Dear Kaggle Administration and Competition Organizers,
\nWe wish to inquire about the use of an external data source, the TUH EEG Seizure Corpus[1], in the context of this competition. Our understanding is that this dataset requires direct confirmation from the distributing entity via email, which we believe may conflict with the competition's rules under Section 7.C (External Data).
\n\n\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).
\n
Despite this, we have observed that the dataset has been previously uploaded to Kaggle Datasets by a third party[2] and is currently being referenced and used in public notebooks as well as discussions[3-6]. This raises a question for us: Even if a dataset is made \"available to everyone\" by a third party, does using it without the distributor's permission violate Kaggle's competition rules? We seek to clarify this point with the organizers, particularly because the competition rules in Section 11 (Winners Obligations) state that \"all programs must include commercially usable source code,\" and we understand that using an external dataset with unclear commercial usage rights might contravene this rule.
\n\n\nTo the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;
\n
We look forward to your clarification on this matter.
\nBest regards.
\nDear Kaggle Administration and Competition Organizers,
\nWe wish to inquire about the use of an external data source, the TUH EEG Seizure Corpus[1], in the context of this competition. Our understanding is that this dataset requires direct confirmation from the distributing entity via email, which we believe may conflict with the competition's rules under Section 7.C (External Data).
\n\n\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).
\n
Despite this, we have observed that the dataset has been previously uploaded to Kaggle Datasets by a third party[2] and is currently being referenced and used in public notebooks as well as discussions[3-6]. This raises a question for us: Even if a dataset is made \"available to everyone\" by a third party, does using it without the distributor's permission violate Kaggle's competition rules? We seek to clarify this point with the organizers, particularly because the competition rules in Section 11 (Winners Obligations) state that \"all programs must include commercially usable source code,\" and we understand that using an external dataset with unclear commercial usage rights might contravene this rule.
\n\n\nTo the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;
\n
We look forward to your clarification on this matter.
\nBest regards.
\nI could not find exact license information of TUH dataset, but I can see a statement in the download page.
\n\n\nThe TUH EEG Corpus is freely available. The only reason we require registration is that we need to track who downloads the data. We also want to be able to inform you of any updates to the releases.
\n
I guess this dataset is allowed to use freely, and hence we can use it, but I am not sure.
", "rawMarkdown": "I could not find exact license information of TUH dataset, but I can see a statement in the [download page](https://isip.piconepress.com/projects/tuh_eeg/html/downloads.shtml).\n>The TUH EEG Corpus is freely available. The only reason we require registration is that we need to track who downloads the data. We also want to be able to inform you of any updates to the releases. \n\nI guess this dataset is allowed to use freely, and hence we can use it, but I am not sure.", "votes": null }, { "id": "2714816", "postDate": "03/25/2024 05:23:42", "content": "@tomooinubushi Oh, I missed that. Thank you for sharing!
", "rawMarkdown": "tomooinubushi Oh, I missed that. Thank you for sharing!", "votes": null }, { "id": "2716200", "postDate": "03/25/2024 22:05:57", "content": "I would appreciate some clarification from the organisers about this. I too applied for access and downloaded the corpus but haven't uploaded it due to the following clauses in the registration form:
\n\n\nThe data available from the NEDC web site is offered free and without restriction, except for a few conditions below that you must abide by:
\n
\n (1)The data recipient should acknowledge the provider of this data using the publication listed in the documentation included with the specific corpus (typically found in the AAREADME file).
\n (2)The data recipient will not release data to a third party. Please have the third party contact us directly by sending an email to help@nedcdata.org.
\n (3)The data recipient will delete the data from all computer systems when finished with the data.
The second point is crucial; releasing the dataset publicly on Kaggle may well be a violation of their ToS. Even if the dataset has been deleted after this competition, third parties may still have downloaded it without going through the official channels.
", "rawMarkdown": "I would appreciate some clarification from the organisers about this. I too applied for access and downloaded the corpus but haven't uploaded it due to the following clauses in the registration form:\n\n>The data available from the NEDC web site is offered free and without restriction, except for a few conditions below that you must abide by: \n(1)The data recipient should acknowledge the provider of this data using the publication listed in the documentation included with the specific corpus (typically found in the AAREADME file). \n(2)The data recipient will not release data to a third party. Please have the third party contact us directly by sending an email to help@nedcdata.org. \n(3)The data recipient will delete the data from all computer systems when finished with the data.\n\nThe second point is crucial; releasing the dataset publicly on Kaggle may well be a violation of their ToS. Even if the dataset has been deleted after this competition, third parties may still have downloaded it without going through the official channels.", "votes": null }, { "id": "2716233", "postDate": "03/25/2024 22:38:06", "content": "Thanks for the questions! We unfortunately could not provide any legal guidance nor interpretation of license of specific dataset. As per the rules, as long as the External Data is \"is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants\" and make sure \"the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense\", they can be used. Hope it helps clarify. Thanks!
", "rawMarkdown": "Thanks for the questions! We unfortunately could not provide any legal guidance nor interpretation of license of specific dataset. As per the rules, as long as the External Data is \"is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants\" and make sure \"the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense\", they can be used. Hope it helps clarify. Thanks!", "votes": null }, { "id": "2716388", "postDate": "03/26/2024 02:31:56", "content": "\n\n\nWe unfortunately could not provide any legal guidance nor interpretation of license of specific dataset.
\n
We are not requesting legal guidance, but we just want to be confirmed competition rule.
\nIf availability of data is considered differently by participant by participant, it will ruin the competitiveness in this competition.
\nWe try to collect evidence as good as possible, so, could you confirm the availability before the competition ends?
\nOr, just providing general guidance is welcome.
\nFor example, as @bigironsphere mentioned, “a dataset that distributor cannot accept distribution to third party is not available in Kaggle, even if someone uploaded the dataset on Kaggle” etc.
Think this is a known issue with what shows up in Datasets on Kaggle. Have seen examples of what would be copyright infringement in some cases and as you have shown is not in accordance with the original source's requirements for use or republishing. Seems like Datasets needs a better way of dealing with what is uploaded wrt license, terms of use. There are some checks in place for objectionable content (more for images but perhaps text too).
\nMaybe the easiest option is for one person making use of the TUH EEG Seizure Corpus to email the address in (2) and just get permission to host it on Kaggle for the duration of this competition?
", "rawMarkdown": "Think this is a known issue with what shows up in Datasets on Kaggle. Have seen examples of what would be copyright infringement in some cases and as you have shown is not in accordance with the original source's requirements for use or republishing. Seems like Datasets needs a better way of dealing with what is uploaded wrt license, terms of use. There are some checks in place for objectionable content (more for images but perhaps text too). \n\nMaybe the easiest option is for one person making use of the TUH EEG Seizure Corpus to email the address in (2) and just get permission to host it on Kaggle for the duration of this competition?", "votes": null }, { "id": "2718136", "postDate": "03/27/2024 01:29:43", "content": "Thanks @tatamikenn! Since competition rules are the agreement between the competition hosts and participants, Kaggle cannot make legal interpretations nor provide guidance on the rules. Let me reach out to the competition hosts for them to chime in. Thanks!
", "rawMarkdown": "Thanks @tatamikenn! Since competition rules are the agreement between the competition hosts and participants, Kaggle cannot make legal interpretations nor provide guidance on the rules. Let me reach out to the competition hosts for them to chime in. Thanks!", "votes": null }, { "id": "2718149", "postDate": "03/27/2024 02:02:47", "content": "@ashleychow Thanks. Could we think the availability of the dataset will be announced in this thread after settlement between competition host and Kaggle team?
", "rawMarkdown": "ashleychow Thanks. Could we think the availability of the dataset will be announced in this thread after settlement between competition host and Kaggle team?", "votes": null }, { "id": "2720491", "postDate": "03/28/2024 11:48:48", "content": "So can someone confirm if at the end of this competition, I can include the data for this competition and maybe a link here to maybe my Git repository along with my script?
", "rawMarkdown": "So can someone confirm if at the end of this competition, I can include the data for this competition and maybe a link here to maybe my Git repository along with my script?", "votes": null }, { "id": "2722666", "postDate": "03/29/2024 17:47:44", "content": "Thanks for your patience! After discussing with the host, as per the rules:
\n\n\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).
\n
\n …
\n As a condition to being awarded a Prize, a Prize winner must fulfill the following obligations:(a) deliver to the Competition Sponsor the final model's software code as used to generate the winning Submission and associated documentation. The delivered software code should follow these documentation guidelines, must be capable of generating the winning Submission, and contain a description of resources required to build and/or run the executable code successfully. To the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;
\n
So if our winning solution is a \"recipe to the cake\", the Datasets being used will be part of the \"recipe\". The winners are required to deliver the \"recipe\" to the host to reproduce the results & validate later. If the usage of this dataset would not grant you the license of Apache 2.0/ \"commercially available software\", then this dataset should not be used.
\nIn other words, using TUH data would be allowed only if the team provides the external data that was used as \"part of the package\" when you submit your code. Since the host will plan to verify the winning solution by running the code, if the winning solution cannot be run as provided (i.e. if data needed to run it is not included), this would disqualify it as a winning entry.
\nIf the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.
\nHope it helps clarify. Thanks!
", "rawMarkdown": "Thanks for your patience! After discussing with the host, as per the rules:\n\n>C. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).\n...\n>As a condition to being awarded a Prize, a Prize winner must fulfill the following obligations:\n\n>(a) deliver to the Competition Sponsor the final model's software code as used to generate the winning Submission and associated documentation. The delivered software code should follow these documentation guidelines, must be capable of generating the winning Submission, and contain a description of resources required to build and/or run the executable code successfully. To the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;\n\nSo if our winning solution is a \"recipe to the cake\", the Datasets being used will be part of the \"recipe\". The winners are required to deliver the \"recipe\" to the host to reproduce the results & validate later. If the usage of this dataset would not grant you the license of Apache 2.0/ \"commercially available software\", then this dataset should not be used. \n\nIn other words, using TUH data would be allowed only if the team provides the external data that was used as \"part of the package\" when you submit your code. Since the host will plan to verify the winning solution by running the code, if the winning solution cannot be run as provided (i.e. if data needed to run it is not included), this would disqualify it as a winning entry.\n\nIf the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.\n\nHope it helps clarify. Thanks!", "votes": null }, { "id": "2722971", "postDate": "03/29/2024 22:30:57", "content": "@ashleychow
\nThank you for the clarification.
I have a question for the last part.
\n\n\nIf the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.
\n
If we use the example of \"recipe\", we can consider the external dataset as \"ingredients\". And at this point, if we can download the TUH dataset only for sending email and confirm the rule as @bigironsphere mentioned, can we think the dataset as \"publicly available and equally accessible\"? The rule says only restrict 1. acknowledgement, 2. redistribution, and 3. deletion. Thus, we can consider the \"recipe\" using this \"ingredients\" are reproducible for host unless they follow these restrictions. Or, do the host plan to redistribute this dataset?
", "rawMarkdown": "ashleychow \nThank you for the clarification.\n\nI have a question for the last part.\n\n> If the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.\n\nIf we use the example of \"recipe\", we can consider the external dataset as \"ingredients\". And at this point, if we can download the TUH dataset only for sending email and confirm the rule as @bigironsphere mentioned, can we think the dataset as \"publicly available and equally accessible\"? The rule says only restrict 1. acknowledgement, 2. redistribution, and 3. deletion. Thus, we can consider the \"recipe\" using this \"ingredients\" are reproducible for host unless they follow these restrictions. Or, do the host plan to redistribute this dataset?", "votes": null }, { "id": "2727870", "postDate": "04/02/2024 01:20:05", "content": "Thanks @tatamikenn! I am checking with the host and will keep you posted. Thanks!
", "rawMarkdown": "Thanks @tatamikenn! I am checking with the host and will keep you posted. Thanks!", "votes": null }, { "id": "2728067", "postDate": "04/02/2024 04:49:43", "content": "@ashleychow Thank you for your reply. However, considering that there is only one week left until the end of the competition, we are concerned about the possibility of changing the policy regarding the use of external data at this point. We have no particular objections to the decision that it cannot be used in this competition, and it is likely that other teams are also developing their solutions with the understanding that it cannot be used. Changing the policy now could lead to confusion, and we believe it is better not to alter the decision.
", "rawMarkdown": "ashleychow Thank you for your reply. However, considering that there is only one week left until the end of the competition, we are concerned about the possibility of changing the policy regarding the use of external data at this point. We have no particular objections to the decision that it cannot be used in this competition, and it is likely that other teams are also developing their solutions with the understanding that it cannot be used. Changing the policy now could lead to confusion, and we believe it is better not to alter the decision.", "votes": null }, { "id": "2734029", "postDate": "04/03/2024 22:04:16", "content": "Hi Bilzard, as far as I can tell, TUH is not technically a public dataset. Therefore, using it for this contest is not allowed. - Brandon
", "rawMarkdown": "Hi Bilzard, as far as I can tell, TUH is not technically a public dataset. Therefore, using it for this contest is not allowed. - Brandon", "votes": null }, { "id": "2734056", "postDate": "04/03/2024 22:49:00", "content": "@damienaltzheim Thanks. I understand.
", "rawMarkdown": "damienaltzheim Thanks. I understand.", "votes": null } ], "comments": [ { "id": 2714725, "author_name": "tomooinubushi", "author_url": "", "post_date": "03/25/2024 03:48:34", "content": "I could not find exact license information of TUH dataset, but I can see a statement in the download page.
\n\n\nThe TUH EEG Corpus is freely available. The only reason we require registration is that we need to track who downloads the data. We also want to be able to inform you of any updates to the releases.
\n
I guess this dataset is allowed to use freely, and hence we can use it, but I am not sure.
", "votes": null, "replies": [ { "id": 2714816, "author_name": "tatamikenn", "author_url": "", "post_date": "03/25/2024 05:23:42", "content": "@tomooinubushi Oh, I missed that. Thank you for sharing!
", "votes": null, "replies": [] } ] }, { "id": 2716200, "author_name": "bigironsphere", "author_url": "", "post_date": "03/25/2024 22:05:57", "content": "I would appreciate some clarification from the organisers about this. I too applied for access and downloaded the corpus but haven't uploaded it due to the following clauses in the registration form:
\n\n\nThe data available from the NEDC web site is offered free and without restriction, except for a few conditions below that you must abide by:
\n
\n (1)The data recipient should acknowledge the provider of this data using the publication listed in the documentation included with the specific corpus (typically found in the AAREADME file).
\n (2)The data recipient will not release data to a third party. Please have the third party contact us directly by sending an email to help@nedcdata.org.
\n (3)The data recipient will delete the data from all computer systems when finished with the data.
The second point is crucial; releasing the dataset publicly on Kaggle may well be a violation of their ToS. Even if the dataset has been deleted after this competition, third parties may still have downloaded it without going through the official channels.
", "votes": null, "replies": [ { "id": 2716727, "author_name": "something4kag", "author_url": "", "post_date": "03/26/2024 07:17:36", "content": "Think this is a known issue with what shows up in Datasets on Kaggle. Have seen examples of what would be copyright infringement in some cases and as you have shown is not in accordance with the original source's requirements for use or republishing. Seems like Datasets needs a better way of dealing with what is uploaded wrt license, terms of use. There are some checks in place for objectionable content (more for images but perhaps text too).
\nMaybe the easiest option is for one person making use of the TUH EEG Seizure Corpus to email the address in (2) and just get permission to host it on Kaggle for the duration of this competition?
", "votes": null, "replies": [] } ] }, { "id": 2716233, "author_name": "ashleychow", "author_url": "", "post_date": "03/25/2024 22:38:06", "content": "Thanks for the questions! We unfortunately could not provide any legal guidance nor interpretation of license of specific dataset. As per the rules, as long as the External Data is \"is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants\" and make sure \"the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense\", they can be used. Hope it helps clarify. Thanks!
", "votes": null, "replies": [ { "id": 2716388, "author_name": "tatamikenn", "author_url": "", "post_date": "03/26/2024 02:31:56", "content": "\n\n\nWe unfortunately could not provide any legal guidance nor interpretation of license of specific dataset.
\n
We are not requesting legal guidance, but we just want to be confirmed competition rule.
\nIf availability of data is considered differently by participant by participant, it will ruin the competitiveness in this competition.
\nWe try to collect evidence as good as possible, so, could you confirm the availability before the competition ends?
\nOr, just providing general guidance is welcome.
\nFor example, as @bigironsphere mentioned, “a dataset that distributor cannot accept distribution to third party is not available in Kaggle, even if someone uploaded the dataset on Kaggle” etc.
Thanks @tatamikenn! Since competition rules are the agreement between the competition hosts and participants, Kaggle cannot make legal interpretations nor provide guidance on the rules. Let me reach out to the competition hosts for them to chime in. Thanks!
", "votes": null, "replies": [ { "id": 2718149, "author_name": "tatamikenn", "author_url": "", "post_date": "03/27/2024 02:02:47", "content": "@ashleychow Thanks. Could we think the availability of the dataset will be announced in this thread after settlement between competition host and Kaggle team?
", "votes": null, "replies": [ { "id": 2722666, "author_name": "ashleychow", "author_url": "", "post_date": "03/29/2024 17:47:44", "content": "Thanks for your patience! After discussing with the host, as per the rules:
\n\n\nC. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).
\n
\n …
\n As a condition to being awarded a Prize, a Prize winner must fulfill the following obligations:(a) deliver to the Competition Sponsor the final model's software code as used to generate the winning Submission and associated documentation. The delivered software code should follow these documentation guidelines, must be capable of generating the winning Submission, and contain a description of resources required to build and/or run the executable code successfully. To the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;
\n
So if our winning solution is a \"recipe to the cake\", the Datasets being used will be part of the \"recipe\". The winners are required to deliver the \"recipe\" to the host to reproduce the results & validate later. If the usage of this dataset would not grant you the license of Apache 2.0/ \"commercially available software\", then this dataset should not be used.
\nIn other words, using TUH data would be allowed only if the team provides the external data that was used as \"part of the package\" when you submit your code. Since the host will plan to verify the winning solution by running the code, if the winning solution cannot be run as provided (i.e. if data needed to run it is not included), this would disqualify it as a winning entry.
\nIf the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.
\nHope it helps clarify. Thanks!
", "votes": null, "replies": [ { "id": 2722971, "author_name": "tatamikenn", "author_url": "", "post_date": "03/29/2024 22:30:57", "content": "@ashleychow
\nThank you for the clarification.
I have a question for the last part.
\n\n\nIf the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.
\n
If we use the example of \"recipe\", we can consider the external dataset as \"ingredients\". And at this point, if we can download the TUH dataset only for sending email and confirm the rule as @bigironsphere mentioned, can we think the dataset as \"publicly available and equally accessible\"? The rule says only restrict 1. acknowledgement, 2. redistribution, and 3. deletion. Thus, we can consider the \"recipe\" using this \"ingredients\" are reproducible for host unless they follow these restrictions. Or, do the host plan to redistribute this dataset?
", "votes": null, "replies": [ { "id": 2727870, "author_name": "ashleychow", "author_url": "", "post_date": "04/02/2024 01:20:05", "content": "Thanks @tatamikenn! I am checking with the host and will keep you posted. Thanks!
", "votes": null, "replies": [ { "id": 2728067, "author_name": "tatamikenn", "author_url": "", "post_date": "04/02/2024 04:49:43", "content": "@ashleychow Thank you for your reply. However, considering that there is only one week left until the end of the competition, we are concerned about the possibility of changing the policy regarding the use of external data at this point. We have no particular objections to the decision that it cannot be used in this competition, and it is likely that other teams are also developing their solutions with the understanding that it cannot be used. Changing the policy now could lead to confusion, and we believe it is better not to alter the decision.
", "votes": null, "replies": [ { "id": 2734029, "author_name": "damienaltzheim", "author_url": "", "post_date": "04/03/2024 22:04:16", "content": "Hi Bilzard, as far as I can tell, TUH is not technically a public dataset. Therefore, using it for this contest is not allowed. - Brandon
", "votes": null, "replies": [ { "id": 2734056, "author_name": "tatamikenn", "author_url": "", "post_date": "04/03/2024 22:49:00", "content": "@damienaltzheim Thanks. I understand.
", "votes": null, "replies": [] } ] } ] } ] } ] } ] } ] } ] } ] } ] }, { "id": 2720491, "author_name": "danan3", "author_url": "", "post_date": "03/28/2024 11:48:48", "content": "So can someone confirm if at the end of this competition, I can include the data for this competition and maybe a link here to maybe my Git repository along with my script?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2714678": "sohier @ashleychow\n\nDear Kaggle Administration and Competition Organizers,\n\nWe wish to inquire about the use of an external data source, the *TUH EEG Seizure Corpus*[1], in the context of this competition. Our understanding is that this dataset requires direct confirmation from the distributing entity via email, which we believe may conflict with the competition's rules under Section 7.C (External Data).\n\n> C. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, **you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants**. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).\n\nDespite this, we have observed that the dataset has been previously uploaded to Kaggle Datasets by a third party[2] and is currently being referenced and used in public notebooks as well as discussions[3-6]. This raises a question for us: Even if a dataset is made \"available to everyone\" by a third party, does using it without the distributor's permission violate Kaggle's competition rules? We seek to clarify this point with the organizers, particularly because the competition rules in Section 11 (Winners Obligations) state that \"all programs must include commercially usable source code,\" and we understand that using an external dataset with unclear commercial usage rights might contravene this rule.\n\n> **To the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense**, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;\n\nWe look forward to your clarification on this matter.\n\nBest regards.\n\n## Reference\n\n- [1] [TUH EEG Seizure Corpus](https://isip.piconepress.com/projects/tuh_eeg/html/downloads.shtml)\n- [2] https://www.kaggle.com/datasets/psyryuvok/the-tuh-eeg-seizure-corpus-tusz-v152\n- [3] https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/479244\n- [4] https://www.kaggle.com/code/seanbearden/processing-the-tuh-eeg-seizure-corpus\n- [5] https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/486002\n- [6] https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/47990", "2714725": "I could not find exact license information of TUH dataset, but I can see a statement in the [download page](https://isip.piconepress.com/projects/tuh_eeg/html/downloads.shtml).\n>The TUH EEG Corpus is freely available. The only reason we require registration is that we need to track who downloads the data. We also want to be able to inform you of any updates to the releases. \n\nI guess this dataset is allowed to use freely, and hence we can use it, but I am not sure.", "2714816": "tomooinubushi Oh, I missed that. Thank you for sharing!", "2716200": "I would appreciate some clarification from the organisers about this. I too applied for access and downloaded the corpus but haven't uploaded it due to the following clauses in the registration form:\n\n>The data available from the NEDC web site is offered free and without restriction, except for a few conditions below that you must abide by: \n(1)The data recipient should acknowledge the provider of this data using the publication listed in the documentation included with the specific corpus (typically found in the AAREADME file). \n(2)The data recipient will not release data to a third party. Please have the third party contact us directly by sending an email to help@nedcdata.org. \n(3)The data recipient will delete the data from all computer systems when finished with the data.\n\nThe second point is crucial; releasing the dataset publicly on Kaggle may well be a violation of their ToS. Even if the dataset has been deleted after this competition, third parties may still have downloaded it without going through the official channels.", "2716233": "Thanks for the questions! We unfortunately could not provide any legal guidance nor interpretation of license of specific dataset. As per the rules, as long as the External Data is \"is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants\" and make sure \"the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense\", they can be used. Hope it helps clarify. Thanks!", "2716388": "ashleychow \n\n> We unfortunately could not provide any legal guidance nor interpretation of license of specific dataset.\n\nWe are not requesting legal guidance, but we just want to be confirmed competition rule.\nIf availability of data is considered differently by participant by participant, it will ruin the competitiveness in this competition.\nWe try to collect evidence as good as possible, so, could you confirm the availability before the competition ends?\nOr, just providing general guidance is welcome.\nFor example, as @bigironsphere mentioned, “a dataset that distributor cannot accept distribution to third party is not available in Kaggle, even if someone uploaded the dataset on Kaggle” etc.", "2716727": "Think this is a known issue with what shows up in Datasets on Kaggle. Have seen examples of what would be copyright infringement in some cases and as you have shown is not in accordance with the original source's requirements for use or republishing. Seems like Datasets needs a better way of dealing with what is uploaded wrt license, terms of use. There are some checks in place for objectionable content (more for images but perhaps text too). \n\nMaybe the easiest option is for one person making use of the TUH EEG Seizure Corpus to email the address in (2) and just get permission to host it on Kaggle for the duration of this competition?", "2718136": "Thanks @tatamikenn! Since competition rules are the agreement between the competition hosts and participants, Kaggle cannot make legal interpretations nor provide guidance on the rules. Let me reach out to the competition hosts for them to chime in. Thanks!", "2718149": "ashleychow Thanks. Could we think the availability of the dataset will be announced in this thread after settlement between competition host and Kaggle team?", "2720491": "So can someone confirm if at the end of this competition, I can include the data for this competition and maybe a link here to maybe my Git repository along with my script?", "2722666": "Thanks for your patience! After discussing with the host, as per the rules:\n\n>C. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your Submissions. However, you will ensure the External Data is publicly available and equally accessible to use by all participants of the Competition for purposes of the competition at no cost to the other participants. The ability to use External Data under this Section 7.C (External Data) does not limit your other obligations under these Competition Rules, including but not limited to Section 11 (Winners Obligations).\n...\n>As a condition to being awarded a Prize, a Prize winner must fulfill the following obligations:\n\n>(a) deliver to the Competition Sponsor the final model's software code as used to generate the winning Submission and associated documentation. The delivered software code should follow these documentation guidelines, must be capable of generating the winning Submission, and contain a description of resources required to build and/or run the executable code successfully. To the extent that the final model’s software code includes generally commercially available software that is not owned by you, but that can be procured by the Competition Sponsor without undue expense, then instead of delivering the code for that software to the Competition Sponsor, you must identify that software, method for procuring it, and any parameters or other information necessary to replicate the winning Submission;\n\nSo if our winning solution is a \"recipe to the cake\", the Datasets being used will be part of the \"recipe\". The winners are required to deliver the \"recipe\" to the host to reproduce the results & validate later. If the usage of this dataset would not grant you the license of Apache 2.0/ \"commercially available software\", then this dataset should not be used. \n\nIn other words, using TUH data would be allowed only if the team provides the external data that was used as \"part of the package\" when you submit your code. Since the host will plan to verify the winning solution by running the code, if the winning solution cannot be run as provided (i.e. if data needed to run it is not included), this would disqualify it as a winning entry.\n\nIf the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.\n\nHope it helps clarify. Thanks!", "2722971": "ashleychow \nThank you for the clarification.\n\nI have a question for the last part.\n\n> If the usage of TUH data has restrictions as mentioned by @bigironsphere above (i.e. \"the data recipient will not release data to a third party\"), the participants are not permitted to include the TUH data in their code / package because that data is not public. In that case, use of TUH data would violate the contest rules.\n\nIf we use the example of \"recipe\", we can consider the external dataset as \"ingredients\". And at this point, if we can download the TUH dataset only for sending email and confirm the rule as @bigironsphere mentioned, can we think the dataset as \"publicly available and equally accessible\"? The rule says only restrict 1. acknowledgement, 2. redistribution, and 3. deletion. Thus, we can consider the \"recipe\" using this \"ingredients\" are reproducible for host unless they follow these restrictions. Or, do the host plan to redistribute this dataset?", "2727870": "Thanks @tatamikenn! I am checking with the host and will keep you posted. Thanks!", "2728067": "ashleychow Thank you for your reply. However, considering that there is only one week left until the end of the competition, we are concerned about the possibility of changing the policy regarding the use of external data at this point. We have no particular objections to the decision that it cannot be used in this competition, and it is likely that other teams are also developing their solutions with the understanding that it cannot be used. Changing the policy now could lead to confusion, and we believe it is better not to alter the decision.", "2734029": "Hi Bilzard, as far as I can tell, TUH is not technically a public dataset. Therefore, using it for this contest is not allowed. - Brandon", "2734056": "damienaltzheim Thanks. I understand." }, "source": "meta" }