mirror of
https://github.com/wassname/Open-Assistant.git
synced 2026-07-15 01:00:53 +08:00
86 KiB
86 KiB
In [ ]:
from bs4 import BeautifulSoup as bs
import pandas as pd
import requests
import jsonIn [ ]:
base_url = "https://ia600107.us.archive.org/view_archive.php?archive=/27/items/stackexchange/{0}&file=Posts.xml"
def get_all_filenames():
response = requests.get("https://archive.org/download/stackexchange")
if response.ok:
soup = bs(response.content, "html.parser")
table = soup.find("table")
link_tags = table.find_all("a")
urls = {}
for link in link_tags:
url = link["href"]
name = url.split(".stackexchange")[0].replace(".", "_").replace("-", "_")
if url.endswith("7z"):
urls[name] = base_url.format(url)
return urls
urls = get_all_filenames()
print(urls.keys())
print(urls.get("ai"))dict_keys(['3dprinting_meta', '3dprinting', 'Stackoverflow_com_Posts_7z', 'academia_meta', 'academia', 'ai_meta', 'ai', 'android_meta', 'android', 'anime_meta', 'anime', 'apple_meta', 'apple', 'arduino_meta', 'arduino', 'askubuntu_com_7z', 'astronomy_meta', 'astronomy', 'aviation_meta', 'aviation', 'avp_meta', 'avp', 'beer_meta', 'beer', 'bicycles_meta', 'bicycles', 'bioacoustics_meta', 'bioacoustics', 'bioinformatics_meta', 'bioinformatics', 'biology_meta', 'biology', 'bitcoin_meta', 'bitcoin', 'blender_meta', 'blender', 'boardgames_meta', 'boardgames', 'bricks_meta', 'bricks', 'buddhism_meta', 'buddhism', 'cardano_meta', 'cardano', 'chemistry_meta', 'chemistry', 'chess_meta', 'chess', 'chinese_meta', 'chinese', 'christianity_meta', 'christianity', 'civicrm_meta', 'civicrm', 'codegolf_meta', 'codegolf', 'codereview_meta', 'codereview', 'coffee_meta', 'coffee', 'cogsci_meta', 'cogsci', 'computergraphics_meta', 'computergraphics', 'conlang_meta', 'conlang', 'cooking_meta', 'cooking', 'craftcms_meta', 'craftcms', 'crafts_meta', 'crafts', 'crypto_meta', 'crypto', 'cs_meta', 'cs', 'cseducators_meta', 'cseducators', 'cstheory_meta', 'cstheory', 'datascience_meta', 'datascience', 'dba_meta', 'dba', 'devops_meta', 'devops', 'diy_meta', 'diy', 'drones_meta', 'drones', 'drupal_meta', 'drupal', 'dsp_meta', 'dsp', 'earthscience_meta', 'earthscience', 'ebooks_meta', 'ebooks', 'economics_meta', 'economics', 'electronics_meta', 'electronics', 'elementaryos_meta', 'elementaryos', 'ell_meta', 'ell', 'emacs_meta', 'emacs', 'engineering_meta', 'engineering', 'english_meta', 'english', 'eosio_meta', 'eosio', 'es_meta_stackoverflow_com_7z', 'es_stackoverflow_com_7z', 'esperanto_meta', 'esperanto', 'ethereum_meta', 'ethereum', 'expatriates_meta', 'expatriates', 'expressionengine_meta', 'expressionengine', 'fitness_meta', 'fitness', 'freelancing_meta', 'freelancing', 'french_meta', 'french', 'gamedev_meta', 'gamedev', 'gaming_meta', 'gaming', 'gardening_meta', 'gardening', 'genealogy_meta', 'genealogy', 'german_meta', 'german', 'gis_meta', 'gis', 'graphicdesign_meta', 'graphicdesign', 'ham_meta', 'ham', 'hardwarerecs_meta', 'hardwarerecs', 'health_meta', 'health', 'hermeneutics_meta', 'hermeneutics', 'hinduism_meta', 'hinduism', 'history_meta', 'history', 'homebrew_meta', 'homebrew', 'hsm_meta', 'hsm', 'interpersonal_meta', 'interpersonal', 'iot_meta', 'iot', 'iota_meta', 'iota', 'islam_meta', 'islam', 'italian_meta', 'italian', 'ja_meta_stackoverflow_com_7z', 'ja_stackoverflow_com_7z', 'japanese_meta', 'japanese', 'joomla_meta', 'joomla', 'judaism_meta', 'judaism', 'korean_meta', 'korean', 'languagelearning_meta', 'languagelearning', 'latin_meta', 'latin', 'law_meta', 'law', 'lifehacks_meta', 'lifehacks', 'linguistics_meta', 'linguistics', 'literature_meta', 'literature', 'magento_meta', 'magento', 'martialarts_meta', 'martialarts', 'materials_meta', 'materials', 'math_meta', 'math', 'matheducators_meta', 'matheducators', 'mathematica_meta', 'mathematica', 'mathoverflow_net_7z', 'mechanics_meta', 'mechanics', 'meta_askubuntu_com_7z', 'meta_mathoverflow_net_7z', 'meta_serverfault_com_7z', 'meta', 'meta_stackoverflow_com_7z', 'meta_superuser_com_7z', 'moderators_meta', 'moderators', 'monero_meta', 'monero', 'money_meta', 'money', 'movies_meta', 'movies', 'music_meta', 'music', 'musicfans_meta', 'musicfans', 'mythology_meta', 'mythology', 'networkengineering_meta', 'networkengineering', 'opendata_meta', 'opendata', 'opensource_meta', 'opensource', 'or_meta', 'or', 'outdoors_meta', 'outdoors', 'parenting_meta', 'parenting', 'patents_meta', 'patents', 'pets_meta', 'pets', 'philosophy_meta', 'philosophy', 'photo_meta', 'photo', 'physics_meta', 'physics', 'pm_meta', 'pm', 'poker_meta', 'poker', 'politics_meta', 'politics', 'portuguese_meta', 'portuguese', 'proofassistants_meta', 'proofassistants', 'pt_meta_stackoverflow_com_7z', 'pt_stackoverflow_com_7z', 'puzzling_meta', 'puzzling', 'quant_meta', 'quant', 'quantumcomputing_meta', 'quantumcomputing', 'raspberrypi_meta', 'raspberrypi', 'retrocomputing_meta', 'retrocomputing', 'reverseengineering_meta', 'reverseengineering', 'robotics_meta', 'robotics', 'rpg_meta', 'rpg', 'ru_meta_stackoverflow_com_7z', 'ru_stackoverflow_com_7z', 'rus_meta', 'rus', 'russian_meta', 'russian', 'salesforce_meta', 'salesforce', 'scicomp_meta', 'scicomp', 'scifi_meta', 'scifi', 'security_meta', 'security', 'serverfault_com_7z', 'sharepoint_meta', 'sharepoint', 'sitecore_meta', 'sitecore', 'skeptics_meta', 'skeptics', 'softwareengineering_meta', 'softwareengineering', 'softwarerecs_meta', 'softwarerecs', 'solana_meta', 'solana', 'sound_meta', 'sound', 'space_meta', 'space', 'spanish_meta', 'spanish', 'sports_meta', 'sports', 'sqa_meta', 'sqa', 'stackapps_com_7z', 'stackoverflow_com_Badges_7z', 'stackoverflow_com_Comments_7z', 'stackoverflow_com_PostHistory_7z', 'stackoverflow_com_PostLinks_7z', 'stackoverflow_com_Tags_7z', 'stackoverflow_com_Users_7z', 'stackoverflow_com_Votes_7z', 'stats_meta', 'stats', 'stellar_meta', 'stellar', 'substrate_meta', 'substrate', 'superuser_com_7z', 'sustainability_meta', 'sustainability', 'tex_meta', 'tex', 'tezos_meta', 'tezos', 'tor_meta', 'tor', 'travel_meta', 'travel', 'tridion_meta', 'tridion', 'ukrainian_meta', 'ukrainian', 'unix_meta', 'unix', 'ux_meta', 'ux', 'vegetarianism_meta', 'vegetarianism', 'vi_meta', 'vi', 'webapps_meta', 'webapps', 'webmasters_meta', 'webmasters', 'windowsphone_meta', 'windowsphone', 'woodworking_meta', 'woodworking', 'wordpress_meta', 'wordpress', 'workplace_meta', 'workplace', 'worldbuilding_meta', 'worldbuilding', 'writers_meta', 'writers']) https://ia600107.us.archive.org/view_archive.php?archive=/27/items/stackexchange/ai.stackexchange.com.7z&file=Posts.xml
In [ ]:
xml_format_map = {
"Id": int,
"PostTypeId": int,
"CreationDate": str,
"Score": int,
"ViewCount": int,
"Body": str,
"AnswerCount": int,
"CommentCount": int,
"ContentLicense": str,
"AcceptedAnswerId": int,
"ParentId": int,
}
# def extract_xml_file(file_url: str):
# table = pd.read_xml(file_url)
# return table
def xml_to_df(response: str):
"""
Collect and Manually import XML into Dataframe
pd.read_xml() errors when XML trees are too large, this is just a hack to
download a XML file and parse into a Dataframe. **Not Tested on huge XML files**
Parameters:
response (Requests.Response): Requests response object with the XML data
Returns:
df (DataFrame): A Dataframe from the XML file
"""
soup = bs(response.content, "xml")
posts = soup.find_all("row")
all_posts = [post.attrs for post in posts]
df = pd.DataFrame(all_posts)
df.AnswerCount.fillna(0, inplace=True)
df.ViewCount.fillna(0, inplace=True)
df.AcceptedAnswerId.fillna(0, inplace=True)
df.ParentId.fillna(0, inplace=True)
df["DataSource"] = response.url
df = df.astype(xml_format_map)
return df
dataset_name = "ai"
xml_posts_path = urls.get(dataset_name)
# df = extract_xml_file(test)
response = requests.get(xml_posts_path)
df = xml_to_df(response)
print(df.dtypes)
df.head()Id int64 PostTypeId int64 AcceptedAnswerId int64 CreationDate object Score int64 ViewCount int64 Body object OwnerUserId object LastEditorUserId object LastEditDate object LastActivityDate object Title object Tags object AnswerCount int64 CommentCount int64 ContentLicense object ParentId int64 ClosedDate object FavoriteCount object CommunityOwnedDate object LastEditorDisplayName object OwnerDisplayName object DataSource object dtype: object
| Id | PostTypeId | AcceptedAnswerId | CreationDate | Score | ViewCount | Body | OwnerUserId | LastEditorUserId | LastEditDate | ... | AnswerCount | CommentCount | ContentLicense | ParentId | ClosedDate | FavoriteCount | CommunityOwnedDate | LastEditorDisplayName | OwnerDisplayName | DataSource | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 1 | 3 | 2016-08-02T15:39:14.947 | 10 | 710 | <p>What does "backprop" mean? Is the "backprop... | 8 | 2444 | 2019-11-16T17:56:22.093 | ... | 5 | 0 | CC BY-SA 4.0 | 0 | NaN | NaN | NaN | NaN | NaN | https://ia600107.us.archive.org/view_archive.p... |
| 1 | 2 | 1 | 9 | 2016-08-02T15:40:20.623 | 14 | 1008 | <p>Does increasing the noise in data help to i... | 8 | 2444 | 2019-02-23T22:36:19.090 | ... | 3 | 0 | CC BY-SA 4.0 | 0 | NaN | NaN | NaN | NaN | NaN | https://ia600107.us.archive.org/view_archive.p... |
| 2 | 3 | 2 | 0 | 2016-08-02T15:40:24.820 | 15 | 0 | <p>"Backprop" is the same as "backpropagation"... | 4 | NaN | NaN | ... | 0 | 0 | CC BY-SA 3.0 | 1 | NaN | NaN | NaN | NaN | NaN | https://ia600107.us.archive.org/view_archive.p... |
| 3 | 4 | 1 | 12 | 2016-08-02T15:41:22.020 | 33 | 1266 | <p>When you're writing your algorithm, how do ... | 8 | 2444 | 2021-01-19T23:54:07.813 | ... | 4 | 0 | CC BY-SA 3.0 | 0 | NaN | NaN | NaN | NaN | NaN | https://ia600107.us.archive.org/view_archive.p... |
| 4 | 6 | 1 | 20 | 2016-08-02T15:43:35.460 | 7 | 279 | <p>Given the following definition of an intell... | 29 | 2444 | 2019-06-15T18:25:58.513 | ... | 2 | 0 | CC BY-SA 4.0 | 0 | NaN | NaN | NaN | NaN | NaN | https://ia600107.us.archive.org/view_archive.p... |
5 rows × 23 columns
In [ ]:
def filter_only_questions_with_accepted_answers(df):
"""**TODO**
Filter only to Questions with Accepted Answers
Filter dataframe by questions that have accepted answers, should also include
all rows of answers for those questions, even if not accepted.
Parameters:
df (DataFrame): containing a "AcceptedAnswerId", "Id", and "ParentId" columns
Returns:
df (DataFrame): current dataframe with filtered results
"""
df = df[(df["AcceptedAnswerId"].notnull()) | (df["ParentId"] == df["Id"])]
def filter_scores_above(df, question_score_threshold: int = 20, answer_score_threshold: int = 20):
"""**TODO**
Filter Dataframe by minimum scores
Filter Question and Answer columns by score thresholds to trim lower scoring results
Parameters:
df (DataFrame): containing a "Score" column
Returns:
df (DataFrame): current dataframe with filtered results
"""
df = df[
((df["Score"] >= question_score_threshold) & (df.PostTypeId == 1))
| ((df["Score"] >= answer_score_threshold) & (df.PostTypeId == 2))
]
def convert_html_to_text(df, column: str = "Body"):
"""
Convert HTML tags to pure text
Feeds HTML text body into BeautifulSoup to parse it to only text. Set aside as
function to provide option to skip
Parameters:
df (DataFrame): containing a "Body" column with HTML
Returns:
df (DataFrame): current dataframe with parsed column
"""
df.dropna(subset=[column], inplace=True)
df[f"{column}Clean"] = df[column].apply(lambda row: bs(row, "html.parser").text)
def clean_tags(df):
"""
Convert Tags into Comma separated
Converts Tag slugs into commas separated tags
Parameters:
df (DataFrame): containing a "Tags" column with slugs
Returns:
df (DataFrame): current dataframe with parsed column
"""
df["TagsClean"] = df["Tags"].str.replace("-", " ").str.replace("><", ", ").str.replace("<", "").str.replace(">", "")
# filter_only_questions_with_accepted_answers(df)
# filter_scores_above(df)
convert_html_to_text(df)
clean_tags(df)
df[["Body", "BodyClean", "Tags", "TagsClean"]]
# print(df.shape)| Body | BodyClean | Tags | TagsClean | |
|---|---|---|---|---|
| 0 | <p>What does "backprop" mean? Is the "backprop... | What does "backprop" mean? Is the "backprop" t... | <neural-networks><backpropagation><terminology... | neural networks, backpropagation, terminology,... |
| 1 | <p>Does increasing the noise in data help to i... | Does increasing the noise in data help to impr... | <neural-networks><machine-learning><statistica... | neural networks, machine learning, statistical... |
| 2 | <p>"Backprop" is the same as "backpropagation"... | "Backprop" is the same as "backpropagation": i... | NaN | NaN |
| 3 | <p>When you're writing your algorithm, how do ... | When you're writing your algorithm, how do you... | <neural-networks><hyperparameter-optimization>... | neural networks, hyperparameter optimization, ... |
| 4 | <p>Given the following definition of an intell... | Given the following definition of an intellige... | <philosophy><definitions><intelligent-agent> | philosophy, definitions, intelligent agent |
| ... | ... | ... | ... | ... |
| 23174 | <p>The purpose of evaluating the state and act... | The purpose of evaluating the state and action... | NaN | NaN |
| 23175 | <p>In machine translation, convolution is a te... | In machine translation, convolution is a techn... | NaN | NaN |
| 23176 | <p>One of the key features of ChatGPT is its a... | One of the key features of ChatGPT is its abil... | NaN | NaN |
| 23177 | <p>Given a neural network model for Covid-19 c... | Given a neural network model for Covid-19 clas... | <neural-networks><homework> | neural networks, homework |
| 23178 | <p>My question is more related to the fundamen... | My question is more related to the fundamental... | <search><constraint-satisfaction-problems> | search, constraint satisfaction problems |
23179 rows × 4 columns
In [ ]:
questions = df[df.PostTypeId == 1]
answers = df[df.PostTypeId == 2]
df = pd.merge(
questions,
answers[
[
"Id",
"CreationDate",
"Score",
"ViewCount",
"CommentCount",
"ContentLicense",
"TagsClean",
"BodyClean",
"ParentId",
]
],
left_on="Id",
right_on="ParentId",
suffixes=("_q", "_a"),
how="left",
)
df["AcceptedAnswerFlag"] = df.apply(lambda row: row["Id_a"] == row["AcceptedAnswerId"], axis=1)
df = df.rename(
columns={
"BodyClean_q": "Question",
"Score_q": "QuestionScore",
"TagsClean_q": "QuestionTags",
"BodyClean_a": "Answer",
"Score_a": "AnswerScore",
"ContentLicense_q": "QuestionContentLicense",
"ContentLicense_a": "AnswerContentLicense",
"CreationDate_q": "CreationDate",
}
)
## Set the number of results to a lower number to only return top N rated Answers.
number_of_results = 25
df = (
df.sort_values(by=["AcceptedAnswerFlag", "AnswerScore"], ascending=[False, False])
.groupby("Question")
.head(number_of_results)
.reset_index(drop=True)
)
df[["Id_q", "Question", "QuestionScore", "QuestionTags", "Id_a", "Answer", "AnswerScore", "AcceptedAnswerFlag"]].head()| Id_q | Question | QuestionScore | QuestionTags | Id_a | Answer | AnswerScore | AcceptedAnswerFlag | |
|---|---|---|---|---|---|---|---|---|
| 0 | 1768 | In Portal 2 we see that AI's can be "killed" b... | 175 | philosophy, decision theory, mythology of ai, ... | 1769.0 | This classic problem exhibits a basic misunder... | 146.0 | True |
| 1 | 10623 | What is self-supervised learning in machine le... | 91 | machine learning, comparison, supervised learn... | 10624.0 | Introduction\nThe term self-supervised learnin... | 90.0 | True |
| 2 | 111 | Obviously, self-driving cars aren't perfect, s... | 100 | philosophy, ethics, autonomous vehicles, decis... | 1790.0 | \nHow could self-driving cars make ethical dec... | 76.0 | True |
| 3 | 14224 | If the original purpose for developing AI was ... | 69 | philosophy, social, explainable ai | 14247.0 | As argued by Selvaraju et al., there are three... | 75.0 | True |
| 4 | 1479 | Do scientists or research experts know from th... | 94 | neural networks, deep learning, convolutional ... | 4044.0 | There are many approaches that aim to make a t... | 69.0 | True |
In [ ]:
testing_id = df.Id_q.mode()[0]
df[(df.Id_q == testing_id) | (df.ParentId_a == testing_id)][
["Id_q", "Question", "ParentId_a", "AcceptedAnswerId", "Id_a", "Answer", "AnswerScore", "AcceptedAnswerFlag"]
]
# df[['Id_q', 'Question', 'ParentId_a', 'AcceptedAnswerId', 'Id_a', 'Answer', 'AnswerScore', 'AcceptedAnswerFlag']]| Id_q | Question | ParentId_a | AcceptedAnswerId | Id_a | Answer | AnswerScore | AcceptedAnswerFlag | |
|---|---|---|---|---|---|---|---|---|
| 7 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15744.0 | I think this is a fairly common misconception ... | 62.0 | True |
| 3662 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15753.0 | I think your premise is flawed.\nYou seem to a... | 19.0 | False |
| 3713 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15747.0 | TL;DR: The subtleties of infinity are made app... | 12.0 | False |
| 3788 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15756.0 | In Haskell, you can type:\nprint [1..]\nand it... | 9.0 | False |
| 3821 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15758.0 | I believe humans can be said to understand inf... | 8.0 | False |
| 3882 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15762.0 | (There's a summary at the bottom for those who... | 7.0 | False |
| 4389 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15783.0 | Then premise assumes that humans "understand" ... | 4.0 | False |
| 4849 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15740.0 | By adding some rules for infinity in arithmeti... | 3.0 | False |
| 4850 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15803.0 | I think the concept that is missing in the dis... | 3.0 | False |
| 5763 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15768.0 | Computers don't understand "infinity" or even ... | 2.0 | False |
| 5764 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15810.0 | The Questions That Computers Can Never Answer ... | 2.0 | False |
| 5765 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15943.0 | John Doucette's answer covers my thoughts on t... | 2.0 | False |
| 7462 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15779.0 | I would think that a computer couldn’t underst... | 1.0 | False |
| 7463 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15787.0 | The "concept" of infinity is 1 thing to unders... | 1.0 | False |
| 7464 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15801.0 | Just food for thought: how about if we try to ... | 1.0 | False |
| 7465 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15930.0 | Its arguable if we humans understand infinity.... | 1.0 | False |
| 7466 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15934.0 | Well -- just to touch on the question of peopl... | 1.0 | False |
| 7467 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15938.0 | Humans certainly don't understand infinity. Cu... | 1.0 | False |
| 9481 | 15730 | As a human being, we can think infinity. In pr... | 15730.0 | 15744 | 15931.0 | I think the property humans have which compute... | 0.0 | False |
In [ ]:
j = (
df.groupby(
["Title", "Question", "QuestionScore", "QuestionTags", "QuestionContentLicense", "DataSource", "CreationDate"]
)
.apply(lambda x: x[["Answer", "AnswerScore", "AcceptedAnswerFlag"]].to_dict("records"))
.reset_index()
.rename(columns={0: "Answers"})
.to_json(orient="records")
)
data = json.loads(j)
for post in data:
if len(post.get("Answers")) >= 4:
print(json.dumps(post, indent=4))
break{
"Title": "1 hidden layer with 1000 neurons vs. 10 hidden layers with 100 neurons",
"Question": "These types of questions may be problem-dependent, but I have tried to find research that addresses the question whether the number of hidden layers and their size (number of neurons in each layer) really matter or not.\nSo my question is, does it really matter if we for example have 1 large hidden layer of 1000 neurons vs. 10 hidden layers with 100 neurons each?\n",
"QuestionScore": 16,
"QuestionTags": "neural networks",
"QuestionContentLicense": "CC BY-SA 3.0",
"DataSource": "https://ia600107.us.archive.org/view_archive.php?archive=/27/items/stackexchange/ai.stackexchange.com.7z&file=Posts.xml",
"CreationDate": "2017-05-04T13:06:37.990",
"Answers": [
{
"Answer": "Basically, having multiple layers (aka a deep network) makes your network more eager to recognize certain aspects of input data. For example, if you have the details of a house (size, lawn size, location etc.) as input and want to predict the price. The first layer may predict:\n\nBig area, higher price\nSmall amount of bedrooms, lower price\n\nThe second layer might conclude:\n\nBig area + small amount of bedrooms = large bedrooms = +- effect\n\nYes, one layer can also 'detect' the stats, however it will require more neurons as it cannot rely on other neurons to do 'parts' of the total calculation required to detect that stat.\nCheck out this answer\n",
"AnswerScore": 13.0,
"AcceptedAnswerFlag": true
},
{
"Answer": "There are so many aspects.\n1. Training:\nTraining deep nets is a hard job due to the vanishing (rearly exploding) gradient problem. So building a 10x100 neural-net is not recommended.\n2. Trained network performance:\n\nInformation loss:\nThe classical usage of neural nets is the classification problem. Which means we want to get some well defined information from the data. (Ex. Is there a face in the picture or not.)\nSo usually classification problem has a lot of input, and few output, whats more the size of the hidden layers are descend from input to output.\nHowever, we loss information using less neurons layer by layer. (Ie. We cannot reproduce the original image based on the fact that is there a face on it or no.) So you must know that you loss information using 100 neurons if the size of the input is (lets say) 1000.\nInformation complexity: However the deeper nets (as Tomas W mentioned) can fetch more complex information from the input data. Inspite of this its not recommended to use 10 fully connected layers. Its recommended to use convolutional/relu/maxpooling or other type of layers. Firest layers can compress the some essential part of the inputs. (Ex is there any line in a specific part of the picture) Second layers can say: There is a specific shape in this place in the picture. Etc etc.\n\nSo deeper nets are more \"clever\" but 10x100 net structure is a good choice.\n",
"AnswerScore": 4.0,
"AcceptedAnswerFlag": false
},
{
"Answer": "If the problem you are solving is linearly separable, one layer of 1000 neurons can do better job than 10 layers with each of 100 neurons.\nIf the problem is non linear and not convex, then you need deep neural nets. \n",
"AnswerScore": 1.0,
"AcceptedAnswerFlag": false
},
{
"Answer": "\nI think you have a confusion in the basics of the neural networks.\n Every layer has a separate activation function and input/output\n connection weights.\n\nThe output of the first hidden layer will be multiplied by a weight, processed by an activation function in the next layer and so on.\nSingle layer neural networks are very limited for simple tasks, deeper NN can perform far better than a single layer. \nHowever, do not use more than layer if your application is not fairly complex. In conclusion, 100 neurons layer does not mean better neural network than 10 layers x 10 neurons but 10 layers are something imaginary unless you are doing deep learning. start with 10 neurons in the hidden layer and try to add layers or add more neurons to the same layer to see the difference. learning with more layers will be easier but more training time is required.\n",
"AnswerScore": 0.0,
"AcceptedAnswerFlag": false
}
]
}
In [ ]:
file_name = dataset_name
def save_data(data: list, file_name: str, file_type: str = ["csv", "json", "jsonl", "parquet"]):
"""
Save Data to file
Save Data list to file as either JSON or JSONL
Parameters:
data (list): list of dictionaries
file_name (str): name of file (no extension)
jsonl (bool): to save file as either JSON or JSONL
"""
file_type = file_type.lower()
if file_type == "csv" and isinstance(data, pd.DataFrame):
data.to_csv(f"/content/{file_name}.csv", index=False)
elif file_type == "json" and isinstance(data, list):
print(json.dumps(data, indent=4), file=open(f"/content/{file_name}.json", "w"))
elif file_type == "jsonl" and isinstance(data, list):
for item in data:
print(json.dumps(item), file=open(f"/content/{file_name}.jsonl", "a"))
elif file_type == "parquet" and isinstance(data, pd.DataFrame):
data.to_parquet(f"/content/{file_name}.parquet", index=False)
else:
print("Data should be either of List type for JSON and JSONL, or Pandas Dataframes for CSV and Parquet")
# save_data(data=data, file_name=file_name, file_type='jsonl')
# save_data(data=df, file_name=file_name, file_type='parquet')Data should be either of List type for JSON and JSONL, or Pandas Dataframes for CSV and Parquet
In [ ]:
from typing import TypeVar, List, Dict, Any, Literal
from json import JSONEncoder
T = TypeVar("T", bound="ConversationTreeNode")
class ConversationTreeNode:
text: str # The text of the node
role: Literal["prompter", "assistant"] # Whether the node is a user prompt/follow-up or an assistant response
children: List[T] # The children of the node (if you have a linear conversation, this will be of length 0 or 1)
metadata: Dict[str, Any] # Node metadata (see below)
def __init__(
self, text: str, role: Literal["prompter", "assistant"], children: List[T], metadata: Dict[str, Any]
) -> None:
self.text = text
self.role = role
self.children = children
self.metadata = metadata
class ConversationTree:
root: ConversationTreeNode # The node containing the initial prompt
metadata: Dict[str, Any] # Tree metadata, different from root node metadata.
def __init__(self, root: ConversationTreeNode, metadata: Dict[str, Any]) -> None:
self.root = root
self.metadata = metadata
# subclass JSONEncoder
class TreeEncoder(JSONEncoder):
def default(self, o):
return o.__dict__In [ ]:
conversation_forest = []
tree_metadata_map = {"Title": str, "QuestionContentLicense": str, "DataSource": str, "CreationDate": str}
question_metadata_map = {"QuestionScore": int, "QuestionTags": str}
answer_metadata_map = {"AnswerScore": int, "AcceptedAnswerFlag": bool}
for item in data:
prompt = item.get("Question")
metadata = {k: v for k, v in item.items() if k in question_metadata_map}
root = ConversationTreeNode(text=prompt, role="prompter", children=[], metadata=metadata)
for answer in item.get("Answers"):
response = answer.get("Answer")
metadata = {k: v for k, v in answer.items() if k in answer_metadata_map}
child = ConversationTreeNode(text=response, role="assistant", children=[], metadata=metadata)
root.children.append(child)
metadata = {k: v for k, v in item.items() if k in tree_metadata_map}
conversation_tree = ConversationTree(root=root, metadata=metadata)
conversation_forest.append(conversation_tree)
conversation_forest_json = [
json.loads(TreeEncoder().encode(conversation_tree)) for conversation_tree in conversation_forest
]
# print(json.dumps(conversation_forest_json[0], indent=4))
print(json.dumps(conversation_forest_json, indent=4), file=open(f"/content/{file_name}.json", "w")){
"root": {
"text": "Science Fiction has frequently shown AI to be a threat to the very existence of mankind. AI systems have often been the antagonists in many works of fiction, from 2001: A Space Odyssey through to The Terminator and beyond.\nThe Media seems to buy into this trope as well. And in recent years we have had people like Elon Musk warn us of the dangers of an impending AI revolution, stating that AI is more dangerous than nukes.\nAnd, apparently, experts think that we will be seeing this AI revolution in the next 100 years.\nHowever, from my (albeit limited) study of AI, I get the impression that they are all wrong. I am going to outline my understanding below, please correct me if I am wrong:\n\nFirstly, all of these things seem to be confusing Artificial Intelligence with Artificial Consciousness. AI is essentially a system to make intelligent decisions, whereas AC is more like the \"self-aware\" systems that are shown in science fiction.\n\nNot AI itself, but intelligence and intelligent decision-making algorithms are something we've been working with and enhancing since before computers have been around. Moving this over to an artificial framework is fairly easy. However, consciousness is still something we are learning about. My guess is we won't be able to re-create something artificially if we barely understand how it works in the real world.\n\nSo, my conclusion is that no AI system will be able to learn enough to start thinking for itself, and that all our warnings of AI are completely unjustified.\n\nThe real danger comes from AC, which we are a long, long way from realizing because we are still a long way off from defining exactly what consciousness is, let alone understanding it.\n\n\n\nSo, my question is, assuming that my understanding is correct, are any efforts are being made by companies or organizations that work with AI to correct these popular misunderstandings in sci-fi, the media, and/or the public?\nOr are the proponents of AI ambivalent towards this public fear-mongering?\nI understand that the fear mongering is going to remain popular for some time, as bad news sells better than good news. I am just wondering if the general attitude from AI organizations is to ignore this popular misconception, or whether a concerted effort is being made to fight against these AI myths (but unfortunately nobody in the media is listening or cares).\n",
"role": "prompter",
"children": [
{
"text": "Nothing. \nIts in almost everyone's favor for it to stay that way financially. Having non-technical individuals associate AI with terminators makes a perception that the field has greater capabilities than it does $\\rightarrow$ this leads to grants, funding, etc... \nIs there any negative? Yes. Misconceptions always have drawbacks. We see the creation of dumb ethics boards and such cough cough Elon Musk.\nBut if history has anything to say about this, as the field gains popularity (which it is dnagerously quick), information will spread by definition, and eventually misconceptions will be laid to rest.\nNote that this answer is biased and based upon my own opinions\n",
"role": "assistant",
"children": [],
"metadata": {
"AnswerScore": 2.0,
"AcceptedAnswerFlag": true
}
}
],
"metadata": {
"QuestionScore": 5,
"QuestionTags": "social, artificial consciousness"
}
},
"metadata": {
"Title": "\"AI will kill us all! The machines will rise up!\" - what is being done to dispel such myths?",
"QuestionContentLicense": "CC BY-SA 4.0",
"DataSource": "https://ia600107.us.archive.org/view_archive.php?archive=/27/items/stackexchange/ai.stackexchange.com.7z&file=Posts.xml",
"CreationDate": "2019-10-16T13:57:37.143"
}
}