![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1723677%2Fadb5c8542e02fb990a6d113efb4be292%2Fe21.png?generation=1582075583564925&alt=media\")
Below are some example words and their respective grapheme root, vowel diacritic, and consonant diacritic displayed separately. In each group of 8 images, they are all the same word.
\n\nThere is a great notebook here that lets you draw the grapheme root and diacritics for each train image. By displaying examples such as below, it helps us understand the challenge that our CNN is facing and lets us build better models. As you see below, finding the pieces within the written word is difficult.
\n\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1723677%2F24fa2287ffd6a4f45d80bda96c33ff56%2Fe61.png?generation=1582075604473522&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1723677%2F61a00e95b869dfdc0a9c898b9c9d2e6b%2Fe103.png?generation=1582075616928841&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1723677%2Fca8b502f80a4839c17f386329a12081c%2Fe22.png?generation=1582075632355465&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1723677%2F3d0b4808be07930a456c461b7b195d38%2Fe79.png?generation=1582075648668289&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1723677%2F65c035b3962dc9ece4cd70d5b70ba309%2Fe167.png?generation=1582075661369858&alt=media\")
Below are some example words and their respective grapheme root, vowel diacritic, and consonant diacritic displayed separately. In each group of 8 images, they are all the same word.
\n\nThere is a great notebook here that lets you draw the grapheme root and diacritics for each train image. By displaying examples such as below, it helps us understand the challenge that our CNN is facing and lets us build better models. As you see below, finding the pieces within the written word is difficult.
\n\n
Excellent post @cdeotte - thanks for sharing! It absolutely helps when you see the handwritten next to the actual root/consonant/vowel. Most of the images I can find the root vowel, but the consonant is harder to spot- unless \"O\" actually means no consonant?
", "rawMarkdown": "Excellent post @cdeotte - thanks for sharing! It absolutely helps when you see the handwritten next to the actual root/consonant/vowel. Most of the images I can find the root vowel, but the consonant is harder to spot- unless \"O\" actually means no consonant?", "votes": null }, { "id": "749954", "postDate": "02/19/2020 03:34:37", "content": "Yes 0 means no consonant diacritic .
", "rawMarkdown": "Yes 0 means no consonant diacritic .", "votes": null }, { "id": "749966", "postDate": "02/19/2020 03:56:35", "content": "That makes sense thanks @phoenix9032
", "rawMarkdown": "That makes sense thanks @phoenix9032", "votes": null }, { "id": "750410", "postDate": "02/19/2020 10:52:46", "content": "Excellent post
", "rawMarkdown": "Excellent post", "votes": null } ], "comments": [ { "id": 749918, "author_name": "robikscube", "author_url": "", "post_date": "02/19/2020 02:14:33", "content": "Excellent post @cdeotte - thanks for sharing! It absolutely helps when you see the handwritten next to the actual root/consonant/vowel. Most of the images I can find the root vowel, but the consonant is harder to spot- unless \"O\" actually means no consonant?
", "votes": null, "replies": [ { "id": 749954, "author_name": "phoenix9032", "author_url": "", "post_date": "02/19/2020 03:34:37", "content": "Yes 0 means no consonant diacritic .
", "votes": null, "replies": [] }, { "id": 749966, "author_name": "robikscube", "author_url": "", "post_date": "02/19/2020 03:56:35", "content": "That makes sense thanks @phoenix9032
", "votes": null, "replies": [] } ] }, { "id": 750410, "author_name": "kaunal11", "author_url": "", "post_date": "02/19/2020 10:52:46", "content": "Excellent post
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "749894": "Below are some example words and their respective grapheme root, vowel diacritic, and consonant diacritic displayed separately. In each group of 8 images, they are all the same word.\n\nThere is a great notebook [here][1] that lets you draw the grapheme root and diacritics for each train image. By displaying examples such as below, it helps us understand the challenge that our CNN is facing and lets us build better models. As you see below, finding the pieces within the written word is difficult.\n\n### Word 1 (r=21, v=1, c=0)\n\n### Word 2 (r=61, v=2, c=0)\n\n\n### Word 3 (r=103, v=1, c=4)\n\n\n### Word 4 (r=22, v=1, c=1)\n\n\n### Word 5 (r=79, v=10, c=0)\n\n\n### Word 6 (r=167, v=5, c=0)\n\n\n[1]: https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-multi-output-cnn", "749918": "Excellent post @cdeotte - thanks for sharing! It absolutely helps when you see the handwritten next to the actual root/consonant/vowel. Most of the images I can find the root vowel, but the consonant is harder to spot- unless \"O\" actually means no consonant?", "749954": "Yes 0 means no consonant diacritic .", "749966": "That makes sense thanks @phoenix9032", "750410": "Excellent post" }, "source": "meta" }