{
  "id": 669163,
  "title": "Encoding mismatches in test.csv",
  "url": "/competitions/deep-past-initiative-machine-translation/discussion/669163",
  "author_name": "",
  "post_date": "2026-01-20T22:40:35.452425500Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I decided to look at test.csv to try to get a better idea of what processing might look like, and after a bit of examening it it I have noticed some mismatches that may have happened while encoding.\nExamples I've found:\nIn test.csv, \nm appears where ṭ should be on some occasions (e.g. mup-up-um instead of   ṭup-up-um).</p>\n<p>aa appears where ša should be (e.g. aa-qí-il instead of ša-qí-il).</p>\n<p>ia appears where iš should be (e.g. kà-ni-ia instead of kà-ni-iš).</p>\n<p>aí appears where ši should be.</p>\n<p>Test.csv has these errors, while train.csv does not, which could hurt BLEU score for example,\nIn train.csv: It is spelled ṭup-pu-um (Correct standard Akkadian), but in test.csv: It appeared as mup-pu-um (which is an anomaly).</p>\n<p><a href=\"https://www.kaggle.com/deeppast\" target=\"_blank\">@deeppast</a>, did you use a different form of encoding in test.csv than in train.csv?</p>",
  "messages": [
    {
      "id": "3394365",
      "postDate": "01/20/2026 22:40:35",
      "content": "<p>I decided to look at test.csv to try to get a better idea of what processing might look like, and after a bit of examening it it I have noticed some mismatches that may have happened while encoding.\nExamples I've found:\nIn test.csv, \nm appears where ṭ should be on some occasions (e.g. mup-up-um instead of   ṭup-up-um).</p>\n<p>aa appears where ša should be (e.g. aa-qí-il instead of ša-qí-il).</p>\n<p>ia appears where iš should be (e.g. kà-ni-ia instead of kà-ni-iš).</p>\n<p>aí appears where ši should be.</p>\n<p>Test.csv has these errors, while train.csv does not, which could hurt BLEU score for example,\nIn train.csv: It is spelled ṭup-pu-um (Correct standard Akkadian), but in test.csv: It appeared as mup-pu-um (which is an anomaly).</p>\n<p><a href=\"https://www.kaggle.com/deeppast\" target=\"_blank\">@deeppast</a>, did you use a different form of encoding in test.csv than in train.csv?</p>",
      "rawMarkdown": "I decided to look at test.csv to try to get a better idea of what processing might look like, and after a bit of examening it it I have noticed some mismatches that may have happened while encoding.\nExamples I've found:\nIn test.csv, \nm appears where ṭ should be on some occasions (e.g. mup-up-um instead of   ṭup-up-um).\n\naa appears where ša should be (e.g. aa-qí-il instead of ša-qí-il).\n\nia appears where iš should be (e.g. kà-ni-ia instead of kà-ni-iš).\n\naí appears where ši should be.\n\nTest.csv has these errors, while train.csv does not, which could hurt BLEU score for example,\nIn train.csv: It is spelled ṭup-pu-um (Correct standard Akkadian), but in test.csv: It appeared as mup-pu-um (which is an anomaly).\n\n@deeppast, did you use a different form of encoding in test.csv than in train.csv?",
      "votes": null
    },
    {
      "id": "3394451",
      "postDate": "01/21/2026 04:19:00",
      "content": "<p>the test.csv you see in Data is not the real test set. It's a dummy file… I'm not sure whether the real test set has these mistakes you mentioned tho, but probably not. Otherwise this is not a serious competition!</p>",
      "rawMarkdown": "the test.csv you see in Data is not the real test set. It's a dummy file... I'm not sure whether the real test set has these mistakes you mentioned tho, but probably not. Otherwise this is not a serious competition!",
      "votes": null
    },
    {
      "id": "3394669",
      "postDate": "01/21/2026 13:16:58",
      "content": "<p>Oh, ok thanks a lot! That really helps. Thanks for clarifying that!</p>",
      "rawMarkdown": "Oh, ok thanks a lot! That really helps. Thanks for clarifying that!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3394451,
      "author_name": "honganzhu",
      "author_url": "",
      "post_date": "01/21/2026 04:19:00",
      "content": "<p>the test.csv you see in Data is not the real test set. It's a dummy file… I'm not sure whether the real test set has these mistakes you mentioned tho, but probably not. Otherwise this is not a serious competition!</p>",
      "votes": null,
      "replies": [
        {
          "id": 3394669,
          "author_name": "leowithrow",
          "author_url": "",
          "post_date": "01/21/2026 13:16:58",
          "content": "<p>Oh, ok thanks a lot! That really helps. Thanks for clarifying that!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3394365": "I decided to look at test.csv to try to get a better idea of what processing might look like, and after a bit of examening it it I have noticed some mismatches that may have happened while encoding.\nExamples I've found:\nIn test.csv, \nm appears where ṭ should be on some occasions (e.g. mup-up-um instead of   ṭup-up-um).\n\naa appears where ša should be (e.g. aa-qí-il instead of ša-qí-il).\n\nia appears where iš should be (e.g. kà-ni-ia instead of kà-ni-iš).\n\naí appears where ši should be.\n\nTest.csv has these errors, while train.csv does not, which could hurt BLEU score for example,\nIn train.csv: It is spelled ṭup-pu-um (Correct standard Akkadian), but in test.csv: It appeared as mup-pu-um (which is an anomaly).\n\n@deeppast, did you use a different form of encoding in test.csv than in train.csv?",
    "3394451": "the test.csv you see in Data is not the real test set. It's a dummy file... I'm not sure whether the real test set has these mistakes you mentioned tho, but probably not. Otherwise this is not a serious competition!",
    "3394669": "Oh, ok thanks a lot! That really helps. Thanks for clarifying that!"
  },
  "source": "meta"
}