Update README.md

README.md

@@ -6,4 +6,103 @@ tags:
 pretty_name: 1m huffpost news article links
 size_categories:
 - 100K<n<1M
----
+---
+The dataset consists of >1m links to huffpost news articles. They have been crawled for an educational approach to learn how content could be gathered through reading the websites sitemap(s).
+
+The code can be found here [Link](https://github.com/manzke/huffpost-crawler/)
+
+# huffpost-crawler
+educational - how to use sitemap.xml for crawling a website (huffpost.com - could be any)
+
+Every public website has (or should have) a sitemap.xml. The sitemap.xml allows robots like Google to find links, which should be crawled.
+
+In the most cases the sitemaps can be found with just adding a /sitemap.xml to the domain or in bigger sites, check the robots.txt (ex. https://www.huffpost.com/robots.txt)
+
+The robots.txt tells you where to find the sitemap.xml and if a robot is allowed to follow the links. A site can also manage to not get indexed, when it has the metatags for nofollow, noindex.
+
+Your crawler should respect them to not read old content, privacy reasons, etc.
+
+What does our crawler do?
+- it loads the robots.txt
+- detects the sitemaps
+- loads every sitemap
+- gets the urls from each sitemap
+- which can be either another sitemap or a url
+- we are loading each sitemap
+- storing every url which is an entry as well as the skipped ones
+
+This led to 1,121,569 unique links for new entries.
+
+## robots.txt
+
+A lot of LLM models have been trained on public data. To forbit it, pages has added
+
+User-agent: GPTBot
+Disallow: /
+
+This tells the GPTBot to not crawl the site. Let's hope it complies with it.
+
+Also take the Crawl-delay into account, which tells you how much time should be between every request to not get banned.
+
+Example robots.txt from huffpost.com
+
+```
+# Cambria robots
+
+User-agent: grapeshot
+Disallow: /member
+Disallow: /*?*err_code=404
+Disallow: /search
+Disallow: /search/?*
+
+User-agent: *
+Crawl-delay: 4
+Disallow: /*?*page=
+Disallow: /member
+Disallow: /*?*err_code=404
+Disallow: /search
+Disallow: /search/?*
+Disallow: /mapi/v4/*/user/*
+Disallow: /embed
+
+User-agent: Googlebot
+Allow: /
+Disallow: /*?*err_code=404
+Disallow: /search
+Disallow: /search/?*
+
+User-agent: google-extended
+Disallow: /
+
+User-agent: GPTBot
+Disallow: /
+
+# archives
+Sitemap: https://www.huffpost.com/sitemaps/sitemap-v1.xml
+Sitemap: https://www.huffpost.com/sitemaps/sitemap-google-news.xml
+Sitemap: https://www.huffpost.com/sitemaps/sitemap-google-video.xml
+Sitemap: https://www.huffpost.com/sitemaps/sections.xml
+
+# huffingtonpost.com archive sitemaps
+Sitemap: https://www.huffpost.com/sitemaps-huffingtonpost/sitemap.xml
+Sitemap: https://www.huffpost.com/sitemaps-huffingtonpost/sections.xml
+``` 
+
+## Inspired by
+``` 
+@article{misra2019sarcasm,
+  title={Sarcasm Detection using Hybrid Neural Network},
+  author={Misra, Rishabh and Arora, Prahal},
+  journal={arXiv preprint arXiv:1908.07414},
+  year={2019}
+}
+
+@book{misra2021sculpting,
+  author = {Misra, Rishabh and Grover, Jigyasa},
+  year = {2021},
+  month = {01},
+  pages = {},
+  title = {Sculpting Data for ML: The first act of Machine Learning},
+  isbn = {9798585463570}
+}
+```