![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F214989%2Feba6581ebd028faff88f08a2b719758e%2Fsample.png?generation=1679297783311464&alt=media\")
In the training data not all users are playing the same game events because there are four different scripts that are randomly chosen when the game starts.
\nAs pointet out by @mattoglesby here, description of game types can be found on jo_wilder in the Script Versions section :
\n\n\nScript Versions
\nStarting in v7, multiple scripts were added to the game for AB tests on snark and humor. The game will randomly choose between 4 different data files: data_dry.js, data_nohumor.js, data_nosnark.js, and the original data.js. They are each built from their respective data folder in assets. The type of script used is only logged once, in startgame.
\nIndex Name Description
\n
\n 0 dry no humor or snark
\n 1 nohumor no humor (includes snark)
\n 2 nosnark no snark (includes humor). No snark can also be thought of as \"obedient\"
\n 3 normal base script (includes snark and humor)
If you want to play a specific game type you must use these custom links:
\nif you want to play a randomly chosen game use this link https://jowilder-master.netlify.app/
\nFor some events there are differenes between texts in each of these versions.
\n
In this dataset I collected text differences between games
", "messages": [ { "id": "2189044", "postDate": "03/20/2023 07:37:29", "content": "In the training data not all users are playing the same game events because there are four different scripts that are randomly chosen when the game starts.
\nAs pointet out by @mattoglesby here, description of game types can be found on jo_wilder in the Script Versions section :
\n\n\nScript Versions
\nStarting in v7, multiple scripts were added to the game for AB tests on snark and humor. The game will randomly choose between 4 different data files: data_dry.js, data_nohumor.js, data_nosnark.js, and the original data.js. They are each built from their respective data folder in assets. The type of script used is only logged once, in startgame.
\nIndex Name Description
\n
\n 0 dry no humor or snark
\n 1 nohumor no humor (includes snark)
\n 2 nosnark no snark (includes humor). No snark can also be thought of as \"obedient\"
\n 3 normal base script (includes snark and humor)
If you want to play a specific game type you must use these custom links:
\nif you want to play a randomly chosen game use this link https://jowilder-master.netlify.app/
\nFor some events there are differenes between texts in each of these versions.
\n
In this dataset I collected text differences between games
", "rawMarkdown": "In the training data not all users are playing the same game events because there are four different scripts that are randomly chosen when the game starts.\n\nAs pointet out by @mattoglesby [here](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/387864#2147760), description of game types can be found on [jo_wilder](https://github.com/fielddaylab/jo_wilder) in the Script Versions section :\n\n> Script Versions\n> \n> Starting in v7, multiple scripts were added to the game for AB tests on snark and humor. The game will randomly choose between 4 different data files: data_dry.js, data_nohumor.js, data_nosnark.js, and the original data.js. They are each built from their respective data folder in assets. The type of script used is only logged once, in startgame.\n> \n> Index Name Description\n> 0 dry no humor or snark\n> 1 nohumor no humor (includes snark)\n> 2 nosnark no snark (includes humor). No snark can also be thought of as \"obedient\"\n> 3 normal base script (includes snark and humor)\n\nIf you want to play a specific game type you must use these custom links:\n* normal: https://jowilder-master.netlify.app/?script_type=original\n* dry: https://jowilder-master.netlify.app/?script_type=dry\n* nohumor: https://jowilder-master.netlify.app/?script_type=nohumor\n* nosnark: https://jowilder-master.netlify.app/?script_type=nosnark\n\nif you want to play a randomly chosen game use this link https://jowilder-master.netlify.app/ \n\nFor some events there are differenes between texts in each of these versions.\n\n\n\nIn this [dataset](https://www.kaggle.com/datasets/steubk/meetings-are-boring) I collected text differences between games", "votes": null }, { "id": "2189796", "postDate": "03/20/2023 19:02:11", "content": "Because I'm focusing on text display time and frequency to create features,
\nYour technique has helped me improve CV of my predictions.
\nThank you!
Thanks a lot. I just got help from your topic for CV in my work
", "rawMarkdown": "Thanks a lot. I just got help from your topic for CV in my work", "votes": null }, { "id": "2190120", "postDate": "03/21/2023 03:52:30", "content": "Thanks for sharing.
\nWe could use such text data to classify versions clearly.
A possible precise approch is like this:
\ncount_dry = \ncount_normal = \ncount_nohumor = \ncount_nosnark = \n _index ((train.index.tolist())):\n i train.loc[_index][].values.tolist():\n i dry:\n count_dry +=\n i normal:\n count_normal +=\n i nohumor:\n count_nohumor +=\n i nosnark:\n count_nosnark +=\nThen check which one is bigger.
\nAnother approch is to select certain four texts to do intersection pairwise to classify four types. It also performs good and is much faster.
", "rawMarkdown": "Thanks for sharing.\nWe could use such text data to classify versions clearly.\n\nA possible precise approch is like this:\n```Python\ncount_dry = 0\ncount_normal = 0\ncount_nohumor = 0\ncount_nosnark = 0\nfor _index in list(set(train.index.tolist())):\n for i in train.loc[_index]['text'].values.tolist():\n if i in dry:\n count_dry +=1\n if i in normal:\n count_normal +=1\n if i in nohumor:\n count_nohumor +=1\n if i in nosnark:\n count_nosnark +=1\n```\nThen check which one is bigger.\n\nAnother approch is to select certain four texts to do intersection pairwise to classify four types. It also performs good and is much faster.", "votes": null }, { "id": "2190143", "postDate": "03/21/2023 04:24:34", "content": "Thanks for sharing valuable insights for the community! Interesting and good technical work.👍
", "rawMarkdown": "Thanks for sharing valuable insights for the community! Interesting and good technical work.👍", "votes": null }, { "id": "2191880", "postDate": "03/22/2023 09:07:36", "content": "Thanks a lot👍
", "rawMarkdown": "Thanks a lot👍", "votes": null } ], "comments": [ { "id": 2189796, "author_name": "tanakaakinori", "author_url": "", "post_date": "03/20/2023 19:02:11", "content": "Because I'm focusing on text display time and frequency to create features,
\nYour technique has helped me improve CV of my predictions.
\nThank you!
Thanks a lot. I just got help from your topic for CV in my work
", "votes": null, "replies": [] }, { "id": 2190120, "author_name": "yunqicao", "author_url": "", "post_date": "03/21/2023 03:52:30", "content": "Thanks for sharing.
\nWe could use such text data to classify versions clearly.
A possible precise approch is like this:
\ncount_dry = \ncount_normal = \ncount_nohumor = \ncount_nosnark = \n _index ((train.index.tolist())):\n i train.loc[_index][].values.tolist():\n i dry:\n count_dry +=\n i normal:\n count_normal +=\n i nohumor:\n count_nohumor +=\n i nosnark:\n count_nosnark +=\nThen check which one is bigger.
\nAnother approch is to select certain four texts to do intersection pairwise to classify four types. It also performs good and is much faster.
", "votes": null, "replies": [] }, { "id": 2190143, "author_name": "tariqbashir", "author_url": "", "post_date": "03/21/2023 04:24:34", "content": "Thanks for sharing valuable insights for the community! Interesting and good technical work.👍
", "votes": null, "replies": [] }, { "id": 2191880, "author_name": "chenxiangjun", "author_url": "", "post_date": "03/22/2023 09:07:36", "content": "Thanks a lot👍
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2189044": "In the training data not all users are playing the same game events because there are four different scripts that are randomly chosen when the game starts.\n\nAs pointet out by @mattoglesby [here](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/387864#2147760), description of game types can be found on [jo_wilder](https://github.com/fielddaylab/jo_wilder) in the Script Versions section :\n\n> Script Versions\n> \n> Starting in v7, multiple scripts were added to the game for AB tests on snark and humor. The game will randomly choose between 4 different data files: data_dry.js, data_nohumor.js, data_nosnark.js, and the original data.js. They are each built from their respective data folder in assets. The type of script used is only logged once, in startgame.\n> \n> Index Name Description\n> 0 dry no humor or snark\n> 1 nohumor no humor (includes snark)\n> 2 nosnark no snark (includes humor). No snark can also be thought of as \"obedient\"\n> 3 normal base script (includes snark and humor)\n\nIf you want to play a specific game type you must use these custom links:\n* normal: https://jowilder-master.netlify.app/?script_type=original\n* dry: https://jowilder-master.netlify.app/?script_type=dry\n* nohumor: https://jowilder-master.netlify.app/?script_type=nohumor\n* nosnark: https://jowilder-master.netlify.app/?script_type=nosnark\n\nif you want to play a randomly chosen game use this link https://jowilder-master.netlify.app/ \n\nFor some events there are differenes between texts in each of these versions.\n\n\n\nIn this [dataset](https://www.kaggle.com/datasets/steubk/meetings-are-boring) I collected text differences between games", "2189796": "Because I'm focusing on text display time and frequency to create features,\nYour technique has helped me improve CV of my predictions.\nThank you!", "2189838": "Thanks a lot. I just got help from your topic for CV in my work", "2190120": "Thanks for sharing.\nWe could use such text data to classify versions clearly.\n\nA possible precise approch is like this:\n```Python\ncount_dry = 0\ncount_normal = 0\ncount_nohumor = 0\ncount_nosnark = 0\nfor _index in list(set(train.index.tolist())):\n for i in train.loc[_index]['text'].values.tolist():\n if i in dry:\n count_dry +=1\n if i in normal:\n count_normal +=1\n if i in nohumor:\n count_nohumor +=1\n if i in nosnark:\n count_nosnark +=1\n```\nThen check which one is bigger.\n\nAnother approch is to select certain four texts to do intersection pairwise to classify four types. It also performs good and is much faster.", "2190143": "Thanks for sharing valuable insights for the community! Interesting and good technical work.👍", "2191880": "Thanks a lot👍" }, "source": "meta" }