chat 2020/5/18

from Nikko Mizutani to Toshiyuki Masui (privately):

I’m working at Microsoft. I’m part of the Bing search team. Currently working on query understanding.

from Nikko Mizutani to Toshiyuki Masui (privately):

SVM, LSI, etc. is actually quite outdated. Currently ranking is dominated by DeepLearning like BERT

from Nikko Mizutani to Toshiyuki Masui (privately):

Deep Learning has actually already been applied by search companies for quite some time now.

from Nikko Mizutani to Toshiyuki Masui (privately):

Currently it is being used by Google actually

from Yung Cheng Yang to Host (privately):

Excuse me, "bird"?

from Toshiyuki Masui to All Participants:

BERT

from Yung Cheng Yang to Host (privately):

ok

from Nikko Mizutani to Toshiyuki Masui (privately):

from Nikko Mizutani to Toshiyuki Masui (privately):

Actually we use the following to measure gains on the SERP(search engine results page):

from 河野勇輝 to Host (privately):

select * from table where cleopatra = 1

from 河野勇輝 to Host (privately):

or cleopatra = True

from Yung Cheng Yang to Host (privately):

just make that sentence a keyword

from 河野勇輝 to Host (privately):

相関類似度っていうんですっけ

from Nikko Mizutani to Toshiyuki Masui (privately):

There is an automated way to do that. I cannot share how it is done due my contract, but you can automate it without a feedback button

from 河野勇輝 to Host (privately):

youtube does that

from 河野勇輝 to Host (privately):

too

from Nikko Mizutani to Toshiyuki Masui (privately):

I built this site(backend and index):

It uses ElasticSearch(a Lucene based index). It employed BM-25 for ranking.

from Nikko Mizutani to Toshiyuki Masui (privately):

It changes as soon as the query changes. Please see the site

from 河野勇輝 to Host (privately):

no hw?

from 河野勇輝 to Host (privately):

what do i do with my scrapbox