15:01:54 <kmlussier> #startmeeting 2016-03-16 - Evergreen focus group discussion on search
15:01:54 <pinesol_green> Meeting started Wed Mar 16 15:01:54 2016 US/Eastern.  The chair is kmlussier. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:54 <pinesol_green> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:54 <pinesol_green> The meeting name has been set to '2016_03_16___evergreen_focus_group_discussion_on_search'
15:02:17 <kmlussier> #topic Introductions
15:02:30 <kmlussier> Please introduce yourselves, as follows
15:02:36 <kmlussier> #info kmlussier is Kathy Lussier, MassLNC
15:02:49 <miker> #info miker == Mike Rylander, Equinox (in and out)
15:02:57 <afterl> #info afterl is Amy Terlaga, Bibliomation
15:03:49 <sarahchilds> #info sarahchilds is Sarah Childs (surprise), Hussey-Mayfield PL, Evergreen Indiana
15:03:49 <kmlussier> While we wait for more people to join us, I'll post a link to our page where we have 1) ground rules for the discussion and 2) logs from our previous chats.
15:04:03 <kmlussier> #link http://wiki.evergreen-ils.org/doku.php?id=scratchpad:search:focus_groups
15:04:22 <rhamby> #info rhamby is Rogan  Hamby, Equinox Software (mostly lurking)
15:04:51 <kmlussier> Basically, this chat is an opportunity to talk about ideas for search. Although it's tempting, especially for technical folks, to get in the nitty gritty of how to make those ideas happen, we don't want to get bogged down in the details today.
15:05:07 <kmlussier> We really want to focus on end results today.
15:05:21 <kmlussier> Also, as is the case with any brainstorming, there are no stupid ideas. Feel free to share whatever ideas you have. And please don't be critical of others ideas.
15:05:29 <kmlussier> However, I do encourage you to build upon each other's ideas.
15:06:01 <kmlussier> I think we have a smaller turnout today than we have on other days, so I hope you all feel talkative! :)
15:06:46 <kmlussier> I'm going to start our first topic, but if anyone comes in late, please feel free to introduce yourselves and join in.
15:07:20 <kmlussier> #topic Strengths of current search
15:07:35 <kmlussier> I want to start by asking people to identify what they like about the current Evergreen search.
15:07:45 <kmlussier> What are its strengths?
15:07:48 <dbs> @info dbs is Dan Scott, Laurentian University
15:07:48 <pinesol_green> dbs: (info <url|feed>) -- Returns information from the given RSS feed, namely the title, URL, description, and last update date, if available.
15:07:52 <dbs> #info dbs is Dan Scott, Laurentian University
15:09:47 <dbs> Keyword searches work as a great default for us for the vast majority of searches
15:09:53 <afterl> Search filters work relatively well
15:10:15 <sarahchilds> Filters make good use of MARC data
15:10:25 <kmlussier> dbs: And when you say they work well, is that because there is good coverage of the record?
15:10:43 <dbs> kmlussier: seems to hit the most important bases, yep
15:10:48 <kmlussier> OK, good.
15:11:15 <kmlussier> And we like the advanced search filters and the fact they make use of the MARC data.
15:11:19 <kmlussier> The reason I ask this question is because, if we make major changes to search, we want to make sure we don't lose the things that are already working well for us.
15:11:32 <kmlussier> So think of the question in terms of things you don't want to lose.
15:11:37 <miker> extensibility of search/filter configuration ... though "yay, lots of knobs!" / "boo, lots of knobs :("
15:12:24 <kmlussier> So there are pros and cons to the amount of configuration we have in search.
15:12:52 <kmlussier> Is there anything Evergreen search has that you think makes it unique?
15:13:45 * miker wonders if he should cheat and pull from yesterday's log :)
15:13:52 <sarahchilds> It seems pretty rare to use fixed fields so extensively for filtering
15:14:24 <kmlussier> miker: No cheating allowed.
15:15:12 <kmlussier> Sorry, I purposely allow for quiet moments because it makes people think harder.
15:15:26 <miker> :)
15:15:36 <kmlussier> But I'll move on to the next topic. If you think of anything else while we're talking, feel free to shout it out.
15:15:38 <afterl> kmlussier:  so looks like we'll be reading yesterday's log on our own ....  homework
15:15:46 <miker> heh
15:16:02 <kmlussier> afterl: I'll be compiling the results at some point if you want the Cliff Notes version.
15:16:19 <kmlussier> #topic Areas for improvement
15:16:22 <afterl> kmlussier:  yes, thanks.  Cliff Notes work
15:16:29 <kmlussier> Where are the areas where you would like to see improvement?
15:17:43 <sarahchilds> I'd like to see search provide suggestions based on authority records
15:18:01 <Bmagic> #info Bmagic = Blake GH, MOBIUS
15:18:06 <kmlussier> sarahchilds: OK, can you expand upon that? Is this in keyword searching?
15:18:19 <kmlussier> Hi Bmagic! Welcome!
15:18:58 <afterl> I'd like a true relevance/popularity blended search.  And yes, that may force kmlussier to explain what's coming down the road, but maybe that's not allowed here (not the time and place)?
15:19:10 <sarahchilds> Maybe. If you do a search for doctors, you could get a Did you mean: Physicians
15:19:39 <kmlussier> sarahchilds: Ah, I see. So it's a "Did you mean?" but leveraging authority data. Gotcha.
15:19:59 <sarahchilds> If you do a search for Samuel Clemens, it would suggest Mark Twain
15:20:15 <dbs> I'd like to see grouping of search results based on external links to, say, OCLC Work Entities
15:20:32 <kmlussier> To follow up on what afterl said, we do have activity metrics coming to search that afterl and I have been testing. But the blending of popularity with relevance is tricky. And we're still working on it / testing it.
15:21:42 <kmlussier> dbs: Can you give me an example of a search that would use that and how the info would be presented to the user?
15:24:18 <dbs> kmlussier: sure. A regular search for "Huckleberry Finn" would give you all of the hits for "Adventures of Huckleberry Finn" based on keywords
15:24:55 <dbs> but using links to OCLC Work Entities, you could group those records even if they were in different languages and the words appeared nowhere in the record
15:25:10 <kmlussier> Ah, nice! Thanks for the explanation dbs!
15:26:14 <kmlussier> Are there features you've seen in other search systems that you would like to see implemented in Evergreen? Things that haven't been mentioned yet.
15:28:04 <dbs> I wish we could offer a single search across not just books but blended with articles, items in our institutional repository, etc, but that's probably out of scope for Evergreen
15:29:09 <kmlussier> dbs: Well, for the purposes of this discussion, I don't think we should say anything is out of scope for the project, because if it's turns out to be something a lot of Evergreen users want to see, then maybe it becomes part of the scope at some point.
15:29:34 <kmlussier> dbs: But I also think a search like that needs to be handled well so that the user doesn't get lost in results.
15:30:12 <JBoyer> dbs: that could be part of a project making Evergreen easier to connect to a discovery system, which would probably touch on search. :)
15:30:33 <kmlussier> Are there other parts of search that just drive your users crazy? Or would you say that, overall, they are fairly happy with the existing search?
15:30:45 <kmlussier> JBoyer: That's a good point.
15:30:50 <kmlussier> Hi terran!
15:31:17 <terran> So sorry I'm late, it's my first day back in the office and I'm pulling my hair out!
15:31:45 <kmlussier> terran: We're just talking about ways we would like to improve search. Do you have any improvements you would like to see?
15:32:11 <afterl> This was talked about awhile back (we were going to do development but it never panned out), and Evergreen doesn't do this now, I don't think, and I will be embarrassed if it can now, but having the search terms highlighted in the results (on the display record) would be nice.
15:32:14 <terran> The main complaint we get is search speed.
15:32:33 <rhamby> overall my experience with users is that they are pretty happy with existing search and keyword search with the one major qualifier to that being that they want records to float the top based on activity (which is in progress)
15:32:50 <rhamby> with that said, many of these other ideas that have come out of these I'm sure would be welcome as well
15:32:52 <kmlussier> terran: Thanks terran!
15:32:54 <sarahchilds> Users hate when they do a search for a particular title and the item they are looking for shows up low in the results
15:33:46 <terran> Most of our users would also like the most popular records to float nearer the top on keyword or title searches.
15:33:55 <kmlussier> sarahchilds: So improvements in relevance. Do you find there are specific problems in relevance that cause those records to appear low in the list?
15:34:54 <terran> And of course, a good "did you mean...?" service to accommodate mis-spellings and related topics
15:35:06 <sarahchilds> One of the issues I get complaints about is that our users expect to be weighed more heavily
15:35:19 <kmlussier> terran / rhamby: On the popularity front, afterl had brought this up earlier, but there is an ongoing project as rhamby noted, but, from early testing, we're also looking at ways to better blend bib relevance with popularity. Which is tricky.
15:35:30 <sarahchilds> expect *title* to be weighed more heavily
15:35:49 <kmlussier> sarahchilds: Thanks! Then field-weighting is important.
15:36:13 <terran> Yes, I strongly agree with weighing title (and primary author) more heavily
15:36:29 <kmlussier> I know there is some ability to configure field-weighting in Evergreen, but I think implementation might not be consistent across the board. As my follow up on these chats, I hope to identify where more documentation may be needed too.
15:36:46 <rhamby> it is and relevance is going to take on different aspects to different audiences.  a public library that heavily circulates new best sellers is going to be well served by new active titles floating the top.  An academic library I would see as very different in how they want to determine relevance.
15:37:29 <sarahchilds> rhamby: agreed
15:38:51 <kmlussier> Things I'm seeing in areas for improvement, then, are "Did you Mean?", leveraging authority data in keyword searches, blending popularity with relevance, grouping of search results based on external links, improved speed, and improved relevance.
15:39:14 <kmlussier> Did I miss anything?
15:40:07 <kmlussier> If not, I'm going to move on to the next topic, which is something we were just discussion.
15:40:15 <kmlussier> #topic Defining relevance
15:40:45 <kmlussier> I often hear that people want to improve relevance, but, as rhamby just mentioned, what is relevant for one person may not be for another.
15:40:55 <terran> <admits to getting completely lost trying to understand field weighting in the current system>
15:41:03 <kmlussier> So I think it's important that we define what we mean by relevance.
15:41:29 <kmlussier> terran: That's a good point. So I wonder if there is a suggested improvement in there for making it easier / more understandable to turn the knobs?
15:41:52 <afterl> I would support that improvement
15:42:20 <kmlussier> What are the factors that you think are important for the system to consider when calculating relevance?
15:42:30 <kmlussier> A few have already been mentioned: field weighting, popularity.
15:42:31 <terran> field weighting: documentation on what the options mean along with a working example or two with detailed explanation would be most helpful
15:42:33 <kmlussier> Are there other factors?
15:43:23 <kmlussier> Or other specific areas where you see relevance falling short?
15:43:23 <terran> relevance: I think it might be useful to look at titles with the most copies being considered more relevant
15:43:44 <sarahchilds> terran: yes!
15:44:16 <kmlussier> terran: Ok, is that because you see that as one way libraries define that one record may be more important than others?
15:45:00 <sarahchilds> lots of copies would be one metric of popularity
15:45:19 <terran> yes - at least for large public libraries and fiction, the more popular an item is, the more copies there will likely be
15:46:06 <kmlussier> Thanks!
15:46:10 <terran> Titles with a large number of holds would also indicate popularity
15:46:16 <kmlussier> Any other factors you think are important for relevance.
15:46:50 <sarahchilds> Also in a consortium, even if everyone has only one copy, if most libraries own it, that indicates popularity as well
15:46:57 <Bmagic> I just caught up with the convo, forgive me, a call came in during this meeting. I have had problems getting the search results I want. Relevance is one reason, and title searches being so limiting is another reason
15:47:05 <terran> sarachilds: true!
15:47:10 <kmlussier> To give a little background on the activity metric, titles with a large number of holds and circs are all types of activity that scores a badge for a record. But I don't want to go to deep into the details here.
15:47:34 <kmlussier> It doesn't cover total # of copies, but you typically find that you get higher circs on those records with more copies.
15:47:36 <Bmagic> When I get 0 search results, I would like it to instead go ahead and do another similar search that will return something
15:47:38 <terran> kmlussier: yes, I'm really interested in that project that you're working on
15:47:57 <kmlussier> Bmagic: Can you tell expand upon what you mean when you say title searches are limiting?
15:48:24 <kmlussier> Bmagic: Also, I would be interested in hearing where you would like to see improvements in relevance.
15:49:17 <Bmagic> The title index isn't the same as keyword by design, but (also by design) it's not encompassing what some of us consider to be the title. I don't know, basically, it needs to "learn" from popular searches
15:49:18 <sarahchilds> Bmagic: I wouldn't want it to *do* the search, but I'd want it to suggest a meaningful alternative
15:49:39 <terran> Bmagic: I like that idea, or to at least suggest options for other searches when there are 0 results.
15:50:08 <kmlussier> Bmagic: Ah, that's interesting. So you're looking for something that can learn from user behavior?
15:50:28 <JBoyer> I realize this would require a ton of work for little visual improvement, but I'll mention it just so it's known: a re-architecting that removes the need for a core limit of records returned would be helpful. (i.e. we have a lib that can't be searched for a common keyword unless the core_limit is close to 80K...)
15:51:15 <JBoyer> "A ton of work" may be an understatement, also. :)
15:51:21 <Bmagic> I think that might be a possible solution. Search phrasing is remembered, and called back on future searches to help the next individual
15:51:29 <sarahchilds> JBoyer: Yes!!! Not returning any results when there are results to be returned is so not cool
15:51:37 <kmlussier> JBoyer: Thanks for that! But I don't want the "ton of work" ideas to go unmentioned, because maybe a ton of work is ultimately needed. Or maybe not.
15:52:06 <terran> JBoyer: I agree.
15:52:29 <kmlussier> OK, we're reaching the eight-minute mark, so I want to move on to my final question.
15:52:43 <kmlussier> #topic Top-priority improvements
15:53:01 <kmlussier> If you could only make two improvements to Evergreen search, what would they be?
15:53:10 <afterl> It's been said before, but SPEED
15:53:13 <kmlussier> Basically, what do you consider to be the most important issues?
15:53:13 <terran> 1) Speed
15:53:28 <terran> 2) did you mean?
15:53:36 <sarahchilds> Hard to argue with speed
15:53:57 <kmlussier> This is why we have 2 instead of 1. The 1st time around it was all speed.
15:54:12 <Bmagic> Speed, 0 or few results automatically expanded to popular related searches
15:54:13 <kmlussier> JBoyer++ # Suggesting that we let people pick two.
15:54:23 <afterl> And I really want to see that popularity/relevance blended search results
15:55:42 <kmlussier> sarahchilds: Do you have a #2?
15:55:44 <terran> Going back to earlier suggestions (since I was late) - I would also like a facet on the search results for audience type (is this possible in the current system already? I haven't delved into how facets work yet)
15:55:59 <sarahchilds> It's too hard to choose!  I love them all :-)
15:56:02 <kmlussier> terran: If it is possible, I don't know how to do it.
15:56:34 <kmlussier> terran: Thank you for adding that suggestion!
15:56:55 <terran> Ooh, also a facet for fiction vs nonfiction
15:57:15 <kmlussier> Ok, I might wrap up a little early. But is there anything else you had hoped we would cover today that we didn't?
15:57:29 <kmlussier> terran: Yes, we also had one from a previous group on publicate date ranges.
15:57:35 <kmlussier> publication, even
15:58:34 <terran> One for language might be helpful for a lot of libraries, too
15:58:45 <kmlussier> OK, I'm going to wrap up the meeting, then, but let me know if anything comes to mind later on.
15:58:51 <kmlussier> Thanks everyone for sharing your ideas today!
15:58:55 <kmlussier> #endmeeting