What kind of questions and answers are actually being made in Quora?
\n\nIn the data provided for the competition, what type of question was tagged as \"insincere\"?
\n\nI scoring words for each class using weights based on target values. Please look at this kernel. (https://www.kaggle.com/tanreinama/finding-negative-words#)
\n\nIn the bottom 20 of the score, words about race (black, white, american, indian, women, muslim) and current incumbent president appeared. In the top 20, words about comparison (best, good, differ) has appeared.
\n\nIn the top 20 (not insincere - a suitable question) people question using 'what' or 'how', whereas 'whi' is used in lower order.
", "messages": [ { "id": 417729, "postDate": "2018-11-08T17:40:19.520Z", "content": "You can check the data description page: https://www.kaggle.com/c/quora-insincere-questions-classification/data
\n\nAn insincere question is defined as a question intended to make a statement rather than look for helpful answers. Some characteristics that can signify that a question is insincere:
\n\nHas a non-neutral tone
Is disparaging or inflammatory
Isn't grounded in reality
Uses sexual content (incest, bestiality, pedophilia) for shock value, and not to seek genuine answers
Yes, of course there is a definition of what type is so.\nBut I want to not general theory, and more specifically which word is prohibited.\nAlso, whether there is a deliberate difference in NG topics between the training set and the test set.
", "rawMarkdown": "Yes, of course there is a definition of what type is so.\nBut I want to not general theory, and more specifically which word is prohibited.\nAlso, whether there is a deliberate difference in NG topics between the training set and the test set." }, { "id": 489188, "postDate": "2019-03-13T15:50:09.583Z", "content": "Wasn't the whole point that there is no hard rules on what makes a question insincere? If there were some set of words that automatically made the question insincere then it wouldn't really be a machine learning problem.
", "rawMarkdown": "Wasn't the whole point that there is no hard rules on what makes a question insincere? If there were some set of words that automatically made the question insincere then it wouldn't really be a machine learning problem.", "votes": 1 } ] }, { "id": 456702, "postDate": "2019-01-16T10:39:03.977Z", "content": "@Tanrei(nama). For full details visit this topic https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/77691
", "rawMarkdown": "@Tanrei(nama). For full details visit this topic https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/77691", "votes": 1 }, { "id": 417421, "postDate": "2018-11-08T08:33:37.700Z", "content": "What kind of questions and answers are actually being made in Quora?
\n\nIn the data provided for the competition, what type of question was tagged as \"insincere\"?
\n\nI scoring words for each class using weights based on target values. Please look at this kernel. (https://www.kaggle.com/tanreinama/finding-negative-words#)
\n\nIn the bottom 20 of the score, words about race (black, white, american, indian, women, muslim) and current incumbent president appeared. In the top 20, words about comparison (best, good, differ) has appeared.
\n\nIn the top 20 (not insincere - a suitable question) people question using 'what' or 'how', whereas 'whi' is used in lower order.
", "rawMarkdown": "What kind of questions and answers are actually being made in Quora?\n\nIn the data provided for the competition, what type of question was tagged as \"insincere\"?\n\nI scoring words for each class using weights based on target values. Please look at this kernel. (https://www.kaggle.com/tanreinama/finding-negative-words#)\n\nIn the bottom 20 of the score, words about race (black, white, american, indian, women, muslim) and current incumbent president appeared. In the top 20, words about comparison (best, good, differ) has appeared.\n\nIn the top 20 (not insincere - a suitable question) people question using 'what' or 'how', whereas 'whi' is used in lower order.", "votes": 1 }, { "id": 417833, "postDate": "2018-11-08T20:59:19.347Z", "content": "I'm noticing another category that's sometimes getting flagged as insincere is some stuff with (generally lengthy) math equations. I'm not sure if it's on the basis that the equations are ridiculous or if it's 'do my homework for me' that's getting it flagged. Look for [math] tag.
", "rawMarkdown": "I'm noticing another category that's sometimes getting flagged as insincere is some stuff with (generally lengthy) math equations. I'm not sure if it's on the basis that the equations are ridiculous or if it's 'do my homework for me' that's getting it flagged. Look for [math] tag.", "votes": 2 }, { "id": 417784, "postDate": "2018-11-08T19:24:31.667Z", "content": "When I read questions that were marked as \"1\" I agree with the marking, even though I do not always know to say why. And that is also for me, for the questions marked as \"0\". What is it that make us \"know\" to tell an insincere question from a sincere one? If we can speculate about it, maybe it will help us in building better models..
", "rawMarkdown": "When I read questions that were marked as \"1\" I agree with the marking, even though I do not always know to say why. And that is also for me, for the questions marked as \"0\". What is it that make us \"know\" to tell an insincere question from a sincere one? If we can speculate about it, maybe it will help us in building better models.." }, { "id": 417710, "postDate": "2018-11-08T17:10:01.027Z", "content": "Very good question. I guess from a practical viewpoint, in this dataset insincere questions are those that are flagged as 1 for the target variable. Besides single words, some of the EDA kernels also give interesting insights into common phrases and other features as well.
", "rawMarkdown": "Very good question. I guess from a practical viewpoint, in this dataset insincere questions are those that are flagged as 1 for the target variable. Besides single words, some of the EDA kernels also give interesting insights into common phrases and other features as well." }, { "id": 417838, "postDate": "2018-11-08T21:13:24.113Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 417775, "postDate": "2018-11-08T19:07:18.160Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 417729, "author_name": "Shujian Liu", "author_url": "", "post_date": "2018-11-08T17:40:19.520000", "content": "You can check the data description page: https://www.kaggle.com/c/quora-insincere-questions-classification/data
\n\nAn insincere question is defined as a question intended to make a statement rather than look for helpful answers. Some characteristics that can signify that a question is insincere:
\n\nHas a non-neutral tone
Is disparaging or inflammatory
Isn't grounded in reality
Uses sexual content (incest, bestiality, pedophilia) for shock value, and not to seek genuine answers
Yes, of course there is a definition of what type is so.\nBut I want to not general theory, and more specifically which word is prohibited.\nAlso, whether there is a deliberate difference in NG topics between the training set and the test set.
", "votes": 0, "replies": [] }, { "id": 489188, "author_name": "DavidSkarbrevik", "author_url": "", "post_date": "2019-03-13T15:50:09.583000", "content": "Wasn't the whole point that there is no hard rules on what makes a question insincere? If there were some set of words that automatically made the question insincere then it wouldn't really be a machine learning problem.
", "votes": 1, "replies": [] } ] }, { "id": 456702, "author_name": "cyberia", "author_url": "", "post_date": "2019-01-16T10:39:03.977000", "content": "@Tanrei(nama). For full details visit this topic https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/77691
", "votes": 1, "replies": [] }, { "id": 417833, "author_name": "LisaGaudette", "author_url": "", "post_date": "2018-11-08T20:59:19.347000", "content": "I'm noticing another category that's sometimes getting flagged as insincere is some stuff with (generally lengthy) math equations. I'm not sure if it's on the basis that the equations are ridiculous or if it's 'do my homework for me' that's getting it flagged. Look for [math] tag.
", "votes": 2, "replies": [] }, { "id": 417784, "author_name": "Abo Sol", "author_url": "", "post_date": "2018-11-08T19:24:31.667000", "content": "When I read questions that were marked as \"1\" I agree with the marking, even though I do not always know to say why. And that is also for me, for the questions marked as \"0\". What is it that make us \"know\" to tell an insincere question from a sincere one? If we can speculate about it, maybe it will help us in building better models..
", "votes": 0, "replies": [] }, { "id": 417710, "author_name": "averagemn", "author_url": "", "post_date": "2018-11-08T17:10:01.027000", "content": "Very good question. I guess from a practical viewpoint, in this dataset insincere questions are those that are flagged as 1 for the target variable. Besides single words, some of the EDA kernels also give interesting insights into common phrases and other features as well.
", "votes": 0, "replies": [] }, { "id": 417838, "author_name": "", "author_url": "", "post_date": "2018-11-08T21:13:24.113000", "content": "", "votes": 1, "replies": [] }, { "id": 417775, "author_name": "", "author_url": "", "post_date": "2018-11-08T19:07:18.160000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "417729": "You can check the data description page: https://www.kaggle.com/c/quora-insincere-questions-classification/data\n\nAn insincere question is defined as a question intended to make a statement rather than look for helpful answers. Some characteristics that can signify that a question is insincere:\n\n - Has a non-neutral tone\n\n - Is disparaging or inflammatory\n\n - Isn't grounded in reality\n\n - Uses sexual content (incest, bestiality, pedophilia) for shock value, and not to seek genuine answers", "456702": "@Tanrei(nama). For full details visit this topic https://www.kaggle.com/c/quora-insincere-questions-classification/discussion/77691", "417421": "What kind of questions and answers are actually being made in Quora?\n\nIn the data provided for the competition, what type of question was tagged as \"insincere\"?\n\nI scoring words for each class using weights based on target values. Please look at this kernel. (https://www.kaggle.com/tanreinama/finding-negative-words#)\n\nIn the bottom 20 of the score, words about race (black, white, american, indian, women, muslim) and current incumbent president appeared. In the top 20, words about comparison (best, good, differ) has appeared.\n\nIn the top 20 (not insincere - a suitable question) people question using 'what' or 'how', whereas 'whi' is used in lower order.", "417833": "I'm noticing another category that's sometimes getting flagged as insincere is some stuff with (generally lengthy) math equations. I'm not sure if it's on the basis that the equations are ridiculous or if it's 'do my homework for me' that's getting it flagged. Look for [math] tag.", "417784": "When I read questions that were marked as \"1\" I agree with the marking, even though I do not always know to say why. And that is also for me, for the questions marked as \"0\". What is it that make us \"know\" to tell an insincere question from a sincere one? If we can speculate about it, maybe it will help us in building better models..", "417710": "Very good question. I guess from a practical viewpoint, in this dataset insincere questions are those that are flagged as 1 for the target variable. Besides single words, some of the EDA kernels also give interesting insights into common phrases and other features as well.", "417838": "", "417775": "" } }