Datasets:

glossAPI
/

eellak-articles

- Single self-contained parquet (eellak.parquet, 25.09 MB)
- Embedded markdown content per article (front-matter stripped)
- NBSP (U+00A0) normalized to regular space across 4,192 files
- Dropped redundant columns (markdown_path, file_size_bytes, user_login)
- Recomputed token_count using nlpaueb/bert-base-greek-uncased-v1
- New README modeled on glossAPI/deliberations-v2 with full per-site
coverage table and known-data-quality notes.

Files changed (2) hide show

README.md +137 -39
eellak.parquet +3 -0

README.md CHANGED Viewed

@@ -1,39 +1,137 @@
----
-language:
-- el
----
-**Dataset Info:**
-This dataset contains blog posts from the Greek Free/Open Source Software (GFOSS) websites (https://ellak.gr/).
-Posts span from 2010 to 2026, providing a rich resource for NLP, content analysis, and research on Greek-language open-source community blogs.
-**Metadata Info:**
-| **Column**          | **Type**      | **Description**                                                                 |
-|-----------------|----------|-----------------------------------------------------------------------------|
-| site            | string   | Website/domain (e.g., advisory.ellak.gr, scientix.ellak.gr)               |
-| id              | string   | Post ID                                                                    |
-| post_author     | string   | Author ID                                                                  |
-| post_date       | datetime | Original publish date                                                      |
-| post_title      | string   | Title of the post (mostly in Greek)                                       |
-| post_name       | string   | URL slug (URL-encoded)                                                     |
-| post_modified   | datetime | Last modified date                                                         |
-| guid            | string   | Permalink/URL to the post                                                 |
-| user_login      | string   | Author's login username                                                   |
-| markdown_path   | string   | Relative path to the markdown file for this post                           |
-| token_count     | int      | Number of tokens in the post content                                       |
-**Dataset Statistics:**
-Total Articles: 7,286
-Total tokens: 8,497,146
-**License:**
-This dataset is released under the Creative Commons Attribution 4.0 International (CC-BY-4.0) license.
-You are free to share and adapt the material, provided that appropriate credit is given to the original sources.
-**Contact:**
-For feedback contact: glossapi.team@eellak.gr

+---
+license: cc-by-sa-4.0
+language:
+- el
+pretty_name: EELLAK Network Articles
+tags:
+- greek
+- open-source
+- open-government
+- open-data
+- open-content
+- creative-commons
+- ellak
+- eellak
+- wordpress
+- blog
+size_categories:
+- 1K<n<10K
+---
+# EELLAK Network Articles - Άρθρα από το δίκτυο ιστοτόπων ΕΕΛ/ΛΑΚ
+**Dataset Info**
+This dataset is a structured snapshot of the multi-site WordPress network operated by **EELLAK** (Εταιρεία Ελεύθερου Λογισμικού / Λογισμικού Ανοιχτού Κώδικα - the Greek Free / Open Source Software Society). EELLAK runs a constellation of 23 thematic subdomains under `ellak.gr`, each of which acts as a long-running editorial blog dedicated to a specific facet of the open-technology and open-knowledge ecosystem in Greece: open-source software, open government, open data, open educational resources, open standards, open hardware, open content, Creative Commons, data privacy, free public WiFi, open design, and related civic-tech topics.
+The dataset contains **7,284 articles** spanning **2008-05-12 - 2026-03-19**, covering the full editorial output of these sites since the foundation of the EELLAK web network. Each row corresponds to a single WordPress post and includes the article body in Markdown plus the originating site, publication metadata, and authorship information.
+The 23 sites covered (and their thematic focus) are:
+| Subdomain | Articles | Focus |
+| --- | ---: | --- |
+| `opensource.ellak.gr` | 1,440 | Free / open-source software news |
+| `opengov.ellak.gr` | 816 | Open government and civic-tech |
+| `edu.ellak.gr` | 683 | Open education in primary / secondary schools |
+| `mycontent.ellak.gr` | 619 | Open content and digital publishing |
+| `creativecommons.ellak.gr` | 498 | Creative Commons Greece chapter |
+| `privacy.ellak.gr` | 436 | Data protection and privacy |
+| `opendata.ellak.gr` | 433 | Open public data |
+| `eellak.ellak.gr` | 377 | EELLAK organizational news |
+| `openstandards.ellak.gr` | 346 | Open standards and interoperability |
+| `gfoss.ellak.gr` | 337 | GFOSS general open-source advocacy |
+| `openbusiness.ellak.gr` | 273 | Open business models |
+| `openhardware.ellak.gr` | 263 | Open hardware |
+| `oer.ellak.gr` | 188 | Open educational resources |
+| `openwifi.ellak.gr` | 138 | Open / municipal WiFi initiatives |
+| `legal.ellak.gr` | 118 | Open-source legal affairs and IP |
+| `advisory.ellak.gr` | 82 | Advisory groups and policy work |
+| `opendesign.ellak.gr` | 82 | Open design |
+| `scientix.ellak.gr` | 56 | Science education (Scientix programme) |
+| `odi.ellak.gr` | 54 | Open Data Institute (Greek node) |
+| `obs.ellak.gr` | 32 | Observatory / monitoring |
+| `mydata.ellak.gr` | 7 | Personal data sovereignty |
+| `mathe.ellak.gr` | 5 | Public-awareness mini-site |
+| `publicfiscal.ellak.gr` | 1 | Public-finance transparency |
+The text is in Modern Greek, with occasional English passages (e.g. quoted upstream announcements) and bilingual titles in the `opendesign.ellak.gr` posts.
+**Metadata Info**
+The parquet has one row per article. The schema is:
+| Column          | Type            | Description                                                                                                          |
+| --------------- | --------------- | -------------------------------------------------------------------------------------------------------------------- |
+| `site`          | string          | Subdomain of `ellak.gr` where the article was published (one of 23 values).                                          |
+| `id`            | string          | WordPress numeric post id, unique within `site`. The pair `(site, id)` is the primary key of the dataset.            |
+| `post_title`    | string          | Article title as published.                                                                                          |
+| `post_name`     | string          | URL-encoded slug of the article (the WordPress `post_name` field; appears in the canonical URL).                     |
+| `post_date`     | timestamp[us]   | Date/time the article was originally published, in Athens local time (`Europe/Athens`).                              |
+| `post_modified` | timestamp[us]   | Date/time of the most recent edit, in Athens local time.                                                             |
+| `post_author`   | string          | WordPress numeric author id of the publishing account (opaque key; usernames are not included).                      |
+| `guid`          | string          | WordPress canonical permalink for the post. See *Known Data Quality Notes* for cross-posted articles.                |
+| `content`       | string          | Markdown body of the article. YAML front-matter has been stripped; non-breaking spaces have been normalized.         |
+| `word_count`    | int64           | Whitespace-separated word count of `content`.                                                                        |
+| `token_count`   | int64           | Subword token count of `content` under the `nlpaueb/bert-base-greek-uncased-v1` WordPiece tokenizer.                 |
+**Note on Timestamps**
+`post_date` and `post_modified` are stored as timezone-naive `timestamp[us]` values, but the wall-clock values represent **Athens local time** (`Europe/Athens` - EET in winter, EEST in summer). This matches the local time configured on the source WordPress instances. To work with these timestamps in a timezone-aware way, localize them on read, e.g.:
+```python
+import pandas as pd
+df = pd.read_parquet("eellak.parquet")
+df["post_date"] = df["post_date"].dt.tz_localize(
+    "Europe/Athens", ambiguous="NaT", nonexistent="shift_forward"
+)
+```
+**Dataset Statistics**
+* Articles (rows): 7.284
+* Sites covered: 23
+* Coverage (`post_date`): 2008-05-12 - 2026-03-19
+* Size: 25,10 MB
+* Words: 4.255.024 (counted on `content`, whitespace-separated)
+* Tokens: 8.497.146 (counted on `content` with the `nlpaueb/bert-base-greek-uncased-v1` WordPiece tokenizer, no special tokens, no truncation)
+The fields included in the word/token count are: `content`. Identifiers, URLs, slugs, timestamps, titles, and author fields are excluded.
+**Note on Content Formatting**
+The `content` field carries the article body in **Markdown**, as it was rendered from the original WordPress posts. A few formatting choices are worth noting:
+- **YAML front matter has been stripped.** Each source `.md` file originally began with a small YAML block (`title`, `date`, `modified`, `author`, `guid`, `site`) that fully duplicated the parquet metadata columns. This block has been removed so that `content` contains only the body text.
+- **Non-breaking spaces (U+00A0) have been normalized to regular spaces.** The WordPress visual editor often inserts NBSP characters during composition; 32.192 such characters across 4.192 articles were replaced with regular spaces. This affects whitespace only and does not change any visible text.
+- **WordPress shortcodes are preserved verbatim.** Tokens such as `[featured-video-plus]`, `[wikiful url=...]`, `[caption]`, `[gallery]`, etc. are kept as-is. They reflect dynamic content (embedded videos, galleries, plugin output) that is not resolvable from the static export. See *Known Data Quality Notes*.
+- **Inline HTML is preserved verbatim.** Some posts contain inline `<table>`, `<iframe>`, `<sup>`/`<sub>` and similar fragments authored by the editors; these are kept as written. Most "HTML-looking" tags found by a tag-shape regex are actually Markdown autolinks of the form `<https://example.com>`, which Markdown explicitly supports.
+- **Markdown links, headings, lists and emphasis are preserved.** No flattening to plain text is performed, so downstream consumers that need raw text can apply their own Markdown parser.
+**Known Data Quality Notes**
+The dataset reflects the source as faithfully as possible. The following quirks exist on the live `ellak.gr` sites and have therefore been **preserved**, not silently corrected:
+- **Cross-posted articles.** 43 articles (86 rows) are mirrored across two subdomains of the EELLAK network: the same canonical `guid` appears once under its origin site (typically `oer.ellak.gr` or `opensource.ellak.gr`) and once under `edu.ellak.gr`. Each mirror has its own `(site, id)` and is treated as a separate row. Users who want a deduplicated view at the article level can group by `guid`.
+- **`post_modified` earlier than `post_date`.** 358 rows have `post_modified < post_date`. This is a known WordPress behavior for back-dated, imported, or migrated posts (e.g. articles re-published after the WordPress instance was upgraded). The values are taken as-is from the source database.
+- **Very short articles.** 16 articles have fewer than 5 words. The vast majority are `opendesign.ellak.gr` portfolio entries whose body on the live site is just an embedded video shortcode (`[featured-video-plus]`); a handful of others are placeholder posts. They are kept for completeness because they exist as published articles on the source sites.
+- **WordPress shortcodes inside `content`.** Posts that rely on plugins (galleries, embedded videos, dynamic widgets) contain unresolved shortcodes. Resolving them would require executing the original WordPress runtime against the original databases.
+**Usage**
+This dataset is useful for:
+- Natural Language Processing (NLP) tasks for Modern Greek, especially in the **technology, civic-tech, open-government, and open-source policy** registers
+- Building Greek language models, lexicons, or domain-adapted embeddings for the open-technology and digital-rights domain
+- Topic modeling, named-entity recognition, and event extraction over a long-running, thematically coherent corpus (2008–2026)
+- Diachronic studies of Greek public discourse on open data, open government, digital privacy, and free software
+- Information-retrieval and Q&A systems over the EELLAK editorial archive
+**License**
+This dataset is licensed under the [Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)](https://creativecommons.org/licenses/by-sa/4.0/) license. EELLAK is the publisher of the source content and authorizes its redistribution under this license. When using this dataset, please attribute:
+> *EELLAK Network Articles*, glossAPI (EELLAK), derived from the network of `ellak.gr` editorial sites of the Greek Free / Open Source Software Society, licensed under CC BY-SA 4.0.
+**Contact**
+For feedback contact: glossapi.team@eellak.gr

eellak.parquet ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7a22756102fe1f04c8896331e641343b1c82472822c45e612849b272c9af3619
+size 26312988