* "guix search" needs your help
@ 2020-05-11 11:05 zimoun
0 siblings, 0 replies; only message in thread
From: zimoun @ 2020-05-11 11:05 UTC (permalink / raw)
To: Guix Devel
Dear,
I am lacking imagination and I need your help for:
- what should be the query you type,
- then what is the package you expect with such query, in order.
Example:
guix search source control management
# expect: git, mercurial, svn, then other
guix search r cytometry
# expect: r-flowcore r-flowsom then other
Using your imagination, I can benchmark some ranking functions.
As reported, the experience with "guix search" is sometimes poor.
Aside the issue of pager or output format, there is 2 explorations
for improvements discussed in this bug report [1]: faster and more
accurate. The benchmark for "faster" is straightforward and the
exploration is on going. :-)
Currently, the packages are ranked accordingly to 'relevance'. There
is good, bad and ugly about this strategy and the state-of-art Okapi
BM25 ranking function has to be evaluated to know if such information
retrieval strategy is worth or not for search of packages. Well,
thanks to Arun guile-xapian bindings, the BM25 ranking is easy to
benchmark since all the tough work is relayed to Xapian.
[1] http://issues.guix.gnu.org/39258
From my point of view, such pair (query terms, expected packages)
would help to improve the tools:
- maybe "guix search"
- maybe "guix lint" by establishing some metrics indicating is the
synopsis and description are relevant -- compared to all the other
packages.
Thank you in advance for help me in the exploration of such ideas.
Best regards,
simon
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2020-05-11 11:05 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-05-11 11:05 "guix search" needs your help zimoun
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/guix.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).