For some of us who did not read the description carefully (including me ), we do not need to rerank the order of code cells. 😯
\n\n", "messages": [ { "id": "1786957", "postDate": "05/13/2022 13:28:55", "content": "Your task is to predict the correct ordering of the cells in a given notebook whose markdown cells have been shuffled.
\n
For some of us who did not read the description carefully (including me ), we do not need to rerank the order of code cells. 😯
\n\n", "rawMarkdown": "For some of us who did not read the description carefully (including me ), we do not need to rerank the order of code cells. 😯\n\n> Your task is to predict the correct ordering of the cells in a given notebook whose **markdown cells have been shuffled**.", "votes": null }, { "id": "1787206", "postDate": "05/13/2022 16:58:50", "content": "Your task is to predict the correct ordering of the cells in a given notebook whose markdown cells have been shuffled.
\n
To clarify, all of the cells need to be ordered, both code and markdown. It's just that you are given the correct relative order of the code cells among themselves. You still need to figure out the correct overall order -- where the markdown cells should be placed among the code cells.
\nSo, the disordered notebook might be:
\ncode_1\ncode_2\ncode_3\nmarkdown_1\nmarkdown_2\nand the correctly ordered notebook might be:
\ncode_1\nmarkdown_2\ncode_2\ncode_3\nmarkdown_1\nThe markdown cells can be in any order, but you would never see code_2 before code_1, for instance.
Might be less interesting than what I tought… I was already thinking about how to feature engineer code. The ranking part still might be interesting.
", "rawMarkdown": "Might be less interesting than what I tought... I was already thinking about how to feature engineer code. The ranking part still might be interesting.", "votes": null }, { "id": "1789846", "postDate": "05/14/2022 09:34:13", "content": "From my local validation, it is very hard to rank the code without domain knowledge.
", "rawMarkdown": "From my local validation, it is very hard to rank the code without domain knowledge.", "votes": null }, { "id": "1793261", "postDate": "05/17/2022 18:12:15", "content": "Or repeatedly invoking the Python interpreter as a loss function ðŸ˜
", "rawMarkdown": "Or repeatedly invoking the Python interpreter as a loss function ðŸ˜", "votes": null }, { "id": "1812803", "postDate": "06/06/2022 08:22:18", "content": "ha ha this is really witty!
", "rawMarkdown": "ha ha this is really witty!", "votes": null }, { "id": "1813424", "postDate": "06/06/2022 19:50:35", "content": "I think the real challenge here is to match natural language in the markdown cells to the code language in the code cells. Limiting it with some domain knowlege will restrict the landscape of solution.
", "rawMarkdown": "I think the real challenge here is to match natural language in the markdown cells to the code language in the code cells. Limiting it with some domain knowlege will restrict the landscape of solution.", "votes": null }, { "id": "1815253", "postDate": "06/08/2022 20:58:44", "content": "Thru Markdown and code, a notebook tells a story. Whilst during development you can jump around a notebook in terms of execution (and state!) which, without a lot of restarting the kernel sometimes makes development a bit non-deterministic (versus CLI/IDE style development), a finished notebook like a document normally reads top to bottom.
\nThis has prompted me to think about how I can maintain an evolving context, for each (best match) markdown cell, as I sequentially work down the code cells. And how to best score that match, given that the relationship between matching markdown and code cells can be one-to-many, and in either direction. In my mind, an evolving context could add a lot of value to the matching objective.
", "rawMarkdown": "Thru Markdown and code, a notebook tells a story. Whilst during development you can jump around a notebook in terms of execution (and state!) which, without a lot of restarting the kernel sometimes makes development a bit non-deterministic (versus CLI/IDE style development), a finished notebook like a document normally reads top to bottom.\n\nThis has prompted me to think about how I can maintain an evolving context, for each (best match) markdown cell, as I sequentially work down the code cells. And how to best score that match, given that the relationship between matching markdown and code cells can be one-to-many, and in either direction. In my mind, an evolving context could add a lot of value to the matching objective.", "votes": null } ], "comments": [ { "id": 1787206, "author_name": "ryanholbrook", "author_url": "", "post_date": "05/13/2022 16:58:50", "content": "To clarify, all of the cells need to be ordered, both code and markdown. It's just that you are given the correct relative order of the code cells among themselves. You still need to figure out the correct overall order -- where the markdown cells should be placed among the code cells.
\nSo, the disordered notebook might be:
\ncode_1\ncode_2\ncode_3\nmarkdown_1\nmarkdown_2\nand the correctly ordered notebook might be:
\ncode_1\nmarkdown_2\ncode_2\ncode_3\nmarkdown_1\nThe markdown cells can be in any order, but you would never see code_2 before code_1, for instance.
Might be less interesting than what I tought… I was already thinking about how to feature engineer code. The ranking part still might be interesting.
", "votes": null, "replies": [ { "id": 1789846, "author_name": "yuanzhezhou", "author_url": "", "post_date": "05/14/2022 09:34:13", "content": "From my local validation, it is very hard to rank the code without domain knowledge.
", "votes": null, "replies": [] }, { "id": 1793261, "author_name": "benvenutto", "author_url": "", "post_date": "05/17/2022 18:12:15", "content": "Or repeatedly invoking the Python interpreter as a loss function ðŸ˜
", "votes": null, "replies": [] }, { "id": 1812803, "author_name": "allohvk", "author_url": "", "post_date": "06/06/2022 08:22:18", "content": "ha ha this is really witty!
", "votes": null, "replies": [] }, { "id": 1813424, "author_name": "mirfanazam", "author_url": "", "post_date": "06/06/2022 19:50:35", "content": "I think the real challenge here is to match natural language in the markdown cells to the code language in the code cells. Limiting it with some domain knowlege will restrict the landscape of solution.
", "votes": null, "replies": [] }, { "id": 1815253, "author_name": "benvenutto", "author_url": "", "post_date": "06/08/2022 20:58:44", "content": "Thru Markdown and code, a notebook tells a story. Whilst during development you can jump around a notebook in terms of execution (and state!) which, without a lot of restarting the kernel sometimes makes development a bit non-deterministic (versus CLI/IDE style development), a finished notebook like a document normally reads top to bottom.
\nThis has prompted me to think about how I can maintain an evolving context, for each (best match) markdown cell, as I sequentially work down the code cells. And how to best score that match, given that the relationship between matching markdown and code cells can be one-to-many, and in either direction. In my mind, an evolving context could add a lot of value to the matching objective.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1786957": "For some of us who did not read the description carefully (including me ), we do not need to rerank the order of code cells. 😯\n\n> Your task is to predict the correct ordering of the cells in a given notebook whose **markdown cells have been shuffled**.", "1787206": "To clarify, all of the cells need to be ordered, both code and markdown. It's just that you are given the correct **relative** order of the code cells among themselves. You still need to figure out the correct **overall** order -- where the markdown cells should be placed among the code cells.\n\nSo, the disordered notebook might be:\n```\ncode_1\ncode_2\ncode_3\nmarkdown_1\nmarkdown_2\n```\nand the correctly ordered notebook might be:\n```\ncode_1\nmarkdown_2\ncode_2\ncode_3\nmarkdown_1\n```\nThe markdown cells can be in any order, but you would never see `code_2` before `code_1`, for instance.", "1789729": "Might be less interesting than what I tought... I was already thinking about how to feature engineer code. The ranking part still might be interesting.", "1789846": "From my local validation, it is very hard to rank the code without domain knowledge.", "1793261": "Or repeatedly invoking the Python interpreter as a loss function ðŸ˜", "1812803": "ha ha this is really witty!", "1813424": "I think the real challenge here is to match natural language in the markdown cells to the code language in the code cells. Limiting it with some domain knowlege will restrict the landscape of solution.", "1815253": "Thru Markdown and code, a notebook tells a story. Whilst during development you can jump around a notebook in terms of execution (and state!) which, without a lot of restarting the kernel sometimes makes development a bit non-deterministic (versus CLI/IDE style development), a finished notebook like a document normally reads top to bottom.\n\nThis has prompted me to think about how I can maintain an evolving context, for each (best match) markdown cell, as I sequentially work down the code cells. And how to best score that match, given that the relationship between matching markdown and code cells can be one-to-many, and in either direction. In my mind, an evolving context could add a lot of value to the matching objective." }, "source": "meta" }