15:00:32 <kmlussier> #startmeeting 2016-03-15 - Evergreen focus group discussion on search
15:00:32 <pinesol_green> Meeting started Tue Mar 15 15:00:32 2016 US/Eastern.  The chair is kmlussier. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:32 <pinesol_green> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:32 <pinesol_green> The meeting name has been set to '2016_03_15___evergreen_focus_group_discussion_on_search'
15:00:49 <kmlussier> Huh, just realized it's March 15. Beware the Ides of March!
15:01:00 <kmlussier> #topic Introductions
15:01:09 <kmlussier> Please introduce yourselves, as follows
15:01:16 <kmlussier> #info kmlussier is Kathy Lussier, MassLNC
15:01:24 <sherbertbc> sherbertbc is Sharon Herbert, Sitka/BC Libraries Cooperative
15:01:31 <ethomsen> #info ethomsen is Elizabeth Thomsen, NOBLE
15:01:49 <rhamby> #info rhamby is Rogan Hamby, Equinox Software
15:01:52 <pams> #info pams Pamela Smith howe library, hanover, nh
15:02:09 <csharp> #info csharp is Chris Sharp, GPLS (partially here)
15:02:55 <kmlussier> While we wait for more people to come in, I'm posting a link to some ground rules I wrote up.
15:03:03 <kmlussier> http://wiki.evergreen-ils.org/doku.p​hp?id=scratchpad:search:focus_groups
15:03:09 <yboston> #info yboston - Yamil Suárez - Berklee College of Music
15:03:10 <kmlussier> bah! That's not going to work
15:03:18 <Onijay> #info onijay is Joni Paxton, Community Library (lurking & learning)
15:03:44 <sherbertbc> #info sherbertbc is Sharon Herbert, Sitka/BC Libraries Cooperative
15:03:59 <kmlussier> http://wiki.evergreen-ils.org/doku.php?id=scratchpad:search:focus_groups
15:04:23 <kmlussier> asically, 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.
15:04:32 <kmlussier> Sorry, that should have said basically.
15:04:50 <kmlussier> We really want to focus on end results today.
15:05:05 <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.
15:05:18 <kmlussier> However, I do encourage you to build upon each other's ideas.
15:05:33 <sherbertbc> kmlussier: +1 on ground rules
15:05:34 <dbwells> #info dbwells is Dan Wells, Hekman Library (Calvin College)
15:05:52 <kmlussier> I'm going to get started on the discussion now. If you come in late, feel free to introduce yourselves in the middle of the discussion.
15:06:08 <kmlussier> #topic Strengths of current search
15:06:26 <kmlussier> I want to start by asking people to identify what they like about the current Evergreen search.
15:06:37 <kmlussier> What are its strengths?
15:06:44 <kmlussier> Is there anything that Evergreen search has that makes it unique (in a good way)?
15:07:55 <miker> #info miker == Mike Rylander, Equinox
15:08:44 <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.
15:09:08 <kmlussier> So think of the question in terms of things you don't want to lose.
15:09:32 <miker> separate scope and depth search ranging
15:09:47 <yboston> Like I have said before, I like that it can handle diacritics, and we can filter by format type
15:09:59 <kmlussier> Excellent! Thanks miker!
15:10:17 <sherbertbc> save to list, filter by availability
15:10:20 <miker> configurable visibility for URIs (with an outstanding patch to improve that)
15:10:46 <kmlussier> sherbertbc: Filtering by availability is a good. Thanks! I don't think it's come up in our past conversations.
15:10:50 <miker> sherbertbc: and, more broadly, filter by status (which is how #available works under the covers)
15:10:53 <kmlussier> miker: Are you speaking of Located URI's?
15:11:01 <miker> kmlussier: I am :)
15:11:05 <csharp> our users (staff and patron) appear to want the ability to have a non-keyword search that just sorts by format, location, and date (e.g. "give me the newest DVDs for my library")
15:11:34 <csharp> oops - may not be the time for that
15:11:42 <kmlussier> csharp: We'll get there in a bit.
15:11:43 <miker> csharp: that's expressible today, but not with UI widgets. fwiw
15:12:10 <kmlussier> Any other key features we need to ensure that we preserve in our search?
15:12:45 <kmlussier> I know there are probably a lot of things we like, but it's not always easy to think of them because we just get used to having certain things.
15:13:56 <pams> we like the call # numeric/browse shelf option
15:14:21 <kmlussier> pams: Great! Thanks for mentioning it.
15:14:31 <ethomsen> We like limit to copy location in Adv. Search
15:15:26 <Onijay> We like to do ISBN searches
15:15:46 <kmlussier> Yes, in addition to the copy location limiting, I would say there are a lot of power limiters from advanced search.
15:15:55 <yboston> I like using search by barcode on occasion
15:16:06 <yboston> also call number search
15:16:26 <kmlussier> Onijay: When you say ISBN searches, is that from the Numeric search or is it that you like the ability to do it from a keyword search?
15:16:47 <Onijay> both
15:16:59 <kmlussier> Onijay: Gotcha. Thanks!
15:17:20 <kmlussier> I'm going to move on to the next topic, but if anyone wants to add other strengths in the course of the conversation, please feel free to do so.
15:17:25 <kmlussier> #topic Areas for improvement
15:17:38 <kmlussier> Where are the areas where you would like to see improvement?
15:17:46 <ethomsen> Speed and relevance
15:18:32 <pams> spelling assist like Google, did you mean?
15:19:04 <kmlussier> ethomsen: Thanks! I don't think there is much to say about speed, except make it faster, but could you talk a little more about where you would like to see improvements in relevance?
15:19:15 <kmlussier> Basically, where do you see relevance coming up short?
15:19:32 <kmlussier> pams: Yes, Did you Mean would be helpful.
15:19:51 <kmlussier> Are there other areas where we think the catalog could provide more assistance to the users in helping them find what they are looking for?
15:19:59 <pams> that's probably the biggest problem our patrons have with our catalog
15:20:00 <ethomsen> Maybe it's just our configuration, but we find very short, older records consistently coming to the top of the results
15:20:31 <kmlussier> ethomsen: Thanks!
15:21:27 <kmlussier> pams: The biggest problem being that the search terms are entered incorrectly, and they need help getting to the right thing?
15:21:37 <ethomsen> Our staff finds it difficult to understand (and therefore explain) why certain records are coming up on top -- can't see what's bumping certain records up so high
15:21:52 * miker has thoughts on that ... but they're implementation-related and will hold off (just sticking a pin in it for later)
15:22:17 <pams> kmlussier: right, a typo or incorrect spelling returns 0 results
15:22:32 * kmlussier likes it when miker has thoughts, but thanks for holding off on details. :)
15:23:03 <miker> (likewise with misspellings... for later :) )
15:23:11 <ethomsen> Or alternate spelling -- we want to maintain local synonym list with matched words and phrases
15:23:11 <ethomsen> color=colour, rain forest=rainforest, 1920s=nineteen twenties, etc.
15:23:11 <ethomsen> (Would like to show users which words have expanded their search and let them opt out of any word)
15:24:16 <kmlussier> ethomsen: That's interesting. You mentioned not being able to see what's bumping certain records. It sounds like you're saying that it's not only that the returned records should be more relevant, but that there's a value to seeing why records are returned in a certain way?
15:24:22 <kmlussier> Am I understanding that correctly?
15:25:07 <ethomsen> At least for staff in troubleshooting mode (or mood)
15:25:59 <dbwells> jobs
15:26:07 <dbwells> sorry :)
15:26:11 <kmlussier> ethomsen: If they were happy with the results, do you think seeing why they're returned in a certain way would be as important?
15:27:00 <kmlussier> OK, so I'm seeing speed, relevance improvements, Did You Mean, a synonym list/thesaurus.
15:27:06 <pams> apostrophes also pose a problem
15:27:09 <ethomsen> It's not really library staff, it's our staff (consortium) trying to understand and to change configs or fix records
15:27:18 <kmlussier> ethomsen: OK, gotcha. Thanks!
15:27:21 <sherbertbc> filtering by format on the search results page could be more obvious, now buried in faceting & labelled as 'genre" not intuitive
15:27:47 <ethomsen> Sometimes overlaying records with better ones demotes them in search results!
15:27:49 <kmlussier> sherbertbc: Are you saying you would like a format facet?
15:28:22 <miker> sherbertbc: or, do you want a "controlled" set for filtering on genre/format?
15:28:24 <kmlussier> ethomsen: I'm going to expand upon that a bit and say that record quality should matter in relevance?
15:28:31 <ethomsen> We want both physical format and genre facets
15:28:53 <kmlussier> sherbertbc: Oh, I think I misunderstood what you were saying. I understand now.
15:29:03 <sherbertbc> more than just re-labelling genre to facet. I think bringing "Filter results by" to the top of sidebar would help
15:29:17 <sherbertbc> miker: not sure if ^^ is answering your question though
15:30:00 <ethomsen> kmlussier: Not sure if that true, just think short records getting too much attention!  To me, if a short and long record both have the word Puffins in the title and subject heading, they should rank the same.
15:30:04 <miker> sherbertbc: to put it in "evergreen features" terms, akin to, say, the language or format filter on the advanced search page?
15:30:18 <kmlussier> sherbertbc: It sounds like you're saying you would like to see usability improvements to limiting on the search results screen.
15:30:27 <sherbertbc> kmlussier: +1
15:30:50 <sherbertbc> yes, the functionality is there, it just isn't very evident to users re: terminology and placement
15:30:51 <kmlussier> I have more to say on that, but will hold back for now. :)
15:31:07 <kmlussier> ethomsen: Thanks
15:31:15 <ethomsen> We definitely want more facets so a user can filter results rather than re-executing search
15:31:52 <kmlussier> I want to go back to something ethomsen said earlier: "Would like to show users which words have expanded their search and let them opt out of any word"
15:32:16 <kmlussier> In this case, it sounds like you're talking about when the system searches for stemmed terms - is that right?
15:32:42 <kmlussier> pams: Also, I would like to hear more from you on where you see problems with apostrophes.
15:33:46 <pams> we have a lot of problems where an author or title isn't searchable without the apostrophe
15:33:46 <ethomsen> Not necessarily stemming, but any future situation where the system is using synonym lists or other tools to be able to change search
15:34:49 <kmlussier> pams: OK, I'll not it down. I know we have some issues with apostrophes too, but I also have some info I need to add to documentation that may be able to help you currently.
15:35:08 <miker> pams: it sounds like more indexing normalizers would be helpful to you ("remove apos", etc)
15:35:21 <miker> pams: does that sound correct?
15:36:19 <kmlussier> The possible feature ethomsen mentioned reminded me of comments made in a previous chat to improve fuzziness. I think the crux of the discussion was that we wanted Evergreen to be smarter with its fuzziness.
15:36:41 <kmlussier> How do you all find the current use of stemming in Evergreen search?
15:37:03 <ethomsen> We have it turned off.
15:37:07 <jeff> death to stemming!
15:37:16 <pams> kmlussier: yes that is correct
15:37:38 <yboston> BTW, I sometimes get errors searching terms that include non-alphanumeric characters like "!@#$%^&
15:37:44 <jeff> er, what i meant to say was: i'm ambivalent about stemming!
15:37:51 <miker> jeff: I read that as "knobs for adjusting stemming!" ;)
15:38:00 <kmlussier> jeff: Yes, that sounded ambivalent.
15:38:32 <kmlussier> I hear smarter fuzziness, whether it's stemming or another way to search word variations. :)
15:39:03 <kmlussier> Any other areas of improvements we want to bring up? Or features we see in other search systems that we would love to see in Evergreen?
15:39:05 <ethomsen> We like controlled fuzziness -- lists we maintain and can adjust
15:39:38 <kmlussier> Or maybe new and innovative things that nobody else has done yet?
15:39:54 <miker> ethomsen: I'll paste a link of interest after the meeting if you can stick around for a minute
15:40:02 <ethomsen> Thanks!
15:40:48 <jeff> miker: I'm still trying to come up with a good analogue or metaphor for tuning search knbos. the best i've come up with is this: https://www.amazon.com/dp/B004NCEL4M/
15:40:48 <miker> kmlussier: I'd like to see search augmented with external, curated disambiguation information
15:41:19 <miker> jeff: heh... I was expecting a sound board with hundreds of knobs ;)
15:41:49 <ethomsen> miker: authority records, linked data...?
15:42:05 <jeff> "you searched for steven king, who has a few children's books. you might have actually wanted to search for stephen king, a prolific author well known for ..."
15:42:14 <miker> ethomsen: and other sources, like wikidata related/broader/narrower lists
15:42:19 <kmlussier> miker: I like the sounds of this. Can you tell us a bit about what the end user would see?
15:42:35 <kmlussier> Or maybe jeff helped to answer my question.
15:42:55 <miker> that's a use case, yep
15:43:19 <miker> more that "did you mean" ... "other things that might be related"
15:44:05 <ethomsen> miker: me, too.  So you search Fenway Park and it knows your search is probably related to Boston, Boston Red Sox, baseball, baseball parks...
15:44:27 <miker> ethomsen: right, like that
15:45:14 <kmlussier> OK, I'm going to wrap up this topic now to move on to a bit more detail on relevance. But I think this one ended on a high note.
15:45:17 <ethomsen> I love Wikipedia's disambiguation pages, so much better for public than authority displays! https://en.wikipedia.org/wiki/Mars_%28disambiguation%29
15:45:52 <kmlussier> #topic Defining relevance
15:45:55 <miker> kmlussier: another enhancement, improve metarecord grouping (in many ways, manual and otherwise), and make them first-class, useful things
15:46:08 <kmlussier> I really hated interrupting the disambiguation discussion for this.
15:46:09 <ethomsen> miker: YES!
15:46:37 <kmlussier> miker: OK, we'll digress for a second. Tell me about the first-class, useful things we'll be turning the into.
15:48:07 <miker> :) sorry ... so, with better grouping mechanisms (more tunable automated grouping, possible to have multiple MRs per record, cataloger-adjustable MR inclusion) I think some classes of searches would benefit from having MRs be the main result list
15:48:30 <miker> and, that result list could (should, IMO) surface information about the constituent records right up front
15:48:56 <miker> have the "work" level info, then "manifestation" info below that for each constituent
15:49:31 <miker> it would be a larger result page, but get the user to the thing they wanted faster ...
15:49:42 <kmlussier> And is this somewhat related to the code4lib talk you saw? Because I need a picture to help me with see this.
15:49:45 <miker> not applicable to all search needs, certainly
15:50:03 <miker> kmlussier: it is.  I haven't had a chance to find the timecode for the start of the talk yet, though
15:50:10 <kmlussier> OK, thanks
15:50:12 <miker> but, the videos for thursday are up now
15:50:13 <ethomsen> How about being able to be looking at a single record and being able to jump from there to other formats/editions, like in Amazon?
15:50:20 <kmlussier> Now I want to move on to relevance.
15:50:20 <miker> ethomsen: YES!
15:50:24 <miker> :)
15:50:42 <miker> kmlussier: sorry for derailing
15:50:53 <kmlussier> No, that's okay, That's what this discussion is for. :)
15:51:39 <kmlussier> I often hear people say that relevance needs to be improved. But relevance for one person may be not-so-relevant for another. I think we really need to define what relevance means for us.
15:52:00 <kmlussier> My question is, what factors should be used to identify a record as being highly relevant in a search?
15:52:29 <kmlussier> We heard a bit of this above. We heard that we don't want them to be older and shorter records. And I think ethomsen mentioned words appearing in specific fields.
15:52:38 <kmlussier> What else should contribute to relevance?
15:53:18 <ethomsen> Obvious aboutness -- searched terms appearing near each other in author, subject, title fields coming up first
15:53:35 <ethomsen> We have great hopes for activity metric, too.
15:54:17 <kmlussier> I'm hearing three things here: Words in specific fields, word proximity, and popularity.
15:54:40 <kmlussier> Other than the shorter / older records, are there other places where we consistently see Evergreen not returning relevance records?
15:55:53 <ethomsen> I would have added more recent titles should tend to appear higher, but I think activity metric will tend to have that effect anyway
15:56:35 <kmlussier> Well, I think recency should be identified anyway. Just to make sure it's not overlooked.
15:57:09 <kmlussier> I'm going to leave this question out for about 30 seconds longer before I move on to the final question.
15:57:50 <kmlussier> #topic Top-priority improvements
15:58:04 <kmlussier> If you had just two improvements you could make to Evergreen search, what would they be?
15:58:07 <ethomsen> We love summary notes and contents notes, but they have so many words in them they often overwhelm the search results with titles of low relevance.  I'd like to be able to be able to turn "Search within contents and summary notes?" on or off.
15:59:21 <kmlussier> ethomsen: Thanks for adding that idea. That particular idea hasn't been raised yet, but it correlates with something I was thinking recently.
15:59:22 <ethomsen> Are we allowed to say speed and relevance?
15:59:32 <kmlussier> ethomsen: Yes, if those are the two highest priorities.
15:59:49 <ethomsen> Speed and relevance
16:00:13 <kmlussier> For the first chat, I only let people pick one. Most people picked speed, so JBoyer suggested I expand to two improvements.
16:00:43 <kmlussier> Anyone else have a top 2? I know there were more people sharing ideas earlier. :)
16:01:45 <pams> I would say relevance & spelling
16:01:51 <kmlussier> Thanks pams!
16:02:35 <kmlussier> We're pass the two minute mark, so I'm going to wrap up. But if anyone else wants to send me there two, just let me know. I'll add it to my notes even if it doesn't make the official logs.
16:02:41 <kmlussier> #endmeeting