Together with my teammate, @federicomatorra, we are a couple of lucky newcomers!. But I have to say itβs was not luck at all, we invested many hours thinking of and implementing a strategy and finally results were favorable.
\nWe were quite sure that the private test set was a completely different issue, and so it was important to think of a good statistical approach, but with the additional restriction of scarce resources and ignorance of the latest developments in NLP.
\nSo, in summary, although we evaluated many ideas, we used spaCy 2 (v2.spacy.io) for text classification in 2 steps, first for separating sentences (ie., sentences that mention a dataset vs no dataset at all) and second for identifying datasets versus ORGs, LOCs, etc., and many other acronyms and names that were only (at least in our view) false positives.
\nTo train the text classification models, we used the datasets names provided for training and managed to create a list of additional dataset names together with corresponding acronyms (positive and negatives) using an abbreviation detector (https://allenai.github.io/scispacy/) and matching for words such as dataset, databank, survey, etc.
\nWe think if was simple enough to be trained and tested fast and flexible enough for adapting to unknown publications.
\nThe notebook that was scored in 6th place it was a hybrid of text matching and the approach mentioned above but as I said, in our opinion, the public and private sets of documents were two completely different universes, so it was not possible to finish in the first ten using only string matching.
\nWe want to thank the organizers and congratulate the immense number of participants and in particular, those who finished in the medallist positions. Also, recognize the work done by Coleridge to set up the training and testing sets.
", "messages": [ { "id": "1362956", "postDate": "06/23/2021 19:10:52", "content": "Together with my teammate, @federicomatorra, we are a couple of lucky newcomers!. But I have to say itβs was not luck at all, we invested many hours thinking of and implementing a strategy and finally results were favorable.
\nWe were quite sure that the private test set was a completely different issue, and so it was important to think of a good statistical approach, but with the additional restriction of scarce resources and ignorance of the latest developments in NLP.
\nSo, in summary, although we evaluated many ideas, we used spaCy 2 (v2.spacy.io) for text classification in 2 steps, first for separating sentences (ie., sentences that mention a dataset vs no dataset at all) and second for identifying datasets versus ORGs, LOCs, etc., and many other acronyms and names that were only (at least in our view) false positives.
\nTo train the text classification models, we used the datasets names provided for training and managed to create a list of additional dataset names together with corresponding acronyms (positive and negatives) using an abbreviation detector (https://allenai.github.io/scispacy/) and matching for words such as dataset, databank, survey, etc.
\nWe think if was simple enough to be trained and tested fast and flexible enough for adapting to unknown publications.
\nThe notebook that was scored in 6th place it was a hybrid of text matching and the approach mentioned above but as I said, in our opinion, the public and private sets of documents were two completely different universes, so it was not possible to finish in the first ten using only string matching.
\nWe want to thank the organizers and congratulate the immense number of participants and in particular, those who finished in the medallist positions. Also, recognize the work done by Coleridge to set up the training and testing sets.
", "rawMarkdown": "Together with my teammate, @federicomatorra, we are a couple of lucky newcomers!. But I have to say itβs was not luck at all, we invested many hours thinking of and implementing a strategy and finally results were favorable.\n\nWe were quite sure that the private test set was a completely different issue, and so it was important to think of a good statistical approach, but with the additional restriction of scarce resources and ignorance of the latest developments in NLP.\n\nSo, in summary, although we evaluated many ideas, we used spaCy 2 (v2.spacy.io) for text classification in 2 steps, first for separating sentences (ie., sentences that mention a dataset vs no dataset at all) and second for identifying datasets versus ORGs, LOCs, etc., and many other acronyms and names that were only (at least in our view) false positives.\n\nTo train the text classification models, we used the datasets names provided for training and managed to create a list of additional dataset names together with corresponding acronyms (positive and negatives) using an abbreviation detector (https://allenai.github.io/scispacy/) and matching for words such as dataset, databank, survey, etc.\n\nWe think if was simple enough to be trained and tested fast and flexible enough for adapting to unknown publications.\n\nThe notebook that was scored in 6th place it was a hybrid of text matching and the approach mentioned above but as I said, in our opinion, the public and private sets of documents were two completely different universes, so it was not possible to finish in the first ten using only string matching.\n\nWe want to thank the organizers and congratulate the immense number of participants and in particular, those who finished in the medallist positions. Also, recognize the work done by Coleridge to set up the training and testing sets.", "votes": null }, { "id": "1363177", "postDate": "06/24/2021 01:58:08", "content": "Congratulations! Keep it up!!!! π
", "rawMarkdown": "Congratulations! Keep it up!!!! π", "votes": null }, { "id": "1363527", "postDate": "06/24/2021 07:14:09", "content": "Congratulations and thanks for sharing. π
", "rawMarkdown": "Congratulations and thanks for sharing. π", "votes": null }, { "id": "1364383", "postDate": "06/24/2021 23:07:10", "content": "Thank you for sharing your good experience, Congratulations!π
", "rawMarkdown": "Thank you for sharing your good experience, Congratulations!π", "votes": null }, { "id": "1364864", "postDate": "06/25/2021 09:09:46", "content": "Awesome work and thanks for sharing.Congratulations! π
", "rawMarkdown": "Awesome work and thanks for sharing.Congratulations! π", "votes": null } ], "comments": [ { "id": 1363177, "author_name": "ramesh084", "author_url": "", "post_date": "06/24/2021 01:58:08", "content": "Congratulations! Keep it up!!!! π
", "votes": null, "replies": [] }, { "id": 1363527, "author_name": "ashrafr", "author_url": "", "post_date": "06/24/2021 07:14:09", "content": "Congratulations and thanks for sharing. π
", "votes": null, "replies": [] }, { "id": 1364383, "author_name": "ssinyu", "author_url": "", "post_date": "06/24/2021 23:07:10", "content": "Thank you for sharing your good experience, Congratulations!π
", "votes": null, "replies": [] }, { "id": 1364864, "author_name": "saurabhbagchi", "author_url": "", "post_date": "06/25/2021 09:09:46", "content": "Awesome work and thanks for sharing.Congratulations! π
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1362956": "Together with my teammate, @federicomatorra, we are a couple of lucky newcomers!. But I have to say itβs was not luck at all, we invested many hours thinking of and implementing a strategy and finally results were favorable.\n\nWe were quite sure that the private test set was a completely different issue, and so it was important to think of a good statistical approach, but with the additional restriction of scarce resources and ignorance of the latest developments in NLP.\n\nSo, in summary, although we evaluated many ideas, we used spaCy 2 (v2.spacy.io) for text classification in 2 steps, first for separating sentences (ie., sentences that mention a dataset vs no dataset at all) and second for identifying datasets versus ORGs, LOCs, etc., and many other acronyms and names that were only (at least in our view) false positives.\n\nTo train the text classification models, we used the datasets names provided for training and managed to create a list of additional dataset names together with corresponding acronyms (positive and negatives) using an abbreviation detector (https://allenai.github.io/scispacy/) and matching for words such as dataset, databank, survey, etc.\n\nWe think if was simple enough to be trained and tested fast and flexible enough for adapting to unknown publications.\n\nThe notebook that was scored in 6th place it was a hybrid of text matching and the approach mentioned above but as I said, in our opinion, the public and private sets of documents were two completely different universes, so it was not possible to finish in the first ten using only string matching.\n\nWe want to thank the organizers and congratulate the immense number of participants and in particular, those who finished in the medallist positions. Also, recognize the work done by Coleridge to set up the training and testing sets.", "1363177": "Congratulations! Keep it up!!!! π", "1363527": "Congratulations and thanks for sharing. π", "1364383": "Thank you for sharing your good experience, Congratulations!π", "1364864": "Awesome work and thanks for sharing.Congratulations! π" }, "source": "meta" }