Discovery/Status updates/2017-10-30

This is the weekly update for the week starting 2017-10-30 and 2017-11-06

Highlights

Blog post about searching using dashes and exclamation points when they're not part of the expected search syntax (by Trey Jones): "So -happy to meet you: Advanced searching techniques on Wikimedia sites". [1]

Trey wrote a blog post about advanced searching techniques, particularly dealing with dashes and exclamation points when they are *not* search syntax. [2]
Trey finished an analysis of enabling mapping hiragana and katakana for English- and Japanese-language wikis. There was less enthusiasm for the mapping and more problems on the Japanese side, so it is not going to be deployed there at this time. However, we're looking to enable it for other languages. [3]
David finished up refactoring the Searcher class and the SearchContext to allow easy code reuse [4]
Eric fixed an issue with a searchmatch span that was altering searches [5]
Eric and David finished up the work needed for evaluating the training speed and accuracy for 1M and 30M sample training sets and updated the resource usage docs [6]
Stas enabled disambiguation page demoting in Wikidata search [7] and fixed several regressions in prefix search [8] [9] [10].

Chelsy finished up the analysis of the A/B test to test relaxing the retrieval query filter [11] and the report has been published [12]
After the quarterly metrics meeting at the end of October, there was a follow-up question that Chelsy answered [13]
Mikhail and Gehel finished up the creation of a Puppet profile/role for doing R-based heavy stats/ML on Wikimedia Cloud [14]

The Wikipedia.org portal page was updated with new translations [15] and new stats [16] on October 30, 2017

--