Permaculture Information Web

[Chad Knepp <pyg AT galatea.org>]

Hi Robyn,

Thanks for the feedback.

Robyn Williamson writes:
 > Hi friends of permaculture
 > 
 > First of all, many thanks, heartiest congratulations and applause to 
 > all of you who managed to get this off the ground.  Well done!

Most of the credit is really due to Plants For A Future
<http://pfaf.org/>, from which our data is based on.

 > However, and this is FYI only, not a criticism - I just did my first 
 > search for "Grevillea".  It returned 370 matches for 'grevillea' but 
 > only one of them was a Grevillea [G. robusta].  There were a few others 
 > in the family Proteaceae but the rest were nothing to do with 
 > Grevilleas and weren't sorted alphabetically.

Yes, but grevillea robusta is the first result for grevillea.
Grevillea robusta is also the only 'grevillea' in the database.  

 > Then I searched "Apple Mint" and it returned 162 matches for 'apple 
 > mint'.  I got lots of mints, apples, pineapples, Mentha species, 
 > Labiatae (now Lamiaceae) and Rosaceae families, fair enough, but about 
 > half maybe two-thirds the way through the list, Chives (Allium 
 > schoenophrasum) showed up, then lots of clovers (Trifolium spp. family 
 > Leguminosae now Fabaceae), after that were nothing to do with mint, 
 > apples or pineapples, and not sorted alphabetically.

Yes, but the top result for 'apple mint' is in fact apple mint.

 > Is this normal?  I don't know if it matters to most people, especially 
 > non-horties, but I for one do not have enough internet time available 
 > to scroll hundreds of matches.  Or maybe this is returning 
 > compatible/guild species as well?

The results returned are a result of the current search algorithm that
could definitely use some tuning and/or be separated into more
specialized searches.  Currently the search works for both names
(common and botanical) and descriptions of plants (indexing the entire
dataset) so the results tend to be fairly generalized.  Additionally
part of the reason for so many results is that the search strings
themselves are broken into substrings in order to accommodate
incorrect spellings.  For example a search for the string 'grevillea
robusta' actually looks through the database for 'gre', 'vil', 'lea',
'robu', and 'sta'.  As you can imagine this results in many more
matches than the only one possible, but it also allows for people to
spell as poorly as I do and still find what they are looking for.

My question is are the top results the ones you where looking for or
not?  If you have to page through several pages to find what you are
looking for then I think we do have a problem.  Most commercial search
engines return many more results than one would have time to page
through, but a good search engine will [hopefully] have the most
relevant material near the top.  Is receiving more results than you
expected for a query really problematic?

For fun, try doing some searches with deliberate misspelling.
Currently, it may not perform as well as it can, because the number of
substrings a word is sperated into is deliberately low in order to
increase performance in general searches.  If/when we seperate out
search queries into ones that look only at names and ones that look
only at descriptions (or other criteria), we can increase the
granularity significantly and have an effective tool even for people
who spell like I do ;-)

 > Also when/how can we start adding our own species?  Are we going to be 
 > able to upload any photos?

Ack... Well, I fully hoped to have collaborative capabilities at least
in the form of comments by Jan 1st, but it looks as if Feb 1st is the
new realistic target.  Relationships and guild information will
follow, and I hope to have something quite usable by April 1st.

 > Best regards
 > Robyn

Thanks again for the feedback.

-- 
Chad Knepp
python -c 'import base64;print base64.decodestring("cHlnQGdhbGF0ZWEub3Jn")'