{
  "id": 673577,
  "title": "Questions about publications.csv and published_texts.csv",
  "url": "/competitions/deep-past-initiative-machine-translation/discussion/673577",
  "author_name": "ND.C",
  "post_date": "2026-02-15T13:27:54.483000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hello. I’m a complete beginner in NLP.\nStill, since this is my first competition, I really want to learn as much as I can.\nI have some questions regarding the dataset, so I’m posting here.</p>\n<p>Q1. In the description of <strong>publications.csv</strong>, it says: “Contains the raw text of about 880 scholarly publications containing translations from Old Assyrian into multiple modern languages. The texts were produced via OCR with LLM postprocessing.” Should this be understood to mean that the <em>page_text</em> field in <strong>publications.csv</strong> consists entirely of translations (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?</p>\n<p>Q2. Also, in <em>page_text</em>, should rows where <em>has_Akkadian</em> is true be understood as containing Akkadian within the translation (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?</p>\n<p>Q3. In the description of <strong>publications.csv</strong>, it says:“You may attempt to extract these translations and align them with the transliterations in <strong>published_texts.csv</strong>.” Does this mean that it would be possible to produce <em>transliteration–translation pairs</em> by using <strong>published_texts.csv</strong> (which contains only transliterations) and <strong>publications.csv</strong> (which contains only translations)? Is that the correct interpretation?</p>\n<p>I’m very curious, thank you! <a href=\"https://www.kaggle.com/deeppast\" target=\"_blank\">@deeppast</a> </p>",
  "messages": [
    {
      "id": 3406369,
      "postDate": "2026-02-15T13:27:54.483Z",
      "content": "<p>Hello. I’m a complete beginner in NLP.\nStill, since this is my first competition, I really want to learn as much as I can.\nI have some questions regarding the dataset, so I’m posting here.</p>\n<p>Q1. In the description of <strong>publications.csv</strong>, it says: “Contains the raw text of about 880 scholarly publications containing translations from Old Assyrian into multiple modern languages. The texts were produced via OCR with LLM postprocessing.” Should this be understood to mean that the <em>page_text</em> field in <strong>publications.csv</strong> consists entirely of translations (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?</p>\n<p>Q2. Also, in <em>page_text</em>, should rows where <em>has_Akkadian</em> is true be understood as containing Akkadian within the translation (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?</p>\n<p>Q3. In the description of <strong>publications.csv</strong>, it says:“You may attempt to extract these translations and align them with the transliterations in <strong>published_texts.csv</strong>.” Does this mean that it would be possible to produce <em>transliteration–translation pairs</em> by using <strong>published_texts.csv</strong> (which contains only transliterations) and <strong>publications.csv</strong> (which contains only translations)? Is that the correct interpretation?</p>\n<p>I’m very curious, thank you! <a href=\"https://www.kaggle.com/deeppast\" target=\"_blank\">@deeppast</a> </p>",
      "rawMarkdown": "Hello. I’m a complete beginner in NLP.\nStill, since this is my first competition, I really want to learn as much as I can.\nI have some questions regarding the dataset, so I’m posting here.\n\n\n\n\nQ1. In the description of **publications.csv**, it says: “Contains the raw text of about 880 scholarly publications containing translations from Old Assyrian into multiple modern languages. The texts were produced via OCR with LLM postprocessing.” Should this be understood to mean that the *page_text* field in **publications.csv** consists entirely of translations (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?\n\nQ2. Also, in *page_text*, should rows where *has_Akkadian* is true be understood as containing Akkadian within the translation (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?\n\nQ3. In the description of **publications.csv**, it says:“You may attempt to extract these translations and align them with the transliterations in **published_texts.csv**.” Does this mean that it would be possible to produce *transliteration–translation pairs* by using **published_texts.csv** (which contains only transliterations) and **publications.csv** (which contains only translations)? Is that the correct interpretation?\n\n\n\n\nI’m very curious, thank you! @deeppast ",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3406369": "Hello. I’m a complete beginner in NLP.\nStill, since this is my first competition, I really want to learn as much as I can.\nI have some questions regarding the dataset, so I’m posting here.\n\n\n\n\nQ1. In the description of **publications.csv**, it says: “Contains the raw text of about 880 scholarly publications containing translations from Old Assyrian into multiple modern languages. The texts were produced via OCR with LLM postprocessing.” Should this be understood to mean that the *page_text* field in **publications.csv** consists entirely of translations (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?\n\nQ2. Also, in *page_text*, should rows where *has_Akkadian* is true be understood as containing Akkadian within the translation (i.e., not annotations, tables, bibliographies, captions, commentary, etc.)?\n\nQ3. In the description of **publications.csv**, it says:“You may attempt to extract these translations and align them with the transliterations in **published_texts.csv**.” Does this mean that it would be possible to produce *transliteration–translation pairs* by using **published_texts.csv** (which contains only transliterations) and **publications.csv** (which contains only translations)? Is that the correct interpretation?\n\n\n\n\nI’m very curious, thank you! @deeppast "
  }
}