turkish-linguistic-features 0.2.0__tar.gz

turkish_linguistic_features-0.2.0/.github/workflows/ci.yml

@@ -0,0 +1,90 @@
+# CI: her push'ta ve her pull request'te lint + tip denetimi + testler.
+# GitHub bu dosyayı `.github/workflows/` altında bulunca kendiliğinden çalıştırır.
+
+name: CI                                  # Actions sekmesinde ve rozette görünen ad
+
+on:                                       # ne zaman çalışsın
+  push:
+    branches: [main]                      # main'e her push
+  pull_request:                           # her pull request (hedef dal fark etmez)
+
+permissions:
+  contents: read                          # yalnız kodu okur; depoya yazamaz
+
+concurrency:                              # aynı dala art arda push'ta eskisini iptal et
+  group: ci-${{ github.ref }}
+  cancel-in-progress: true
+
+defaults:
+  run:
+    shell: bash                           # Windows'ta da aynı komut sözdizimi (Git Bash)
+
+jobs:
+  # ── İş 1: kodu çalıştırmadan denetle ─────────────────────────────────
+  lint:
+    runs-on: ubuntu-latest                # GitHub'ın verdiği temiz bir Linux makinesi
+    steps:
+      - uses: actions/checkout@v7         # depoyu makineye indir
+      - uses: actions/setup-python@v7     # Python'u kur
+        with:
+          python-version: "3.12"
+          cache: pip                      # pip indirmelerini koşular arası sakla
+          cache-dependency-path: |        # bu dosyalar değişince önbellek yenilenir
+            requirements.txt
+            pyproject.toml
+      - name: Install dependencies
+        run: pip install -r requirements.txt   # ruff ve mypy de burada
+      - name: Ruff
+        run: ruff check .
+      - name: Mypy
+        run: mypy                         # ayarı pyproject.toml [tool.mypy]
+
+  # ── İş 2: testler, birden çok Python sürümü ve işletim sisteminde ─────
+  test:
+    name: test (${{ matrix.os }}, py${{ matrix.python-version }})
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false                    # biri kırılsa da ötekiler sonuna kadar koşsun
+      matrix:                             # aşağıdaki her birleşim ayrı bir makinede
+        os: [ubuntu-latest]
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+        include:                          # + Windows: zeyrek/spaCy import sırası Windows'a özgü
+          - os: windows-latest
+            python-version: "3.11"
+    env:
+      # Dil verisi eksikse testler ATLANMASIN, başarısız olsun (tests/conftest.py).
+      # Kurulum adımı sessizce bozulursa CI yeşil "skipped" ile geçmesin.
+      TLF_REQUIRE_MODELS: "1"
+      TR_MODEL_URL: https://huggingface.co/turkish-nlp-suite/tr_core_news_md/resolve/main/tr_core_news_md-1.0-py3-none-any.whl
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-python@v7
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+          cache-dependency-path: |
+            requirements.txt
+            pyproject.toml
+
+      - name: Install dependencies
+        run: pip install -r requirements.txt
+
+      # Türkçe model 156 MB ve Hugging Face'ten geliyor: wheel dosyası bir kez
+      # indirilip önbellekte tutulur. Anahtar model sürümüne bağlı: TR_MODEL_URL'deki
+      # sürüm değişirse anahtarı da değiştir, yoksa eski wheel kullanılır.
+      - name: Cache Turkish model wheel
+        uses: actions/cache@v6
+        with:
+          path: ~/tlf-models
+          key: tr-core-news-md-1.0
+      - name: Install language data
+        run: |
+          mkdir -p ~/tlf-models
+          whl=~/tlf-models/tr_core_news_md-1.0-py3-none-any.whl
+          [ -f "$whl" ] || curl -fsSL -o "$whl" "$TR_MODEL_URL"
+          pip install "$whl"
+          python -m spacy download en_core_web_sm
+          python -m nltk.downloader cmudict      # İngilizce hece sayımı (textstat)
+
+      - name: Tests
+        run: python -m pytest

turkish_linguistic_features-0.2.0/.github/workflows/docs.yml

@@ -0,0 +1,41 @@
+# Docs: main'e her push'ta dokümantasyonu MkDocs ile kurar ve gh-pages dalına yazar.
+# GitHub Pages ayarı: Settings → Pages → Source "Deploy from a branch",
+# Branch "gh-pages" / "(root)". ("main /docs" seçilirse GitHub ham Markdown'ı
+# Jekyll ile kurar; menü, sekmeler ve bilgi kutuları bozulur.)
+
+name: Docs
+
+on:
+  push:
+    branches: [main]
+    paths:                                # yalnız dokümantasyon değişince
+      - "docs/**"
+      - "mkdocs.yml"
+      - ".github/workflows/docs.yml"
+  workflow_dispatch:                      # Actions sekmesinden elle de çalıştırılabilir
+
+permissions:
+  contents: write                         # gh-pages dalına yazabilmek için
+
+concurrency:                              # iki yayın aynı anda gh-pages'e yazmasın
+  group: docs
+  cancel-in-progress: false
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0                  # gh-deploy'un gh-pages dalını görebilmesi için
+      - uses: actions/setup-python@v7
+        with:
+          python-version: "3.12"
+      - name: Install MkDocs
+        run: pip install "mkdocs-material>=9.5"   # requirements.txt ile aynı alt sınır
+      - name: Configure git
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+      - name: Build and deploy
+        run: mkdocs gh-deploy --strict --force

turkish_linguistic_features-0.2.0/.github/workflows/publish.yml

@@ -0,0 +1,44 @@
+# Yayın: GitHub'da bir release yayımlanınca paketi derleyip PyPI'a yükler.
+# Actions sekmesinden elle çalıştırılırsa ("Run workflow") TestPyPI'a yükler — deneme için.
+#
+# Token ya da şifre kullanmaz: PyPI'ın "trusted publishing" özelliği (OIDC). PyPI'da
+# (ve TestPyPI'da) proje ayarlarına bir kez şu yayıncı tanımlanır:
+#   sahip: efeziya1 · depo: turkish-linguistic-features · iş akışı: publish.yml
+#   ortam: pypi (TestPyPI için: testpypi)
+# Sürüm numarası turkish_linguistic_features/__init__.py'deki __version__'dan gelir;
+# PyPI aynı numarayı ikinci kez kabul etmez.
+
+name: Publish
+
+on:
+  release:
+    types: [published]                    # release taslağı değil, yayımlanınca
+  workflow_dispatch:                      # elle: TestPyPI'a deneme yüklemesi
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    # Elle çalıştırılınca testpypi, release'te pypi ortamı.
+    environment: ${{ github.event_name == 'release' && 'pypi' || 'testpypi' }}
+    permissions:
+      id-token: write                     # trusted publishing için kimlik belirteci
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-python@v7
+        with:
+          python-version: "3.12"
+      - name: Build
+        run: |
+          pip install build
+          python -m build                 # dist/ altına sdist ve wheel
+      - name: Publish to TestPyPI
+        if: github.event_name != 'release'
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          repository-url: https://test.pypi.org/legacy/
+      - name: Publish to PyPI
+        if: github.event_name == 'release'
+        uses: pypa/gh-action-pypi-publish@release/v1

turkish_linguistic_features-0.2.0/.gitignore

@@ -0,0 +1,18 @@
+__pycache__/
+*.py[cod]
+.venv/
+dist/
+build/
+*.egg-info/
+.pytest_cache/
+.mypy_cache/
+examples/output/
+.vscode/
+
+# MkDocs derleme çıktısı — üretilen dosya, depoda durmaz
+site/
+
+# Yerel çalışma dosyaları — diskte durur, depoya girmez
+plan/
+_referans/
+scripts/check_plan.py

turkish_linguistic_features-0.2.0/.pre-commit-config.yaml

@@ -0,0 +1,9 @@
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v5.0.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: check-yaml
+      - id: check-added-large-files
+        args: [--maxkb=500]

turkish_linguistic_features-0.2.0/CHANGELOG.md

@@ -0,0 +1,159 @@
+# Changelog
+
+## [0.2.0] - 2026-10-13
+
+This release renames and removes feature keys; tables built with 0.1.0 will not line up
+column for column. Keys: Turkish 208 (the same number, not the same keys), English 180 → 181. Cited features: Turkish
+141 → 197, English 116 → 170; 11 keys in each language are plain definitions without a source.
+
+### Added
+
+- The package is on PyPI: `pip install turkish-linguistic-features`. The language data
+  is still installed separately.
+- Nine counts. `lexical`: `word_count` (words by the library's word), `type_count` (V),
+  `hapax_count` (V1, words occurring once) and `dislegomena_count` (V2, twice); `sentence`:
+  `sent_count`; `paragraph`: `para_count`; `phonetic`: `syllable_count`; `punctuation`:
+  `char_count` (whitespace included) and `punct_count` (marks). Divide other counts by them
+  to compare texts of different length. Citations: de Vel (2000) Table 2, attributes 1, 3 and
+  8; Zheng et al. (2006) Table 3, nos. 1, 54, 60–62, 247 and 248; Tweedie & Baayen (1998)
+  pp.325 and 329; de Vel et al. (2001) Table 2; Kincaid et al. (1975) p.38.
+- `ngram_matches(text, phrase, lang="tr")`: what one `custom_ngrams` phrase matched, with
+  counts, most frequent first: `{"eski ev": 2, "büyük bahçe": 1}`. The matching is the
+  one `analyze` uses, so the counts add up to the phrase's `ngram_{...}_count`. The public API
+  has eleven names.
+- Five lexical richness measures: `cttr` (Carroll's corrected TTR), `summer_s` (Summer's S),
+  `maas_a2` (Maas' a²), `herdan_vm` (Herdan's Vm) and `honore_r` (Honoré's R, 100 · ln N /
+  (1 − V1/V)). Citations: Carroll (1964) and Somers (1966) as cited in Torruella & Capsada
+  (2013) p.448; Maas (1972), Herdan (1955) and Honoré (1979) as cited in Tweedie & Baayen
+  (1998) eqs. (7), (18) and (11).
+- `describe_feature(key)["definitions"]`: for every term the `formula` uses (`sentence`,
+  `word`, `type`, `syllable`, `pos_tag` … 21 terms), the rule the library counts it with, who
+  defines that rule (`tlf`, `spacy`, `zeyrek`, `textstat`, `wordfreq`) and one sentence on
+  how it is counted. `describe_feature(key, lang=...)` picks the language's entry where a
+  definition differs by language (`syllable`, `type`).
+- `examples/10_kelime_oruntuleri.py`: counting your own word and tag phrases and seeing what
+  they matched.
+
+### Changed
+
+- `analyze_corpus` still stops at the first segment it cannot analyse, but the `ValueError` now
+  names that segment's `label`, `source` and `segment_id`; the original error is its `__cause__`.
+- Python 3.13 is tested in CI and listed in the package classifiers.
+- **Key names.** A name now says what the feature measures: a share between 0 and 1 ends in
+  `_ratio` unless the measure has an established name (`ttr`, `hdd`, `lexical_density`);
+  means and medians end the key; one family, one prefix. Values do not change.
+  - Punctuation: `punc_,_ratio` → `punct_comma_ratio`, `punc_._ratio` →
+    `punct_period_ratio`, `punc_;_ratio` → `punct_semicolon_ratio`, `punc_!_ratio` →
+    `punct_exclamation_ratio`, `punc_:_ratio` → `punct_colon_ratio`, `punc_-_ratio` →
+    `punct_dash_ratio`, `punc_ellipsis/paren/quote/question_ratio` → `punct_…_ratio`,
+    `punct_density` → `punct_char_ratio`, `digit_vs_all` → `digit_ratio`. Column names with
+    `.`, `-` or `!` broke attribute access in pandas and were rewritten by R.
+  - Means and medians: `avg_sent_len_word` → `sent_len_mean`, `avg_sent_len_char` →
+    `sent_len_char_mean`, `med_sent_len` → `sent_len_median`, `avg_word_length` →
+    `word_len_mean`, `sentence_syllable_mean` → `sent_syllable_mean`.
+  - Shares: `pos_noun` … → `pos_noun_ratio` … (12), `sentfinal_noun` … →
+    `sentfinal_noun_ratio` … (14), `char_a` … → `char_a_ratio` …, `hapax_percentage` →
+    `hapax_token_ratio` (a share, not a percentage), `question_per_sent` →
+    `question_sent_ratio`, `pronoun_freq` → `pronoun_ratio`, `n_lemma_count` → `lemma_count`.
+  - Morphology: keys from spaCy's UD tags drop `morph_` (`morph_case_loc` →
+    `case_loc_ratio`, 18 keys); keys from Zeyrek take `zeyrek_` (`case_loc_ratio` →
+    `zeyrek_case_loc_ratio`, `tense_past_def` → `zeyrek_tense_past_def_ratio`,
+    `agglutination_depth` → `zeyrek_agglutination_depth`, 23 keys).
+  - Deutsch, Jasbi & Shieber (2020) measures take their source names: `pos_dist_std` →
+    `posddev`, `pos_kl_div` → `posdiv`.
+  - Feature functions named after a key are renamed the same way.
+- **One word.** Every feature except the dependency group counts the same word: a
+  whitespace-separated piece with edge punctuation stripped, containing a letter or digit
+  (`e-posta`, `%50` and numbers are one word each). On the TOMA set it equals the expert's
+  word count in all 57 texts. Features that need a tag per word take the POS tag,
+  morphological tags, lemma and Zeyrek analysis from the first token inside the word; spaCy
+  splits 0.3% of Turkish and 2.7% of English words (`Türk-Amerikan`, `it's`). Punctuation no
+  longer enters the denominators of the `pos_*_ratio` shares and `pronoun_ratio`.
+  `syntactic_dep` keeps spaCy's tokens and parser sentences.
+- **Turkish lemmas come from Zeyrek**, not spaCy: the dictionary entry of Zeyrek's first
+  analysis, lowercased, without the infinitive `-mak/-mek`; circumflexes stay (`millî`). A word
+  Zeyrek cannot analyse keeps the part before its apostrophe (`Pittsburgh'tan` →
+  `pittsburgh`). spaCy's Turkish lemma left inflected forms as lemmas in 15% of words.
+  English lemmas still come from spaCy. Affected: `lemma_count`, `*_variation`, `wordfreq_*`,
+  `surface_per_lemma` and the `frequency_structure` group.
+- **One sentence rule.** `. ? ! …` end a sentence; `:` only when what follows starts like a new
+  sentence (capital letter, quote, dash or opening bracket). The `sentence` group,
+  `question_sent_ratio`, `posdiv` and `sent_syllable_mean` use it instead of spaCy's parser
+  sentences. `sents_per_para_mean` is now `sent_count` / `para_count`; it counted runs of
+  `.!?…` in each paragraph, so `Dr.` ended a sentence. The readability formulas use the same conditional colon (Kincaid's rule too);
+  the Çetinkaya rule keeps its source's unconditional `:`, and `syntactic_dep` keeps the
+  parser.
+- **Sentence-length thresholds** are calibrated on newspaper columns with the rules above
+  (15th/85th percentile; Turkish 162 columnists / 197,990 sentences, English 30 columnists /
+  52,745 sentences): Turkish 4/18 → 4/17, English 7/39 → 8/32.
+- **`segment_text` and `analyze_corpus(segment_size=...)` count words**, the same word
+  `analyze` counts: a `segment_size=1000` segment is exactly 1000 words. They used to count
+  spaCy tokens, punctuation included (100 tokens were about 83 words). Segment boundaries
+  change.
+- **`segment_text(size=...)` is now `segment_text(segment_size=...)`**, the name
+  `analyze_corpus` already uses for the same thing. Passing `size=` raises `TypeError`.
+- **Punctuation shares.** The ten `punct_*_ratio` keys are each mark type's share of all
+  punctuation marks (0-1, summing to 1), no longer marks per word, which could exceed 1. How
+  much punctuation a text has is `punct_char_ratio`. A text without punctuation gives `nan`.
+- **`custom_ngrams`** keys are `ngram_{...}_count` and the value is the number of matches, not
+  matches per window. A phrase item written as an UPPERCASE UD tag (`NOUN`, `VERB` …) matches
+  any word with that tag: `["ADJ", "NOUN"]` counts an adjective followed by a noun. Matches stay
+  inside a sentence. An empty text gives `nan`, like every other feature.
+- **Natural logarithm everywhere.** `dugast_u`, `lambda_pa`, `adjusted_modulus` (were log₁₀)
+  and `entropy`, `punct_entropy`, `sent_len_entropy`, `zeyrek_suffix_bigram_entropy`, `posdiv`
+  (were log₂) change by a constant factor; the ranking of texts does not change. The scale
+  `"bits"` is now `"nats"`.
+- **Syllable counts** read Turkish ordinals (`3. kat` → üçüncü), times and scores (`10:30` →
+  on otuz) and numbers glued to letters (`3kg` → üç kilogram), and add the reading of a listed
+  symbol to the number next to it in both languages (`%50` → yüzde elli, `$5` → five dollars).
+- **Citations.** Features that had no citation now cite a source read in full, among them
+  `sent_len_mean`
+  (Flesch 1948), `word_len_mean` (Mendenhall 1887), `sent_len_median` (Yule 1939),
+  `hapax_ratio` (de Vel 2000), the letter and punctuation shares (Zheng et al. 2006) and
+  `yule_k`, which now cites Yule (1944) itself. Where the library's definition differs from
+  the source, the citation says so. Bibliography 45 → 56 works.
+- **Verification tolerance** is 1% of the published value instead of a fixed 0.05. One row
+  moves from ✅ to 🟡 with its reason written out (`ttr`, QUITA Text 2, a misprint in the
+  source). Turkish report: 46 ✅ + 2 🟡; English: 35 ✅ + 3 🟡; no ❌.
+- **Examples.** `05` ranks labels by the share of short plus long sentences; `06` counts
+  sentences with the library's own rule; `02` no longer lists correlated feature pairs.
+
+### Removed
+
+- `pos_punct`: with shares counted per word there is no punctuation token to count;
+  punctuation is measured in the `punctuation` group.
+- `ttr_moving_slope` and `FeatureParams.ttr_slope_chunk_size` (passing it now raises
+  `TypeError`).
+- The eight coefficients of variation (`word_length_cv`, `sentence_length_cv`,
+  `para_len_cv`, `sents_per_para_cv`, `syllable_cv`, `sentence_syllable_cv`,
+  `verb_dist_cv`, `suffix_chain_cv`), `entropy_std` and `sent_len_skewness`: spread
+  statistics without a source of their own.
+- `para_count_norm`: exactly 1000 / `para_len_mean`. The plain count is `para_count`.
+- `nominal_verbal_ratio`: count `custom_ngrams=[["NOUN"], ["VERB"]]` and divide.
+- The `"cv"` and `"signed"` scales, which no key uses any more.
+- `exceptions.MissingDependencyError`: it was never raised. A missing optional package gives
+  `MissingDependencyWarning` and `nan`; missing language data gives `ModelNotFoundError`.
+- The undocumented `LINGUISTIC_FEATURES_NO_ZEYREK_WARMUP` environment variable: skipping the
+  Zeyrek warm-up and then analysing Turkish in the same process could crash on Windows.
+
+### Fixed
+
+- `…` (one character) now ends a sentence like `...` in every rule set; before, the same text
+  got a different sentence count and readability score depending on how the ellipsis was
+  typed.
+- Words with a curly apostrophe (`Zeynep’i`) were left unanalysed by Zeyrek and dropped out
+  of the `morphological_zeyrek` features; `’` and `‘` are turned into `'` first.
+- The minimum spaCy version is 3.8 (`spacy>=3.8,<4`, was 3.5): both language models need it.
+- The install hint for a missing optional package names the package itself
+  (`pip install 'wordfreq>=3.0'`), not an extra of a package not yet on PyPI.
+- Registry and documentation texts: three descriptions ran words together, the parameter
+  guide showed `params` as a list (it is a tuple), the `analyze` docstring example gave 3.5
+  for a mean of 3.0, four bibliography entries had Turkish fragments.
+- The limitations page explains how Zeyrek orders its analyses and that the first one can
+  depend on `PYTHONHASHSEED` (200 of 8,215 words on the TOMA set).
+
+## [0.1.0] - 2026-09-28
+
+### Added
+
+- Initial public release.

turkish_linguistic_features-0.2.0/CITATION.cff

@@ -0,0 +1,32 @@
+cff-version: 1.2.0
+message: "If you use this software in your research, please cite it using this metadata."
+type: software
+title: "turkish-linguistic-features"
+abstract: >-
+  A toolkit for extracting quantitative linguistic features from Turkish and
+  English text: 208 features for Turkish, 181 for English. 197 of the Turkish
+  features cite a source in the literature; the rest are plain definitions.
+authors:
+  - given-names: "Efe Ziya"
+    family-names: "Okul"
+version: "0.2.0"
+date-released: 2026-10-13
+doi: 10.5281/zenodo.23009540
+license: MIT
+repository-code: "https://github.com/efeziya1/turkish-linguistic-features"
+keywords:
+  - quantitative-linguistics
+  - computational-linguistics
+  - corpus-linguistics
+  - stylometry
+  - turkish
+  - english
+  - feature-extraction
+  - text-analysis
+  - natural-language-processing
+  - readability
+  - lexical-richness
+  - lexical-diversity
+  - syntactic-complexity
+  - morphology
+  - digital-humanities

turkish_linguistic_features-0.2.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Efe Ziya Okul
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

turkish_linguistic_features-0.2.0/PKG-INFO

@@ -0,0 +1,137 @@
+Metadata-Version: 2.5
+Name: turkish-linguistic-features
+Version: 0.2.0
+Summary: A comprehensive toolkit for extracting quantitative linguistic features from Turkish and English text.
+Project-URL: Repository, https://github.com/efeziya1/turkish-linguistic-features
+Project-URL: Documentation, https://efeziya1.github.io/turkish-linguistic-features/
+Project-URL: Issues, https://github.com/efeziya1/turkish-linguistic-features/issues
+Author-email: Efe Ziya Okul <efeziya@live.com>
+License: MIT
+License-File: LICENSE
+Keywords: computational-linguistics,corpus-linguistics,digital-humanities,english,feature-extraction,lexical-diversity,lexical-richness,morphology,natural-language-processing,quantitative-linguistics,readability,stylometry,syntactic-complexity,text-analysis,turkish
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Text Processing :: Linguistic
+Requires-Python: >=3.10
+Requires-Dist: nltk>=3.8
+Requires-Dist: numpy>=1.21
+Requires-Dist: spacy<4,>=3.8
+Requires-Dist: textstat<0.8,>=0.7.10
+Requires-Dist: zeyrek<0.2,>=0.1.3
+Provides-Extra: all
+Requires-Dist: pandas>=1.5; extra == 'all'
+Requires-Dist: wordfreq>=3.0; extra == 'all'
+Provides-Extra: lexical-freq
+Requires-Dist: wordfreq>=3.0; extra == 'lexical-freq'
+Provides-Extra: pandas
+Requires-Dist: pandas>=1.5; extra == 'pandas'
+Description-Content-Type: text/markdown
+
+# turkish-linguistic-features
+
+[![CI](https://github.com/efeziya1/turkish-linguistic-features/actions/workflows/ci.yml/badge.svg)](https://github.com/efeziya1/turkish-linguistic-features/actions/workflows/ci.yml)
+[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.23009540.svg)](https://doi.org/10.5281/zenodo.23009540)
+
+Extracts 208 quantitative linguistic features from Turkish text and 181 from
+English. Every feature has its formula written out; 197 of the Turkish features
+and 170 of the English ones cite a source in the literature, and the rest are
+plain definitions (such as the dash's share of punctuation marks). The
+[verification report](https://efeziya1.github.io/turkish-linguistic-features/verification-report/) shows which ones have been
+checked against a number their source published. Documentation:
+https://efeziya1.github.io/turkish-linguistic-features/
+
+**Status: early development (0.x).** Changes: [CHANGELOG](https://github.com/efeziya1/turkish-linguistic-features/blob/main/CHANGELOG.md).
+
+```python
+import turkish_linguistic_features as tlf
+
+oz = tlf.analyze("Renksiz yeşil fikirler öfkeyle uyur.", lang="tr")
+oz["word_len_mean"]   # 6.2
+oz["atesman"]         # 89.355  (Ateşman 1997 readability)
+```
+
+## Documentation
+
+**→ [Full documentation](https://efeziya1.github.io/turkish-linguistic-features/)** — tutorial, how-to
+guides, feature reference and limitations, in **Turkish and English**.
+
+| | |
+|---|---|
+| Türkçe | [tr/](https://efeziya1.github.io/turkish-linguistic-features/tr/) |
+| English | [en/](https://efeziya1.github.io/turkish-linguistic-features/en/) |
+| Feature reference | [reference/features/](https://efeziya1.github.io/turkish-linguistic-features/reference/features/) |
+| Verification report | [TR](https://efeziya1.github.io/turkish-linguistic-features/dogrulama-raporu/) · [EN](https://efeziya1.github.io/turkish-linguistic-features/verification-report/) |
+
+## Install
+
+```bash
+pip install turkish-linguistic-features
+```
+
+### Language data
+
+`pip install` sets up the Python dependencies but not the language data.
+Install the data for the language(s) you use, once:
+
+```bash
+# Turkish — 156 MB. The model is by turkish-nlp-suite and is not in spaCy's
+# registry, so install the wheel directly:
+pip install https://huggingface.co/turkish-nlp-suite/tr_core_news_md/resolve/main/tr_core_news_md-1.0-py3-none-any.whl
+
+# English — the spaCy model (12 MB) and the CMU Pronouncing Dictionary for
+# syllable counts (1 MB):
+python -m spacy download en_core_web_sm
+python -m nltk.downloader cmudict
+
+# Optional — for the two wordfreq_* features:
+pip install 'wordfreq>=3.0'
+```
+
+If something is missing, `analyze()` stops with an error that shows the
+command to run.
+
+Verified with spaCy 3.8.16, `en_core_web_sm` 3.8.0 and `tr_core_news_md` 1.0.
+The Turkish wheel says `1.0` in its filename but `3.4.2` in its metadata, and
+loading it prints a `W094` warning; both are harmless
+([details](https://efeziya1.github.io/turkish-linguistic-features/en/getting-started/)).
+
+## The whole API
+
+Eleven public names:
+
+```python
+tlf.analyze(text, lang="tr")                     # 208 features (TR) / 181 (EN)
+tlf.analyze(text, groups=["readability"])        # one group only
+tlf.analyze_corpus("corpus/")                    # one row per file
+tlf.analyze_corpus("corpus/", segment_size=1000) # or split into chunks
+tlf.save_csv(rows, "features.csv")               # write them out
+tlf.segment_text(text, segment_size=1000, lang="tr")  # split a single text
+tlf.analyze(text, custom_ngrams=[["ADJ", "NOUN"]])  # count your own phrases
+tlf.ngram_matches(text, ["ADJ", "NOUN"])         # ...and see what they matched
+tlf.describe_feature("mtld")                     # what a key measures, and its source
+tlf.FeatureParams(mattr_window=100)              # metric constants
+
+# tlf.LinguisticFeaturesError  tlf.ModelNotFoundError
+# tlf.MissingDependencyWarning  tlf.ParagraphStructureWarning
+```
+
+Full signatures: [Public API](https://efeziya1.github.io/turkish-linguistic-features/reference/api/).
+
+## Development
+
+```bash
+pip install -r requirements.txt
+pytest                          # pytest -m "" also runs the slow test
+ruff check .
+python scripts/generate_verification_report.py  # regenerate the verification reports
+python scripts/generate_feature_reference.py    # regenerate the feature reference
+```
+
+## License
+
+MIT — see `LICENSE`.

turkish_linguistic_features-0.2.0/README.md

@@ -0,0 +1,103 @@
+# turkish-linguistic-features
+
+[![CI](https://github.com/efeziya1/turkish-linguistic-features/actions/workflows/ci.yml/badge.svg)](https://github.com/efeziya1/turkish-linguistic-features/actions/workflows/ci.yml)
+[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.23009540.svg)](https://doi.org/10.5281/zenodo.23009540)
+
+Extracts 208 quantitative linguistic features from Turkish text and 181 from
+English. Every feature has its formula written out; 197 of the Turkish features
+and 170 of the English ones cite a source in the literature, and the rest are
+plain definitions (such as the dash's share of punctuation marks). The
+[verification report](https://efeziya1.github.io/turkish-linguistic-features/verification-report/) shows which ones have been
+checked against a number their source published. Documentation:
+https://efeziya1.github.io/turkish-linguistic-features/
+
+**Status: early development (0.x).** Changes: [CHANGELOG](https://github.com/efeziya1/turkish-linguistic-features/blob/main/CHANGELOG.md).
+
+```python
+import turkish_linguistic_features as tlf
+
+oz = tlf.analyze("Renksiz yeşil fikirler öfkeyle uyur.", lang="tr")
+oz["word_len_mean"]   # 6.2
+oz["atesman"]         # 89.355  (Ateşman 1997 readability)
+```
+
+## Documentation
+
+**→ [Full documentation](https://efeziya1.github.io/turkish-linguistic-features/)** — tutorial, how-to
+guides, feature reference and limitations, in **Turkish and English**.
+
+| | |
+|---|---|
+| Türkçe | [tr/](https://efeziya1.github.io/turkish-linguistic-features/tr/) |
+| English | [en/](https://efeziya1.github.io/turkish-linguistic-features/en/) |
+| Feature reference | [reference/features/](https://efeziya1.github.io/turkish-linguistic-features/reference/features/) |
+| Verification report | [TR](https://efeziya1.github.io/turkish-linguistic-features/dogrulama-raporu/) · [EN](https://efeziya1.github.io/turkish-linguistic-features/verification-report/) |
+
+## Install
+
+```bash
+pip install turkish-linguistic-features
+```
+
+### Language data
+
+`pip install` sets up the Python dependencies but not the language data.
+Install the data for the language(s) you use, once:
+
+```bash
+# Turkish — 156 MB. The model is by turkish-nlp-suite and is not in spaCy's
+# registry, so install the wheel directly:
+pip install https://huggingface.co/turkish-nlp-suite/tr_core_news_md/resolve/main/tr_core_news_md-1.0-py3-none-any.whl
+
+# English — the spaCy model (12 MB) and the CMU Pronouncing Dictionary for
+# syllable counts (1 MB):
+python -m spacy download en_core_web_sm
+python -m nltk.downloader cmudict
+
+# Optional — for the two wordfreq_* features:
+pip install 'wordfreq>=3.0'
+```
+
+If something is missing, `analyze()` stops with an error that shows the
+command to run.
+
+Verified with spaCy 3.8.16, `en_core_web_sm` 3.8.0 and `tr_core_news_md` 1.0.
+The Turkish wheel says `1.0` in its filename but `3.4.2` in its metadata, and
+loading it prints a `W094` warning; both are harmless
+([details](https://efeziya1.github.io/turkish-linguistic-features/en/getting-started/)).
+
+## The whole API
+
+Eleven public names:
+
+```python
+tlf.analyze(text, lang="tr")                     # 208 features (TR) / 181 (EN)
+tlf.analyze(text, groups=["readability"])        # one group only
+tlf.analyze_corpus("corpus/")                    # one row per file
+tlf.analyze_corpus("corpus/", segment_size=1000) # or split into chunks
+tlf.save_csv(rows, "features.csv")               # write them out
+tlf.segment_text(text, segment_size=1000, lang="tr")  # split a single text
+tlf.analyze(text, custom_ngrams=[["ADJ", "NOUN"]])  # count your own phrases
+tlf.ngram_matches(text, ["ADJ", "NOUN"])         # ...and see what they matched
+tlf.describe_feature("mtld")                     # what a key measures, and its source
+tlf.FeatureParams(mattr_window=100)              # metric constants
+
+# tlf.LinguisticFeaturesError  tlf.ModelNotFoundError
+# tlf.MissingDependencyWarning  tlf.ParagraphStructureWarning
+```
+
+Full signatures: [Public API](https://efeziya1.github.io/turkish-linguistic-features/reference/api/).
+
+## Development
+
+```bash
+pip install -r requirements.txt
+pytest                          # pytest -m "" also runs the slow test
+ruff check .
+python scripts/generate_verification_report.py  # regenerate the verification reports
+python scripts/generate_feature_reference.py    # regenerate the feature reference
+```
+
+## License
+
+MIT — see `LICENSE`.