locked New test/mobile version


 

Hi All,

As you know, I've been working on a new version of the site that's better suited for mobile browsers. Along with those changes, additional changes and fixes have been made that affect viewing the site on desktop as well. While not perfect (yet!), the new version is much better than the existing version of the site. Also, having two version is confusing for everybody, so I want to switch over to the new version as soon as possible, which I hope to do on Friday. I'd encourage you to play with the new version before then, on either desktop or mobile, and let me know if you have any feedback.

To switch between versions, go to https://groups.io/test

Thanks,
Mark


 

So test version = mobile version right now? It doesn't detect if I'm on a mobile device and switch to another version? Thanks.

Sent from my iPhone

On Jul 13, 2016, at 9:53 AM, Mark Fletcher <markf@corp.groups.io> wrote:

Hi All,

As you know, I've been working on a new version of the site that's better suited for mobile browsers. Along with those changes, additional changes and fixes have been made that affect viewing the site on desktop as well. While not perfect (yet!), the new version is much better than the existing version of the site. Also, having two version is confusing for everybody, so I want to switch over to the new version as soon as possible, which I hope to do on Friday. I'd encourage you to play with the new version before then, on either desktop or mobile, and let me know if you have any feedback.

To switch between versions, go to https://groups.io/test

Thanks,
Mark

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

p.s. If all I care about it testing the new search function, can I do that from test version on a laptop? That's all I really need to know. Thanks. :-)

On Wed, Jul 13, 2016 at 10:12 AM, J_catlady <j.olivia.catlady@...> wrote:
So test version = mobile version right now? It doesn't detect if I'm on a mobile device and switch to another version? Thanks.

Sent from my iPhone

On Jul 13, 2016, at 9:53 AM, Mark Fletcher <markf@corp.groups.io> wrote:

Hi All,

As you know, I've been working on a new version of the site that's better suited for mobile browsers. Along with those changes, additional changes and fixes have been made that affect viewing the site on desktop as well. While not perfect (yet!), the new version is much better than the existing version of the site. Also, having two version is confusing for everybody, so I want to switch over to the new version as soon as possible, which I hope to do on Friday. I'd encourage you to play with the new version before then, on either desktop or mobile, and let me know if you have any feedback.

To switch between versions, go to https://groups.io/test

Thanks,
Mark

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.



--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

On Wed, Jul 13, 2016 at 10:16 AM, J_catlady <j.olivia.catlady@...> wrote:
p.s. If all I care about it testing the new search function, can I do that from test version on a laptop? That's all I really need to know. Thanks. :-)

Yep. You want to use the test version.

Thanks,
Mark 


 

Just doing a cursory check so far. Trying to find something to complain about and having a *really hard time*:-) But here's what I would say so far:

1. The search still returns approximate results. For example, I did a search on the word "disabled" and also got all messages containing the word "disability." I was specifically looking for that one particular word. Even if I put it in quotation marks, I get back all the other stuff. I understand my issue is not specifically related to the new version of search, but I am taking the opportunity to bump it up because I have found it very annoying. WHY can't I get back all messages, and only all messages, containing the exact term I am looking for?

2. After collapsing, threads with only one message get no number after their titles. I think they should get a "1" to make it very clear. I realize that this, too, is not related to the new version of search (or to search at all), but rather to the way threads are displayed in general. But I find it disconcerting in the context of search (and not otherwise) not to specifically have a number displayed for every thread, even if that number is "1."

I am still trying hard to find other things to complain about. Nice job.;) 

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

Here's a little thing to complain about: entering only part of a word brings up no results. For example, if there are messages with the term "supercalafragilistic," that word works fine as a search term. But "supercala" brings up nothing. Etc. I tried this for a bunch of words - "messa" instead of "message," "wate" instead of "water," etc.

I think this is wrong (and my guess is that it's a simple implementation bug).

If intentional policy and not a simple glitch, I think requiring whole words in order to register a match is especially contradictory or inconsistent, in light of the policy of returning highly approximate (rather than just exact) matches in general.

I think both should be eliminated/fixed.

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

J,

On Wed, Jul 13, 2016 at 2:32 PM, J_catlady <j.olivia.catlady@...> wrote:

Here's a little thing to complain about: entering only part of a word brings up no results. For example, if there are messages with the term "supercalafragilistic," that word works fine as a search term. But "supercala" brings up nothing. Etc. I tried this for a bunch of words - "messa" instead of "message," "wate" instead of "water," etc.

Shal brought this up as an issue with activity search as well. It's not a bug in that it's doing what it was programmed to do. Doing partial matches is a tricky thing. Does Yahoo search do true partial matching of arbitrary length queries? I'd be shocked if they do. It's a question of storage space and processing required. 

That said, it definitely needs to happen for at least some things. Unfortunately, the software I use for search, ElasticSearch, has somewhat impenetrable documentation and despite a few hours playing with it, I haven't found the right incantation to get partial searches working yet. I'll most likely have to hire them to help me figure things out (which is their business model.... hmm, now I think I understand the bad documentation.....). All this is to say that: it's a big project, it's important, but it won't happen right away.
 
In search results, thread matches now always display message count badges, even for threads of one message. 

Thanks,
Mark


 

Mark,

On Wed, Jul 13, 2016 at 03:54 pm, Mark Fletcher wrote:

Does Yahoo search do true partial matching of arbitrary length queries?

I don't know about "arbitrary length," but I just tried it one of the few yahoo groups I'm still in, for various words with between about 5 and 15 letters, and it did produce the matches.

What about the approximate matches? That's my main gripe. In combination with no partial matches, what we're getting is (a) no matches for strings (parts of words) that ARE actually contained in messages and on top of that, (b) matches for strings that ARE NOT contained in the search term! It's a double whammy! Is there at least the possibility of making search terms in quotation marks return only exact matches? Isn't that standard, or am I just really behind the times? Is that also a problem with the search software? Yes, maybe you either need to hire these guys or (better yet?) use different software...

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

p.s. It's hard to understand what the problem with partial matches could be. Does there need to be a space after the search term in the message??? Why does "meeting" match in a message that says "The meeting is at 2:00" but "meeti" not match?
--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

On Wed, Jul 13, 2016 at 4:11 PM, J_catlady <j.olivia.catlady@...> wrote:
p.s. It's hard to understand what the problem with partial matches could be. Does there need to be a space after the search term in the message??? Why does "meeting" match in a message that says "The meeting is at 2:00" but "meeti" not match?

I am far from an expert in search, but the partial match thing has to do with something called stemming. The software looks to stem words in the query. That means it removes things like 'ing' or 'es'.

Mark 


 

Ok. I'm not seeing it but I guess I have to take this one in faith. What about 'r' in 'water' and 'fragilicious' in 'supercalacrsgilicious'? It matches 'water' but not 'wate' and 'supercala' but not 'supercalacragilicious'. The only thing I see in common is the spaces after the actual words but not after the partial strings.

Sent from my iPhone

On Jul 14, 2016, at 4:42 PM, Mark Fletcher <markf@corp.groups.io> wrote:

On Wed, Jul 13, 2016 at 4:11 PM, J_catlady <j.olivia.catlady@...> wrote:
p.s. It's hard to understand what the problem with partial matches could be. Does there need to be a space after the search term in the message??? Why does "meeting" match in a message that says "The meeting is at 2:00" but "meeti" not match?

I am far from an expert in search, but the partial match thing has to do with something called stemming. The software looks to stem words in the query. That means it removes things like 'ing' or 'es'.

Mark 

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

Oops I reversed those. It matches supercalafragilistic but not supercala. Ignore iPhone typos;)

Sent from my iPhone

On Jul 14, 2016, at 5:06 PM, J_catlady <j.olivia.catlady@...> wrote:

Ok. I'm not seeing it but I guess I have to take this one in faith. What about 'r' in 'water' and 'fragilicious' in 'supercalacrsgilicious'? It matches 'water' but not 'wate' and 'supercala' but not 'supercalacragilicious'. The only thing I see in common is the spaces after the actual words but not after the partial strings.

Sent from my iPhone

On Jul 14, 2016, at 4:42 PM, Mark Fletcher <markf@corp.groups.io> wrote:

On Wed, Jul 13, 2016 at 4:11 PM, J_catlady <j.olivia.catlady@...> wrote:
p.s. It's hard to understand what the problem with partial matches could be. Does there need to be a space after the search term in the message??? Why does "meeting" match in a message that says "The meeting is at 2:00" but "meeti" not match?

I am far from an expert in search, but the partial match thing has to do with something called stemming. The software looks to stem words in the query. That means it removes things like 'ing' or 'es'.

Mark 

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

Mark,

On Thu, Jul 14, 2016 at 04:42 pm, Mark Fletcher wrote:

it removes things like 'ing' or 'es'

Thinking about this more, I think this also accounts for the approximate matches (which I hate!) I vote for new search software. :-) 
--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

Wikipedia has this to say about stemming;

"There are two error measurements in stemming algorithms, overstemming and understemming. Overstemming is an error where two separate inflected words are stemmed to the same root, but should not have been—a false positive. Understemming is an error where two separate inflected words should be stemmed to the same root, but are not—a false negative. Stemming algorithms attempt to minimize each type of error, although reducing one type can lead to increasing the other."

These are precisely (a) and (b) from my complaint before: searching on "disabled" also yields "disability" etc. (false positive), and searching on "wate" produces nothing even if "water" is in the text (false negative). Arguments justifying either one would negate arguments justifying the other - e.g., "we include as much as possible" negates "we don't include partial matches", and vice versa. That jives with the statement above that  "reducing one type can lead to increasing the other."

I wonder what kind of "minimizing each type of error" was done or attempted by the group who developed the software used by Groups.io (I've already forgotten the name of it). They need to do better IMHO!

--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


 

J,

p.s. It's hard to understand what the problem with partial matches
could be.
Search in a system like this (built on a database) is generally based on indexes, not a direct scan of the content for each search (that would be way to slow for a large-scale system).

Why does "meeting" match in a message that says "The meeting is at 2:00"
but "meeti" not match?
So the partial match issue boils down to what's in the index. In your example having all the possible partial matches ("eeting", "meetin", "eting", "meeti", ...) would dramatically expand the size of the index and slows down the searching.

There may be a clever way to solve the problem, possibly by having a meta-index of the partials contained in indexed words, but this is really beyond my depth where database design and use are concerned. Possibly an answer lies in ElasticSearch's impenetrable documentation, but I don't envy Mark's task in trying to prise it out.


Shal
https://groups.io/g/GroupManagersForum


 

J,

What about the approximate matches?
This goes with the indexed search mechanism. One way to compensate for the lack of partial matches in the index is to index word stems, and use language-based rules (dictionaries) for transforming the search query into a word stem.

For example, I did a search on the word "disabled" and also got all
messages containing the word "disability."
You would likely have seen results for "disables", "disabling" and any other word with disable as a stem.

Even if I put it in quotation marks, I get back all the other stuff.
Here I agree with you strongly. Even if word stemming is used to improve general results, there ought to be a way for the user to specify "exact matches only".

It is looking to me like quotation marks are now doing one expected thing: when searching for a phrase their use appears to return only results with those words in sequence (I tried "quotation marks" in beta@).

But they don't do the other expected thing of returning only exact word matches (rather than stemmed matches).


Shal
https://groups.io/g/GroupManagersForum


 

Yep.

Sent from my iPhone

On Jul 16, 2016, at 11:20 AM, Shal Farley <shals2nd@gmail.com> wrote:

J,

What about the approximate matches?
This goes with the indexed search mechanism. One way to compensate for the lack of partial matches in the index is to index word stems, and use language-based rules (dictionaries) for transforming the search query into a word stem.

For example, I did a search on the word "disabled" and also got all
messages containing the word "disability."
You would likely have seen results for "disables", "disabling" and any other word with disable as a stem.

Even if I put it in quotation marks, I get back all the other stuff.
Here I agree with you strongly. Even if word stemming is used to improve general results, there ought to be a way for the user to specify "exact matches only".

It is looking to me like quotation marks are now doing one expected thing: when searching for a phrase their use appears to return only results with those words in sequence (I tried "quotation marks" in beta@).

But they don't do the other expected thing of returning only exact word matches (rather than stemmed matches).


Shal
https://groups.io/g/GroupManagersForum



--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.


Nightowl >8#
 

Shal wrote:>>There may be a clever way to solve the problem, possibly by having a meta-index of the partials contained in indexed words, but this is really beyond my depth where database design and use are concerned. Possibly an answer lies in ElasticSearch's impenetrable documentation, but I don't envy Mark's task in trying to prise it out.<<

Shal, what about a criteria that does an "Exact Match" search? That's what sites like the Post Dispatch uses if you don't want parts of words or 'ings or 's.

Brenda


Nightowl >8#
 

Hehee... Shal, I just wrote about trying for "Exact Matches" right before reading this post. ;)

Guess we're both on the same track here. :)

Brenda


 

Brenda, We're on the same page here. I've been bemoaning this problem for months. I think I will continue to post concrete examples showing specifically why this is annoying and truly problematic, because I'm not sure the seriousness of the issue has gotten through yet. It is a huge detriment to the product IMO.

Sent from my iPhone

On Jul 16, 2016, at 11:46 AM, Nightowl <featheredleader@att.net> wrote:

Hehee... Shal, I just wrote about trying for "Exact Matches" right before reading this post. ;)

Guess we're both on the same track here. :)

Brenda



--
J

Messages are the sole opinion of the author. 

It's dumb to buy smart water.