{ "id": 193933, "title": "Understanding some concepts", "url": "/competitions/riiid-test-answer-prediction/discussion/193933", "author_name": "", "post_date": "2020-10-29T17:54:53.590070Z", "votes": -5, "comment_count": 7, "views": 0, "content": "

I joined he competition recently. Please excuse me for very basic questions asked below

\n

1) How should we interpret the Answers -1 for content id that is for lectures and 1 .
\n2) Any relationship between the tag of Questions and Lectures
\n3) Difference between Bundle Id of question and task container id
\n4) difference between two type of lectures which Concept/solving questions
\n5) What kind of evaluations student undergoes only Questions asked or any thing to do with lecture also..

\n

6) What is difference between Part in Question df and Part in Lectures ?

", "messages": [ { "id": "1064108", "postDate": "10/29/2020 17:54:53", "content": "

I joined he competition recently. Please excuse me for very basic questions asked below

\n

1) How should we interpret the Answers -1 for content id that is for lectures and 1 .
\n2) Any relationship between the tag of Questions and Lectures
\n3) Difference between Bundle Id of question and task container id
\n4) difference between two type of lectures which Concept/solving questions
\n5) What kind of evaluations student undergoes only Questions asked or any thing to do with lecture also..

\n

6) What is difference between Part in Question df and Part in Lectures ?

", "rawMarkdown": "I joined he competition recently. Please excuse me for very basic questions asked below\n\n1) How should we interpret the Answers -1 for content id that is for lectures and 1 .\n2) Any relationship between the tag of Questions and Lectures\n3) Difference between Bundle Id of question and task container id \n4) difference between two type of lectures which Concept/solving questions\n5) What kind of evaluations student undergoes only Questions asked or any thing to do with lecture also..\n\n6) What is difference between Part in Question df and Part in Lectures ?", "votes": null }, { "id": "1064122", "postDate": "10/29/2020 18:30:09", "content": "

Most questions are already answered in Data or in Forum.

\n

1) Lectures don't have answers. Consider -1 as NA as mentioned here
\n2) Check the first point here
\n3) Answered here
\n4) They are just (four) different types of lectures
\n5) There is no evaluation for lectures as mentioned here
\n6) No difference, both refer to same parts

", "rawMarkdown": "Most questions are already answered in [Data](https://www.kaggle.com/c/riiid-test-answer-prediction/data) or in [Forum](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion).\n\n1) Lectures don't have answers. Consider -1 as NA as mentioned [here](https://www.kaggle.com/c/riiid-test-answer-prediction/data)\n2) Check the first point [here](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/191106)\n3) Answered [here](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/190828#1051526)\n4) They are just (four) different types of lectures\n5) There is no evaluation for lectures as mentioned [here](https://www.kaggle.com/c/riiid-test-answer-prediction/data)\n6) No difference, both refer to same [parts](https://www.iibc-global.org/english/toeic/test/lr/about/format.html)", "votes": null }, { "id": "1064665", "postDate": "10/30/2020 11:43:58", "content": "

@rohanrao i understand quite a it things now..

\n

1) Timestamp mentions user interaction and first event , are the user interaction and event same stuff ?

\n

2) Why there is word First mentioned here…
\n3) Does the clock always resets for any new event tracked for user , stops when the event ends ?

", "rawMarkdown": "rohanrao i understand quite a it things now..\n\n1) Timestamp mentions user interaction and first event , are the user interaction and event same stuff ?\n\n2) Why there is word First mentioned here...\n3) Does the clock always resets for any new event tracked for user , stops when the event ends ?", "votes": null }, { "id": "1065253", "postDate": "10/31/2020 04:50:58", "content": "

1) Interaction = Event
\n2) Because timestamp is relative to the first completed event of the user. That's why every user starts at timestamp = 0
\n3) No, the clock for a user starts only once when the first event is completed.

\n

Here's an example for 2) and 3)

\n

If a user completed events at following actual times:
\nFirst - 10:53am on any day
\nSecond - 10:57am on same day
\nThird - 11:22am on same day

\n

The timestamp values in data would be
\nFirst - timestamp = 0
\nSecond - timestamp = 240000 (4 minutes)
\nThird - timestamp = = 1740000 (29 minutes)

", "rawMarkdown": "1) Interaction = Event\n2) Because timestamp is relative to the **first** completed event of the user. That's why every user starts at timestamp = 0\n3) No, the clock for a user starts only once when the first event is completed.\n\nHere's an example for 2) and 3)\n\nIf a user completed events at following actual times:\nFirst - 10:53am on any day\nSecond - 10:57am on same day\nThird - 11:22am on same day\n\nThe timestamp values in data would be\nFirst - timestamp = 0\nSecond - timestamp = 240000 (4 minutes)\nThird - timestamp = = 1740000 (29 minutes)", "votes": null }, { "id": "1065303", "postDate": "10/31/2020 06:18:26", "content": "

1) but what if second event comes in days after first event.. can this be possibility too.. in that case what is meaningful info it can carry.. i think in that case one would want to check relative time elapsed wrt to previous event ?

\n

2) From training point of view.. can one remove all the lecture events since they dont have a target ?

", "rawMarkdown": "1) but what if second event comes in days after first event.. can this be possibility too.. in that case what is meaningful info it can carry.. i think in that case one would want to check relative time elapsed wrt to previous event ?\n\n2) From training point of view.. can one remove all the lecture events since they dont have a target ?", "votes": null }, { "id": "1065831", "postDate": "10/31/2020 20:54:29", "content": "

1) Indeed, ~ 87% of timestamps are > 24H so you have to deal with it.
\n2) Yes you can remove lectures events doing :

\n

train[train[\"content_type_id\"] == 0]

", "rawMarkdown": "1) Indeed, ~ 87% of timestamps are > 24H so you have to deal with it.\n2) Yes you can remove lectures events doing :\n\n`train[train[\"content_type_id\"] == 0]`", "votes": null }, { "id": "1067181", "postDate": "11/02/2020 11:42:24", "content": "

@matthieuplante thanku

\n

1) can we assume that Private+Public test set will contain uniquely Question id + Lecture id that would all exist in questions.csv/lecture.csv ?

\n

2) Can we assume Tag id represents possibly the subject domain ?

\n

3) Why do we have multiple tags in Question.csv for each question ?

", "rawMarkdown": "matthieuplante thanku\n\n1) can we assume that Private+Public test set will contain uniquely Question id + Lecture id that would all exist in questions.csv/lecture.csv ?\n\n2) Can we assume Tag id represents possibly the subject domain ?\n\n3) Why do we have multiple tags in Question.csv for each question ?", "votes": null }, { "id": "1067463", "postDate": "11/02/2020 14:48:28", "content": "

1) Yeah I think I saw that in a discussion

\n

2) Yes

\n

3) Because a question can concern multiple domains. We don't have the signification of the tags but for example a question can concern the financial domain and the newspaper domain so it will have 2 tags (Not the best example but I think you will understand ;) )

", "rawMarkdown": "1) Yeah I think I saw that in a discussion \n\n2) Yes\n\n3) Because a question can concern multiple domains. We don't have the signification of the tags but for example a question can concern the financial domain and the newspaper domain so it will have 2 tags (Not the best example but I think you will understand ;) )", "votes": null } ], "comments": [ { "id": 1064122, "author_name": "rohanrao", "author_url": "", "post_date": "10/29/2020 18:30:09", "content": "

Most questions are already answered in Data or in Forum.

\n

1) Lectures don't have answers. Consider -1 as NA as mentioned here
\n2) Check the first point here
\n3) Answered here
\n4) They are just (four) different types of lectures
\n5) There is no evaluation for lectures as mentioned here
\n6) No difference, both refer to same parts

", "votes": null, "replies": [ { "id": 1064665, "author_name": "jaideepvalani", "author_url": "", "post_date": "10/30/2020 11:43:58", "content": "

@rohanrao i understand quite a it things now..

\n

1) Timestamp mentions user interaction and first event , are the user interaction and event same stuff ?

\n

2) Why there is word First mentioned here…
\n3) Does the clock always resets for any new event tracked for user , stops when the event ends ?

", "votes": null, "replies": [] }, { "id": 1065253, "author_name": "rohanrao", "author_url": "", "post_date": "10/31/2020 04:50:58", "content": "

1) Interaction = Event
\n2) Because timestamp is relative to the first completed event of the user. That's why every user starts at timestamp = 0
\n3) No, the clock for a user starts only once when the first event is completed.

\n

Here's an example for 2) and 3)

\n

If a user completed events at following actual times:
\nFirst - 10:53am on any day
\nSecond - 10:57am on same day
\nThird - 11:22am on same day

\n

The timestamp values in data would be
\nFirst - timestamp = 0
\nSecond - timestamp = 240000 (4 minutes)
\nThird - timestamp = = 1740000 (29 minutes)

", "votes": null, "replies": [] }, { "id": 1065303, "author_name": "jaideepvalani", "author_url": "", "post_date": "10/31/2020 06:18:26", "content": "

1) but what if second event comes in days after first event.. can this be possibility too.. in that case what is meaningful info it can carry.. i think in that case one would want to check relative time elapsed wrt to previous event ?

\n

2) From training point of view.. can one remove all the lecture events since they dont have a target ?

", "votes": null, "replies": [] }, { "id": 1065831, "author_name": "matthieuplante", "author_url": "", "post_date": "10/31/2020 20:54:29", "content": "

1) Indeed, ~ 87% of timestamps are > 24H so you have to deal with it.
\n2) Yes you can remove lectures events doing :

\n

train[train[\"content_type_id\"] == 0]

", "votes": null, "replies": [] }, { "id": 1067181, "author_name": "jaideepvalani", "author_url": "", "post_date": "11/02/2020 11:42:24", "content": "

@matthieuplante thanku

\n

1) can we assume that Private+Public test set will contain uniquely Question id + Lecture id that would all exist in questions.csv/lecture.csv ?

\n

2) Can we assume Tag id represents possibly the subject domain ?

\n

3) Why do we have multiple tags in Question.csv for each question ?

", "votes": null, "replies": [] }, { "id": 1067463, "author_name": "matthieuplante", "author_url": "", "post_date": "11/02/2020 14:48:28", "content": "

1) Yeah I think I saw that in a discussion

\n

2) Yes

\n

3) Because a question can concern multiple domains. We don't have the signification of the tags but for example a question can concern the financial domain and the newspaper domain so it will have 2 tags (Not the best example but I think you will understand ;) )

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1064108": "I joined he competition recently. Please excuse me for very basic questions asked below\n\n1) How should we interpret the Answers -1 for content id that is for lectures and 1 .\n2) Any relationship between the tag of Questions and Lectures\n3) Difference between Bundle Id of question and task container id \n4) difference between two type of lectures which Concept/solving questions\n5) What kind of evaluations student undergoes only Questions asked or any thing to do with lecture also..\n\n6) What is difference between Part in Question df and Part in Lectures ?", "1064122": "Most questions are already answered in [Data](https://www.kaggle.com/c/riiid-test-answer-prediction/data) or in [Forum](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion).\n\n1) Lectures don't have answers. Consider -1 as NA as mentioned [here](https://www.kaggle.com/c/riiid-test-answer-prediction/data)\n2) Check the first point [here](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/191106)\n3) Answered [here](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/190828#1051526)\n4) They are just (four) different types of lectures\n5) There is no evaluation for lectures as mentioned [here](https://www.kaggle.com/c/riiid-test-answer-prediction/data)\n6) No difference, both refer to same [parts](https://www.iibc-global.org/english/toeic/test/lr/about/format.html)", "1064665": "rohanrao i understand quite a it things now..\n\n1) Timestamp mentions user interaction and first event , are the user interaction and event same stuff ?\n\n2) Why there is word First mentioned here...\n3) Does the clock always resets for any new event tracked for user , stops when the event ends ?", "1065253": "1) Interaction = Event\n2) Because timestamp is relative to the **first** completed event of the user. That's why every user starts at timestamp = 0\n3) No, the clock for a user starts only once when the first event is completed.\n\nHere's an example for 2) and 3)\n\nIf a user completed events at following actual times:\nFirst - 10:53am on any day\nSecond - 10:57am on same day\nThird - 11:22am on same day\n\nThe timestamp values in data would be\nFirst - timestamp = 0\nSecond - timestamp = 240000 (4 minutes)\nThird - timestamp = = 1740000 (29 minutes)", "1065303": "1) but what if second event comes in days after first event.. can this be possibility too.. in that case what is meaningful info it can carry.. i think in that case one would want to check relative time elapsed wrt to previous event ?\n\n2) From training point of view.. can one remove all the lecture events since they dont have a target ?", "1065831": "1) Indeed, ~ 87% of timestamps are > 24H so you have to deal with it.\n2) Yes you can remove lectures events doing :\n\n`train[train[\"content_type_id\"] == 0]`", "1067181": "matthieuplante thanku\n\n1) can we assume that Private+Public test set will contain uniquely Question id + Lecture id that would all exist in questions.csv/lecture.csv ?\n\n2) Can we assume Tag id represents possibly the subject domain ?\n\n3) Why do we have multiple tags in Question.csv for each question ?", "1067463": "1) Yeah I think I saw that in a discussion \n\n2) Yes\n\n3) Because a question can concern multiple domains. We don't have the signification of the tags but for example a question can concern the financial domain and the newspaper domain so it will have 2 tags (Not the best example but I think you will understand ;) )" }, "source": "meta" }