{ "id": 228216, "title": "Welcome to Show US the Data", "url": "/competitions/coleridgeinitiative-show-us-the-data/discussion/228216", "author_name": "Julia Lane", "post_date": "2021-03-23T20:12:48.895000", "votes": 34, "comment_count": 16, "views": 0, "content": "

Welcome to the Show US the data competition

\n

We are thrilled that you are interested in working on this challenge - to help us make the use of public data and evidence more transparent.

\n

We are a not-for-profit startup that is dedicated to improving the use of data and evidence for the public good. We face many threats that governments must address with data: pandemics, climate change and coastal inundation, Alzheimer’s disease, child hunger, threatened biodiversity, and many others. Yet much of the information about data necessary to inform evidence and science is locked inside publications. The results of this competition can tie evidence to action - If data scientists can find how datasets are used in publications quickly, and at scale, the government can catalog answers to our most pressing questions and apply them to the world’s critical issues.

\n

Beyond the prizes, your work will help governments do their job better. Governments the world over are being force to modernize their data management practices and be more open and transparent in how data are used. You can help build scientific basis for evidence based decision making.

\n

Our team worked super hard to make this competition happen. We'll be around to answer any questions. We look forward to seeing how you do!

\n

Hosts:
\nThe Coleridge Initiative

\n\n

And thanks to our collaborators at Kaggle

\n", "messages": [ { "id": 1250164, "postDate": "2021-03-23T20:12:48.897Z", "content": "

Welcome to the Show US the data competition

\n

We are thrilled that you are interested in working on this challenge - to help us make the use of public data and evidence more transparent.

\n

We are a not-for-profit startup that is dedicated to improving the use of data and evidence for the public good. We face many threats that governments must address with data: pandemics, climate change and coastal inundation, Alzheimer’s disease, child hunger, threatened biodiversity, and many others. Yet much of the information about data necessary to inform evidence and science is locked inside publications. The results of this competition can tie evidence to action - If data scientists can find how datasets are used in publications quickly, and at scale, the government can catalog answers to our most pressing questions and apply them to the world’s critical issues.

\n

Beyond the prizes, your work will help governments do their job better. Governments the world over are being force to modernize their data management practices and be more open and transparent in how data are used. You can help build scientific basis for evidence based decision making.

\n

Our team worked super hard to make this competition happen. We'll be around to answer any questions. We look forward to seeing how you do!

\n

Hosts:
\nThe Coleridge Initiative

\n\n

And thanks to our collaborators at Kaggle

\n", "rawMarkdown": "## Welcome to the Show US the data competition\n\nWe are thrilled that you are interested in working on this challenge - to help us make the use of public data and evidence more transparent.\n\nWe are a not-for-profit startup that is dedicated to improving the use of data and evidence for the public good. We face many threats that governments must address with data: pandemics, climate change and coastal inundation, Alzheimer’s disease, child hunger, threatened biodiversity, and many others. Yet much of the information about data necessary to inform evidence and science is locked inside publications. The results of this competition can tie evidence to action - If data scientists can find how datasets are used in publications quickly, and at scale, the government can catalog answers to our most pressing questions and apply them to the world’s critical issues.\n\nBeyond the prizes, your work will help governments do their job better. Governments the world over are being force to modernize their data management practices and be more open and transparent in how data are used. You can help build scientific basis for evidence based decision making. \n\nOur team worked super hard to make this competition happen. We'll be around to answer any questions. We look forward to seeing how you do! \n\n**Hosts:**\n*The Coleridge Initiative*\n- Julia Lane (@julialane) \n- Ernesto Gimeno (@ernestogimeno)\n- Haritz (@haritzpuerto)\n\nAnd thanks to our collaborators at Kaggle\n- Maggie Demkin (@maggiemd)\n- Phil Culliton (@philculliton)\n\n", "votes": 33 }, { "id": 1250983, "postDate": "2021-03-24T12:07:25.320Z", "content": "

There is no dataleak. There seems to be an error in the leaderboard. We are waiting for @philculliton to fix it.

", "rawMarkdown": "There is no dataleak. There seems to be an error in the leaderboard. We are waiting for @philculliton to fix it. ", "votes": 4, "replies": [ { "id": 1250992, "postDate": "2021-03-24T12:17:52.423Z", "content": "

@julialane That's good to hear! Please let me know if you need any help in clarifying what my code does so that you can more rapidly debug the issue. For reference, here is my notebook that achieved a perfect 1.0 score.

", "rawMarkdown": "@julialane That's good to hear! Please let me know if you need any help in clarifying what my code does so that you can more rapidly debug the issue. For reference, here is [my notebook](https://www.kaggle.com/josephassaker/coleridge-initiative-eda-na-ve-submission) that achieved a perfect 1.0 score.", "votes": 3 } ] }, { "id": 1251564, "postDate": "2021-03-24T22:54:11.427Z", "content": "

Thanks to all - especially to @philculliton - for being patient with the launch blips. I should also have mentioned that @michaelstebbins is part of our team. Many of you will know him for being the author of the Holdren Open Access memo under the Obama Administration. Much of this competition is inspired by those ideas

", "rawMarkdown": "Thanks to all - especially to @philculliton - for being patient with the launch blips. I should also have mentioned that @michaelstebbins is part of our team. Many of you will know him for being the author of the [Holdren Open Access memo](https://obamawhitehouse.archives.gov/blog/2013/02/22/expanding-public-access-results-federally-funded-research) under the Obama Administration. Much of this competition is inspired by those ideas ", "votes": 1 }, { "id": 1250733, "postDate": "2021-03-24T08:37:37.507Z", "content": "

The first place is 0.997, and I feel that this game is not interesting. 🤒

", "rawMarkdown": "The first place is 0.997, and I feel that this game is not interesting. 🤒", "votes": 1, "replies": [ { "id": 1250872, "postDate": "2021-03-24T10:30:50.193Z", "content": "

Well now we got someone scoring 1.000 on the first day ( more like first 12 hours ) .

", "rawMarkdown": "Well now we got someone scoring 1.000 on the first day ( more like first 12 hours ) ." }, { "id": 1250946, "postDate": "2021-03-24T11:33:00.997Z", "content": "

@zy1343930734 Please don't get discouraged! Check out my notebook to see how you can very easily achieve 1.0 score with a naïve approach.

\n

Keep in mind that the primary goal of my notebook is EDA, but I somehow stumbled across a 1.0 score from the first attempt 😅

", "rawMarkdown": "@zy1343930734 Please don't get discouraged! Check out [my notebook](https://www.kaggle.com/josephassaker/coleridge-initiative-eda-na-ve-submission) to see how you can very easily achieve 1.0 score with a naïve approach.\n\nKeep in mind that the primary goal of my notebook is EDA, but I somehow stumbled across a 1.0 score from the first attempt 😅", "votes": 3 }, { "id": 1250947, "postDate": "2021-03-24T11:35:14.677Z", "content": "

Thank you!!! Come on.

", "rawMarkdown": "Thank you!!! Come on." }, { "id": 1250949, "postDate": "2021-03-24T11:35:38.577Z", "content": "

lol 😅gogogo

", "rawMarkdown": "lol 😅gogogo", "votes": 1 } ] }, { "id": 1250984, "postDate": "2021-03-24T12:07:59.783Z", "content": "

And thank you @kamaldas335

", "rawMarkdown": "And thank you @kamaldas335 ", "votes": 2 }, { "id": 1250855, "postDate": "2021-03-24T10:18:22.363Z", "content": "

Hi @julialane, @ernestogimeno, @haritzpuerto and Kaggle team @maggiemd, @philculliton,

\n

As @ammarali32 and @josephassaker have shown

\n

Notebook: https://www.kaggle.com/josephassaker/coleridge-initiative-eda-na-ve-submission
\ndataset: https://www.kaggle.com/ammarali32/coleridge-dataset-labels-matching
\nand
\nthread: https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228347

\n

appears we may get 1.000 using a Naïve Dataset Title Matching Submission and /or external dataset

\n

we need to relook at the metrics/dataset

\n

Thanks!

", "rawMarkdown": "Hi @julialane, @ernestogimeno, @haritzpuerto and Kaggle team @maggiemd, @philculliton,\n\nAs @ammarali32 and @josephassaker have shown\n\nNotebook: https://www.kaggle.com/josephassaker/coleridge-initiative-eda-na-ve-submission\ndataset: https://www.kaggle.com/ammarali32/coleridge-dataset-labels-matching\nand\nthread: https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228347\n\nappears we may get 1.000 using a Naïve Dataset Title Matching Submission and /or external dataset\n\nwe need to relook at the metrics/dataset\n\nThanks!", "votes": 2, "replies": [ { "id": 1250884, "postDate": "2021-03-24T10:35:30.413Z", "content": "

Totally agreed! I would suggest to further diversify the test set to be more representative. That being said, I'm not sad about my weirdo temporary position 1 in the leaderboard 😅

", "rawMarkdown": "Totally agreed! I would suggest to further diversify the test set to be more representative. That being said, I'm not sad about my weirdo temporary position 1 in the leaderboard 😅", "votes": 3 }, { "id": 1250892, "postDate": "2021-03-24T10:41:07.577Z", "content": "

The leaderboard standing are based on 12% of the data so the final standing will most probably differ . So i think increasing it to atleast 30% will help.

", "rawMarkdown": "The leaderboard standing are based on 12% of the data so the final standing will most probably differ . So i think increasing it to atleast 30% will help.", "votes": 3 }, { "id": 1250948, "postDate": "2021-03-24T11:35:19.957Z", "content": "

Releasing more data won't help on its own, but could be part of a solution. For the leaderboard to work well, the training, test and hidden test data sets must not have labels in common.

\n

This data leak seems fixable (e.g. by re-sorting the documents among the data sets, and/or by psudeonomysing the labels in the hidden test set).

\n

Let me know if I can help @julialane, @ernestogimeno, @haritzpuerto. I recognise that if I help fix the data leak I can't compete.

", "rawMarkdown": "Releasing more data won't help on its own, but could be part of a solution. For the leaderboard to work well, the training, test and hidden test data sets must not have labels in common.\n\nThis data leak seems fixable (e.g. by re-sorting the documents among the data sets, and/or by psudeonomysing the labels in the hidden test set).\n\nLet me know if I can help @julialane, @ernestogimeno, @haritzpuerto. I recognise that if I help fix the data leak I can't compete.", "votes": 3 }, { "id": 1250950, "postDate": "2021-03-24T11:36:09.263Z", "content": "

Well yeah, I totally agree. Because even if the private data have some external datasets. The public leaderboard should carry some useful information. Also, I think it is better to increase the external data because in 12% of the data there is none even one external sample that could indicate that the number of external datasets is small. And that is also a problem because we can easily gather external datasets and transfer the problem to string matching again. They should also change the metric I guess.

", "rawMarkdown": "Well yeah, I totally agree. Because even if the private data have some external datasets. The public leaderboard should carry some useful information. Also, I think it is better to increase the external data because in 12% of the data there is none even one external sample that could indicate that the number of external datasets is small. And that is also a problem because we can easily gather external datasets and transfer the problem to string matching again. They should also change the metric I guess.", "votes": 2 }, { "id": 1250951, "postDate": "2021-03-24T11:37:09.803Z", "content": "

I agree you

", "rawMarkdown": "I agree you", "votes": 1 } ] }, { "id": 1250941, "postDate": "2021-03-24T11:30:24.213Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1250983, "author_name": "Julia Lane", "author_url": "", "post_date": "2021-03-24T12:07:25.320000", "content": "

There is no dataleak. There seems to be an error in the leaderboard. We are waiting for @philculliton to fix it.

", "votes": 4, "replies": [ { "id": 1250992, "author_name": "Joseph Assaker", "author_url": "", "post_date": "2021-03-24T12:17:52.423000", "content": "

@julialane That's good to hear! Please let me know if you need any help in clarifying what my code does so that you can more rapidly debug the issue. For reference, here is my notebook that achieved a perfect 1.0 score.

", "votes": 3, "replies": [] } ] }, { "id": 1251564, "author_name": "Julia Lane", "author_url": "", "post_date": "2021-03-24T22:54:11.427000", "content": "

Thanks to all - especially to @philculliton - for being patient with the launch blips. I should also have mentioned that @michaelstebbins is part of our team. Many of you will know him for being the author of the Holdren Open Access memo under the Obama Administration. Much of this competition is inspired by those ideas

", "votes": 1, "replies": [] }, { "id": 1250733, "author_name": "DECEM", "author_url": "", "post_date": "2021-03-24T08:37:37.507000", "content": "

The first place is 0.997, and I feel that this game is not interesting. 🤒

", "votes": 1, "replies": [ { "id": 1250872, "author_name": "Puru Behl", "author_url": "", "post_date": "2021-03-24T10:30:50.193000", "content": "

Well now we got someone scoring 1.000 on the first day ( more like first 12 hours ) .

", "votes": 0, "replies": [] }, { "id": 1250946, "author_name": "Joseph Assaker", "author_url": "", "post_date": "2021-03-24T11:33:00.997000", "content": "

@zy1343930734 Please don't get discouraged! Check out my notebook to see how you can very easily achieve 1.0 score with a naïve approach.

\n

Keep in mind that the primary goal of my notebook is EDA, but I somehow stumbled across a 1.0 score from the first attempt 😅

", "votes": 3, "replies": [] }, { "id": 1250947, "author_name": "DECEM", "author_url": "", "post_date": "2021-03-24T11:35:14.677000", "content": "

Thank you!!! Come on.

", "votes": 0, "replies": [] }, { "id": 1250949, "author_name": "DECEM", "author_url": "", "post_date": "2021-03-24T11:35:38.577000", "content": "

lol 😅gogogo

", "votes": 1, "replies": [] } ] }, { "id": 1250984, "author_name": "Julia Lane", "author_url": "", "post_date": "2021-03-24T12:07:59.783000", "content": "

And thank you @kamaldas335

", "votes": 2, "replies": [] }, { "id": 1250855, "author_name": "Kamal Das", "author_url": "", "post_date": "2021-03-24T10:18:22.363000", "content": "

Hi @julialane, @ernestogimeno, @haritzpuerto and Kaggle team @maggiemd, @philculliton,

\n

As @ammarali32 and @josephassaker have shown

\n

Notebook: https://www.kaggle.com/josephassaker/coleridge-initiative-eda-na-ve-submission
\ndataset: https://www.kaggle.com/ammarali32/coleridge-dataset-labels-matching
\nand
\nthread: https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228347

\n

appears we may get 1.000 using a Naïve Dataset Title Matching Submission and /or external dataset

\n

we need to relook at the metrics/dataset

\n

Thanks!

", "votes": 2, "replies": [ { "id": 1250884, "author_name": "Joseph Assaker", "author_url": "", "post_date": "2021-03-24T10:35:30.413000", "content": "

Totally agreed! I would suggest to further diversify the test set to be more representative. That being said, I'm not sad about my weirdo temporary position 1 in the leaderboard 😅

", "votes": 3, "replies": [] }, { "id": 1250892, "author_name": "Puru Behl", "author_url": "", "post_date": "2021-03-24T10:41:07.577000", "content": "

The leaderboard standing are based on 12% of the data so the final standing will most probably differ . So i think increasing it to atleast 30% will help.

", "votes": 3, "replies": [] }, { "id": 1250948, "author_name": "Nicholas James Bailey", "author_url": "", "post_date": "2021-03-24T11:35:19.957000", "content": "

Releasing more data won't help on its own, but could be part of a solution. For the leaderboard to work well, the training, test and hidden test data sets must not have labels in common.

\n

This data leak seems fixable (e.g. by re-sorting the documents among the data sets, and/or by psudeonomysing the labels in the hidden test set).

\n

Let me know if I can help @julialane, @ernestogimeno, @haritzpuerto. I recognise that if I help fix the data leak I can't compete.

", "votes": 3, "replies": [] }, { "id": 1250950, "author_name": "ammarali32", "author_url": "", "post_date": "2021-03-24T11:36:09.263000", "content": "

Well yeah, I totally agree. Because even if the private data have some external datasets. The public leaderboard should carry some useful information. Also, I think it is better to increase the external data because in 12% of the data there is none even one external sample that could indicate that the number of external datasets is small. And that is also a problem because we can easily gather external datasets and transfer the problem to string matching again. They should also change the metric I guess.

", "votes": 2, "replies": [] }, { "id": 1250951, "author_name": "DECEM", "author_url": "", "post_date": "2021-03-24T11:37:09.803000", "content": "

I agree you

", "votes": 1, "replies": [] } ] }, { "id": 1250941, "author_name": "", "author_url": "", "post_date": "2021-03-24T11:30:24.213000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1250164": "## Welcome to the Show US the data competition\n\nWe are thrilled that you are interested in working on this challenge - to help us make the use of public data and evidence more transparent.\n\nWe are a not-for-profit startup that is dedicated to improving the use of data and evidence for the public good. We face many threats that governments must address with data: pandemics, climate change and coastal inundation, Alzheimer’s disease, child hunger, threatened biodiversity, and many others. Yet much of the information about data necessary to inform evidence and science is locked inside publications. The results of this competition can tie evidence to action - If data scientists can find how datasets are used in publications quickly, and at scale, the government can catalog answers to our most pressing questions and apply them to the world’s critical issues.\n\nBeyond the prizes, your work will help governments do their job better. Governments the world over are being force to modernize their data management practices and be more open and transparent in how data are used. You can help build scientific basis for evidence based decision making. \n\nOur team worked super hard to make this competition happen. We'll be around to answer any questions. We look forward to seeing how you do! \n\n**Hosts:**\n*The Coleridge Initiative*\n- Julia Lane (@julialane) \n- Ernesto Gimeno (@ernestogimeno)\n- Haritz (@haritzpuerto)\n\nAnd thanks to our collaborators at Kaggle\n- Maggie Demkin (@maggiemd)\n- Phil Culliton (@philculliton)\n\n", "1250983": "There is no dataleak. There seems to be an error in the leaderboard. We are waiting for @philculliton to fix it. ", "1251564": "Thanks to all - especially to @philculliton - for being patient with the launch blips. I should also have mentioned that @michaelstebbins is part of our team. Many of you will know him for being the author of the [Holdren Open Access memo](https://obamawhitehouse.archives.gov/blog/2013/02/22/expanding-public-access-results-federally-funded-research) under the Obama Administration. Much of this competition is inspired by those ideas ", "1250733": "The first place is 0.997, and I feel that this game is not interesting. 🤒", "1250984": "And thank you @kamaldas335 ", "1250855": "Hi @julialane, @ernestogimeno, @haritzpuerto and Kaggle team @maggiemd, @philculliton,\n\nAs @ammarali32 and @josephassaker have shown\n\nNotebook: https://www.kaggle.com/josephassaker/coleridge-initiative-eda-na-ve-submission\ndataset: https://www.kaggle.com/ammarali32/coleridge-dataset-labels-matching\nand\nthread: https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228347\n\nappears we may get 1.000 using a Naïve Dataset Title Matching Submission and /or external dataset\n\nwe need to relook at the metrics/dataset\n\nThanks!", "1250941": "" } }