stopword removal, version your lemmatizer, and verify with ΔS, λ, and coverage targets.

Open these first

Core acceptance targets

  • ΔS(question, retrieved) ≤ 0.45 on three paraphrases and two seeds
  • Coverage of target section ≥ 0.70
  • λ convergent when switching between inflected forms and lemmas
  • Removing stopwords never drops a required entity or negation token
  • Rank@k does not regress more than 2 points after morphology changes

What usually breaks

Symptom Likely cause Open this
Named entities vanish after preprocessing stopword list removes particles that are part of names proper_noun_aliases.md
Wrong meaning for negated statements stopword filter removes “not” class tokens retrieval-traceability.md
Duplicate docs across lemma and surface form inconsistent stemming across index and query tokenizer_mismatch.md
CJK recall collapses after adding stopwords imported Latin stopword list applied to CJK script_mixing.md
Turkish i and ı behave inconsistently locale fold differs across stages locale_drift.md
Romanized queries fail while native works alias view absent and morphology applied only to native romanization_transliteration.md

Language family controls

Latin and Germanic

  • Case fold with locale aware rules.
  • Use light stemming or lemmatization, not both.
  • Keep negation terms out of stopwords. Maintain a protected list.

Romance

  • Normalize accents consistently.
  • Prefer lemmatization for verbs and nouns.
  • Maintain a protected list for names with articles or particles.

CJK

  • Do not apply generic stopword lists.
  • Use bigram or language specific tokenizers.
  • Keep entities in dedicated fields without stopword removal.

Semitic RTL

  • Normalize diacritics and width consistently.
  • Use lemmatization that preserves roots only if evaluation passes ΔS and coverage.
  • Keep a protected list for clitics that change meaning.

Indic

  • Avoid generic lists from other languages.
  • Use language specific analyzers and verify with bilingual eval.
  • Protect named entities that share forms with common words.

Cyrillic and Greek

  • Apply accent and width normalization.
  • Prefer lemmatization over aggressive stemming.
  • Maintain a protected entity list for inflected forms.

Deterministic pipeline checklist

  1. Version every component: stoplist_v, stemmer_v, lemma_v, normalize_v.
  2. Define no stop zones for entity fields and citation fields.
  3. Keep words filter with a protected list per language code.
  4. Apply normalization before morphology, not after.
  5. Use the same pipeline for indexing and querying.
  6. Log a morphology fingerprint in traces and eval reports.

Copy snippets

A. Protected term filter sketch

{
  "analysis": {
    "filter": {
      "keep_entities_en": {
        "type": "keep",
        "keep_words": ["New York", "AT&T", "Côte d'Ivoire", "Íñigo", "İstanbul"]
      }
    }
  }
}

B. Minimal morphology config record

{
  "language": "tr",
  "stoplist_v": "tr_core_1.2",
  "lemma_v": "tr_lemma_0.9",
  "normalize_v": "nfkc_fold_tr",
  "no_stop_fields": ["title_exact", "entity_exact"],
  "protected_list_hash": "sha256:..."
}

C. Trace fields to log

{
  "ΔS": 0.42,
  "λ_state": "<>",
  "coverage": 0.74,
  "language": "ar",
  "morph_fingerprint": "stop:ar_1.1|lemma:ar_0.8|norm:nfkc_1.0|keep:hash"
}

Eval protocol

  • Use bilingual and code switching sets from code_switching_eval.md.
  • For each query, test with and without stopword removal and with two morphology settings.
  • Accept only if ΔS ≤ 0.45, coverage ≥ 0.70, λ convergent, and no loss of entity recall.
  • Report Rank@k deltas for lemma vs surface forms.

When to escalate

Copy paste prompt for the LLM step

You have TXTOS and the WFGY Problem Map loaded.

Task:
1) For {lang, script}, choose stopword and morphology settings that protect entities and negation.
2) Run cite then explain. If ΔS(question, retrieved) ≥ 0.60, propose the minimal structural fix.
3) Return JSON:
{ "stoplist_v": "...", "lemma_or_stem": "lemma|stem|none", "protected_terms": [...], "ΔS": 0.xx, "coverage": 0.xx, "λ_state": "→|←|<>|×" }
Keep it auditable and short.

🔗 Quick-Start Downloads (60 sec)

Tool Link 3-Step Setup
WFGY 1.0 PDF Engine Paper 1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + <your question>”
TXT OS (plain-text OS) TXTOS.txt 1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly

Explore More

Layer Page What it’s for
⭐ Proof WFGY Recognition Map External citations, integrations, and ecosystem proof
⚙️ Engine WFGY 1.0 Original PDF tension engine and early logic sketch (legacy reference)
⚙️ Engine WFGY 2.0 Production tension kernel for RAG and agent systems
⚙️ Engine WFGY 3.0 TXT based Singularity tension engine (131 S class set)
🗺️ Map Problem Map 1.0 Flagship 16 problem RAG failure taxonomy and fix map
🗺️ Map Problem Map 2.0 Global Debug Card for RAG and agent pipeline diagnosis
🗺️ Map Problem Map 3.0 Global AI troubleshooting atlas and failure pattern map
🧰 App TXT OS .txt semantic OS with fast bootstrap
🧰 App Blah Blah Blah Abstract and paradox Q&A built on TXT OS
🧰 App Blur Blur Blur Text to image generation with semantic control
🏡 Onboarding Starter Village Guided entry point for new users

If this repository helped, starring it improves discovery so more builders can find the docs and tools.
GitHub Repo stars