Yioop!

Release Notes: This release adds initial support for word suggestions as a user types in queries. The bigramming used to speed common two word queries now works with n word grams. N word gram filter files can now be created using Wikipedia raw page count dumps. This version adds support for * and $ in allowed- and disallowed-to-crawl sites. Using this, the user can crawl sites to a fixed depth. Robots.txt processing now supports * and $ in robot.txt paths. Support for NOSNIPPET, NOARCHIVE, and X-Robots-tag HTTP headers has also been implemented. A tool for editing search summaries after a crawl has also been added.

Release Notes: The crawler now has its own DNS caching mechanism independent of cURL's. Yioop now has a detection mechanism for when websites are becoming congested. The user can also set a quota on the number of URLs downloaded per hour from sites. A webcrawl statistics page can now be generated for a crawl. Bugs in robots.txt handling and in archive handling, which were introduced in 0.82, have been fixed. The demo site now features an example crawl of 100 million pages crawled with the previous version of the software.

Release Notes: This release improved scalability by allowing multiple machines to maintain portions of the "to crawl next" queue. Query processing can also be split amongst machines, with different machines being responsible for documents of a given hash. Yioop! now supports mirroring of machines. Two word phrases as determined by an XML file such as Wikipedia URL dump can now be treated as a logical unit. The Yioop! model-view-controller framework has been made easier to extend and documentation for it has been added to the website.

Release Notes: This version supports starting, stopping, and viewing log files of the queue server and fetchers from a Web interface. One can now inject new URLs into an active crawl via a Web interface. This version of Yioop! supports re-crawling of pages after a fixed number of days. Also, the file extensions that are crawled, the number of bytes downloaded per page, and how Yioop! weighs different page components can now all be controlled through a Web interface rather than just the config.php file. Improvements have also been made to how HTML Processor extracts text to index.

Release Notes: Character n-grams are now supported for many languages that did not have a stemmer. Language detection was improved and better UTF-8 preparation was provided for downloads. Yioop!'s ability to following redirects, including bit.ly redirects, was improved. Proximity scoring of text in documents has also been enhanced.