There was a recent competition by Coleridge, which had an objective to identify citations and mentions in scientific publications. It contains about 5 thousand articles which may be useful in additional model training for the present competition.
\nLink to competition overview: https://coleridgeinitiative.org/richcontext/richcontextcompetition/
\nGH link with data: https://github.com/Coleridge-Initiative/rich-context-competition
\nrelated link: https://study.sagepub.com/richcontext/student-resources/the-%E2%80%98rich-context%E2%80%99-competition
If you find data from \"Rich Context\" useful -- please share you findings here.
", "messages": [ { "id": "1266997", "postDate": "04/08/2021 08:26:20", "content": "There was a recent competition by Coleridge, which had an objective to identify citations and mentions in scientific publications. It contains about 5 thousand articles which may be useful in additional model training for the present competition.
\nLink to competition overview: https://coleridgeinitiative.org/richcontext/richcontextcompetition/
\nGH link with data: https://github.com/Coleridge-Initiative/rich-context-competition
\nrelated link: https://study.sagepub.com/richcontext/student-resources/the-%E2%80%98rich-context%E2%80%99-competition
If you find data from \"Rich Context\" useful -- please share you findings here.
", "rawMarkdown": "There was a recent competition by Coleridge, which had an objective to identify citations and mentions in scientific publications. It contains about 5 thousand articles which may be useful in additional model training for the present competition.\n\nLink to competition overview: [https://coleridgeinitiative.org/richcontext/richcontextcompetition/](url)\nGH link with data: [https://github.com/Coleridge-Initiative/rich-context-competition](url)\nrelated link: [https://study.sagepub.com/richcontext/student-resources/the-%E2%80%98rich-context%E2%80%99-competition](url)\n\nIf you find data from \"Rich Context\" useful -- please share you findings here.", "votes": null }, { "id": "1267317", "postDate": "04/08/2021 13:13:53", "content": "Hi Thank you so much for reminding everyone to look at this. We pointed to the book https://study.sagepub.com/richcontext in the current competition description. There is no overlap in datasets, but if there were, it wouldn't help. This is not a competition designed to look up dataset references. It is to find references based on the semantic context used by scientists to describe the datasets used in their empirical analyses. The book describes that broader vision.
\nI can't tell you how excited we - and many government agency staff - are at the response to the competition. If you want to see some of the work that is related, you should come to the next meeting of the Advisory Committee on Data for Evidence Building on April 23 https://www.bea.gov/evidence
\nLook forward to your results - based on the private leaderboard, many of you are making great progress!
", "rawMarkdown": "Hi Thank you so much for reminding everyone to look at this. We pointed to the book https://study.sagepub.com/richcontext in the current competition description. There is no overlap in datasets, but if there were, it wouldn't help. This is not a competition designed to look up dataset references. It is to find references based on the semantic context used by scientists to describe the datasets used in their empirical analyses. The book describes that broader vision.\n\nI can't tell you how excited we - and many government agency staff - are at the response to the competition. If you want to see some of the work that is related, you should come to the next meeting of the Advisory Committee on Data for Evidence Building on April 23 https://www.bea.gov/evidence \n\nLook forward to your results - based on the private leaderboard, many of you are making great progress!", "votes": null }, { "id": "1267366", "postDate": "04/08/2021 13:35:57", "content": "Hi, @julialane, I edited this thread title to be more neutral. I agree that simple title matching is a rather limited technique. My suggestion (and plan) was to use \"Rich Context\" data to enhance model training.
", "rawMarkdown": "Hi, @julialane, I edited this thread title to be more neutral. I agree that simple title matching is a rather limited technique. My suggestion (and plan) was to use \"Rich Context\" data to enhance model training.", "votes": null }, { "id": "1267426", "postDate": "04/08/2021 14:05:18", "content": "@poedator You're the experts - so just will be happy to see the results. Thanks a million.
", "rawMarkdown": "poedator You're the experts - so just will be happy to see the results. Thanks a million.", "votes": null }, { "id": "1293743", "postDate": "05/05/2021 06:09:23", "content": "Is the dataset labels in the previous competition exhaustively labelled?
\nI saw 5499 labels for 2500 publications, it seems too few.
If exhaustively labelled, training models then validating on the previous competition data seems good strategy to evaluate our models.
", "rawMarkdown": "Is the dataset labels in the previous competition exhaustively labelled?\nI saw 5499 labels for 2500 publications, it seems too few.\n\nIf exhaustively labelled, training models then validating on the previous competition data seems good strategy to evaluate our models.", "votes": null } ], "comments": [ { "id": 1267317, "author_name": "julialane", "author_url": "", "post_date": "04/08/2021 13:13:53", "content": "Hi Thank you so much for reminding everyone to look at this. We pointed to the book https://study.sagepub.com/richcontext in the current competition description. There is no overlap in datasets, but if there were, it wouldn't help. This is not a competition designed to look up dataset references. It is to find references based on the semantic context used by scientists to describe the datasets used in their empirical analyses. The book describes that broader vision.
\nI can't tell you how excited we - and many government agency staff - are at the response to the competition. If you want to see some of the work that is related, you should come to the next meeting of the Advisory Committee on Data for Evidence Building on April 23 https://www.bea.gov/evidence
\nLook forward to your results - based on the private leaderboard, many of you are making great progress!
", "votes": null, "replies": [ { "id": 1267366, "author_name": "poedator", "author_url": "", "post_date": "04/08/2021 13:35:57", "content": "Hi, @julialane, I edited this thread title to be more neutral. I agree that simple title matching is a rather limited technique. My suggestion (and plan) was to use \"Rich Context\" data to enhance model training.
", "votes": null, "replies": [] }, { "id": 1293743, "author_name": "resistance0108", "author_url": "", "post_date": "05/05/2021 06:09:23", "content": "Is the dataset labels in the previous competition exhaustively labelled?
\nI saw 5499 labels for 2500 publications, it seems too few.
If exhaustively labelled, training models then validating on the previous competition data seems good strategy to evaluate our models.
", "votes": null, "replies": [] } ] }, { "id": 1267426, "author_name": "julialane", "author_url": "", "post_date": "04/08/2021 14:05:18", "content": "@poedator You're the experts - so just will be happy to see the results. Thanks a million.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1266997": "There was a recent competition by Coleridge, which had an objective to identify citations and mentions in scientific publications. It contains about 5 thousand articles which may be useful in additional model training for the present competition.\n\nLink to competition overview: [https://coleridgeinitiative.org/richcontext/richcontextcompetition/](url)\nGH link with data: [https://github.com/Coleridge-Initiative/rich-context-competition](url)\nrelated link: [https://study.sagepub.com/richcontext/student-resources/the-%E2%80%98rich-context%E2%80%99-competition](url)\n\nIf you find data from \"Rich Context\" useful -- please share you findings here.", "1267317": "Hi Thank you so much for reminding everyone to look at this. We pointed to the book https://study.sagepub.com/richcontext in the current competition description. There is no overlap in datasets, but if there were, it wouldn't help. This is not a competition designed to look up dataset references. It is to find references based on the semantic context used by scientists to describe the datasets used in their empirical analyses. The book describes that broader vision.\n\nI can't tell you how excited we - and many government agency staff - are at the response to the competition. If you want to see some of the work that is related, you should come to the next meeting of the Advisory Committee on Data for Evidence Building on April 23 https://www.bea.gov/evidence \n\nLook forward to your results - based on the private leaderboard, many of you are making great progress!", "1267366": "Hi, @julialane, I edited this thread title to be more neutral. I agree that simple title matching is a rather limited technique. My suggestion (and plan) was to use \"Rich Context\" data to enhance model training.", "1267426": "poedator You're the experts - so just will be happy to see the results. Thanks a million.", "1293743": "Is the dataset labels in the previous competition exhaustively labelled?\nI saw 5499 labels for 2500 publications, it seems too few.\n\nIf exhaustively labelled, training models then validating on the previous competition data seems good strategy to evaluate our models." }, "source": "meta" }