Search result affected by a number at end of title #119

New issue

Closed

opened 2018-09-28 00:24:42 +02:00 by tzarebczan · 8 comments

tzarebczan commented

2018-09-28 00:24:42 +02:00

(Migrated from github.com)

This query for AdmiralBulldog Bristleback Aghanim's [YOU CAN'T RUN!] Dota 2 https://lighthouse.lbry.io/search?s=AdmiralBulldog%20Bristleback%20Aghanim%27s%20%5BYOU%20CAN%27T%20RUN!%5D%20Dota%202&size=20&from=0 - brings in high effective amount claims like one, four, six at the top fo the list

Removing the 2 at the end AdmiralBulldog Bristleback Aghanim's [YOU CAN'T RUN!] Dota:
https://lighthouse.lbry.io/search?s=AdmiralBulldog%20Bristleback%20Aghanim%27s%20%5BYOU%20CAN%27T%20RUN!%5D%20Dota%20&size=20&from=0 - brings better results (lbry://three is up there because of you most likely - probably need to blacklist that word.

Not sure what other instances of numbers would affect query results but it may be worth digging into.

This query for `AdmiralBulldog Bristleback Aghanim's [YOU CAN'T RUN!] Dota 2` https://lighthouse.lbry.io/search?s=AdmiralBulldog%20Bristleback%20Aghanim%27s%20%5BYOU%20CAN%27T%20RUN!%5D%20Dota%202&size=20&from=0 - brings in high effective amount claims like one, four, six at the top fo the list Removing the 2 at the end `AdmiralBulldog Bristleback Aghanim's [YOU CAN'T RUN!] Dota`: https://lighthouse.lbry.io/search?s=AdmiralBulldog%20Bristleback%20Aghanim%27s%20%5BYOU%20CAN%27T%20RUN!%5D%20Dota%20&size=20&from=0 - brings better results (lbry://three is up there because of `you` most likely - probably need to blacklist that word. Not sure what other instances of numbers would affect query results but it may be worth digging into.

tiger5226 commented

2018-10-04 02:28:51 +02:00

(Migrated from github.com)

we can blacklist numbers via a cast check. @nikooo777 is that performant in js or is there something else you would recommend?

nikooo777 commented

2018-10-06 00:04:23 +02:00

(Migrated from github.com)

why is the number affecting the results? aren't sequences of less than 3 chars ignored?

tiger5226 commented

2018-10-06 00:59:53 +02:00

(Migrated from github.com)

ahhh thats a good point niko! I will have to check the code to see if any part of the query is not being washed properly.

marcdeb1 commented

2018-10-24 10:46:37 +02:00

(Migrated from github.com)

It seems that the after the query is washed, the result is "admiralbulldog bristleback aghanim'[you can'run!] dota". It is removing spaces that shouldn't be removed, for example it should be "you can't run". This is because of the regex.
Secondly, because of this, sometimes a trailing whitespace remains (when "2" is removed), this causes one of the queries to return all results (query_string with * wildcard). Because of the effective_amount score, top claims will appear first.
I will propose a PR to solve this issue.

It seems that the after the query is washed, the result is "admiralbulldog bristleback aghanim'[you can'run!] dota". It is removing spaces that shouldn't be removed, for example it should be "you can't run". This is because of the regex. Secondly, because of this, sometimes a trailing whitespace remains (when "2" is removed), this causes one of the queries to return all results (query_string with * wildcard). Because of the effective_amount score, top claims will appear first. I will propose a PR to solve this issue.

tiger5226 commented

2018-10-25 01:09:28 +02:00

(Migrated from github.com)

Boom! nice investigation!

tzarebczan commented

2018-11-08 00:38:10 +01:00

(Migrated from github.com)

https://github.com/lbryio/lighthouse/pull/125 did not fix this issue. Removing the 2 at the end still gives better results (without the top claims).

tiger5226 commented

2018-11-08 16:03:57 +01:00

(Migrated from github.com)

@tzarebczan I checked the server, and there was an issue with the update shell script. It was in dev but not production. production should be up to date again.

alyssaoc commented

2019-02-20 15:53:14 +01:00

(Migrated from github.com)

closing, reopen when tom comes back from Japan