Unexpected behavior in Concordance

Page 1 of 2 (22 items) 1 2 Next >
This post has 21 Replies | 1 Follower

Posts 1896
Donnie Hale | Forum Activity | Posted: Mon, Aug 22 2016 5:26 PM

I'm getting to use Concordance for the first time today. This is on L7 with indexing complete.

I'm attaching 3 images in the order in which I was working, with comments. Based on the UI, it looks like the Concordance is supposed to be a faceted style of experience; i.e. continue selecting options (part of speech, case, etc) to narrow what is being included in the results list. However, that is not the behavior I'm seeing.

Here's the first image, showing a baseline concordance results in the ESV of John. I've highlighted key things I was using and looking at.

Here's the second image, where I had clicked the part of speech "Adverb." Note that the number of adverbs matches the count in the first image, as expected; and the count of nouns has changed to 1. To me that implies that there is a single adverb which is functioning as a noun in John.

Here's the third image, in which I've selected "Noun." My expectation was that this would restrict the results to a single lemma, since the count was 1. However, the count of nouns goes back to the count in the first image; and the number of lemmas in the results list increased rather than decreased.

From this I conclude that either there is a bug in the Concordance feature or, more likely, the UI which implies a faceted behavior is not actually faceted. If it is the latter, then I really don't know what to expect when I click something. What did the count of 1 for nouns really represent, and why would I ever consider clicking it if it was just going to increase to its original count.

Can anyone help me out?

Thanks,

Donnie

Posts 27034
Forum MVP
MJ. Smith | Forum Activity | Replied: Tue, Aug 23 2016 12:00 AM

The count on the left is the number of unique roots; the count on the right is the number of occurrences  Does that resolve your problem?

When only adverb is checked you see beside noun the number of roots that are nouns AND adverbs. When you select nouns you get the number of roots that are nouns OR adverbs .... at least that's what I've assumed

Orthodox Bishop Hilarion Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."

Posts 22095
Forum MVP
Graham Criddle | Forum Activity | Replied: Tue, Aug 23 2016 1:45 AM

MJ. Smith:
When only adverb is checked you see beside noun the number of roots that are nouns AND adverbs. When you select nouns you get the number of roots that are nouns OR adverbs .... at least that's what I've assumed

I believe this is a correct statement of what is happening.

Like Donnie I find this surprising behaviour - I would have expect it to AND the terms not OR them.

This would be consistent with the behaviour in, for example, Psalms Explorer.

Posts 13369
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Aug 23 2016 3:29 AM

MJ. Smith:
When you select nouns you get the number of roots that are nouns OR adverbs

Yes, this is correct. As there's never an occasion when you really want words that are both nouns AND adverbs, the OR function is much more useful. Checklists (rather than radio buttons or links) are the visual clue that it's an OR function.

The reason it appears as though there is a word that is both a noun and an adverb is because you have an English Bible selected. Just as with morph searching, you won't get 100% accurate results. If you select a Greek Bible, everything will appear as it should, and selecting adverbs will give a zero total for nouns.

Posts 1896
Donnie Hale | Forum Activity | Replied: Tue, Aug 23 2016 6:01 AM

Mark Barnes:
Checklists (rather than radio buttons or links) are the visual clue that it's an OR function.

That's not really accurate. Go to any web site that uses faceted browsing and it uses checkboxes where items are not completely mutually exclusive. Selecting multiple items continually narrows the result set.

I'll play around with the feature some more, but it is counter-intuitive based on how we've been "trained" in the use of UIs that on the surface look like faceted UIs.

Thanks for the replies,

Donnie

Posts 8084
LogosEmployee

Donnie Hale:

That's not really accurate. Go to any web site that uses faceted browsing and it uses checkboxes where items are not completely mutually exclusive. Selecting multiple items continually narrows the result set.

Items under the same header are ORs (and expand the result set); items under different headers are ANDs (and narrow the result set).

Donnie Hale:
I'll play around with the feature some more, but it is counter-intuitive based on how we've been "trained" in the use of UIs that on the surface look like faceted UIs.

AFAICT, this is exactly how amazon.com's faceted product browsing works (whenever I've used it).

Posts 22095
Forum MVP
Graham Criddle | Forum Activity | Replied: Tue, Aug 23 2016 8:14 AM

Bradley Grainger (Faithlife):
Items under the same header are ORs (and expand the result set); items under different headers are ANDs (and narrow the result set).

But in Psalms Explorer, for example, clicking multiple items under the Themes heading narrow the result set.

Is that different because you can legitimately have overlap?

EDIT: Apologies, the Psalms Explorer doesn't use checkboxes

Posts 13369
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Aug 23 2016 8:19 AM

Donnie Hale:

Mark Barnes:
Checklists (rather than radio buttons or links) are the visual clue that it's an OR function.

That's not really accurate. Go to any web site that uses faceted browsing and it uses checkboxes where items are not completely mutually exclusive. Selecting multiple items continually narrows the result set.

Posts 13369
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Aug 23 2016 8:23 AM

Graham Criddle:
But in Psalms Explorer, for example, clicking multiple items under the Themes heading narrow the result set.

Psalms Explorer doesn't have checkboxes; it ANDs. That's a design choice. It could have had checkboxes and ORd. Who knows what people would have preferred, but as you says, as there could be overlap AND seems to work pretty well there.

Posts 3150
SineNomine | Forum Activity | Replied: Tue, Aug 23 2016 8:30 AM

Bradley Grainger (Faithlife):
AFAICT, this is exactly how amazon.com's faceted product browsing works (whenever I've used it).

This is also how most if not all library websites (academic and public) work when it comes to checkboxes.

Posts 1896
Donnie Hale | Forum Activity | Replied: Tue, Aug 23 2016 10:33 AM

Mark (and Bradley),

The confusion is in the count that is displayed. In your Amazon example above, pick any of those checkboxes, and your results are going to change by that number. If it's an OR, then it will add that amount (e.g. 71 Russell Hobbs). If it's an AND, it will reduce to that amount (e.g. 6 Mini Blenders).

In the screen shot I provided, the number next to "Nouns" was 1, and clicking it added far more than all occurrences of just 1 noun lemma.

Does that make sense?

Donnie

Posts 8084
LogosEmployee
Bradley Grainger (Faithlife) | Forum Activity | Replied: Tue, Aug 23 2016 11:47 AM

Donnie Hale:

In the screen shot I provided, the number next to "Nouns" was 1, and clicking it added far more than all occurrences of just 1 noun lemma.

Gotcha. Yes, I see how this is confusing.

The "Noun (1)" label does two things:

  1. Shows you the current number of rows in the current view that match "Noun", i.e., 1.
  2. Allows you to expand the current view to include Nouns, at which point the count increases to 405 (because there will be 405 rows shown that match nouns).

I can see how it might be more consistent to make all the checkboxes show the number of results that would be shown if they were checked, rather than providing information on what is currently being displayed.

I'll pass this feedback along.

Posts 8084
LogosEmployee
Bradley Grainger (Faithlife) | Forum Activity | Replied: Tue, Aug 23 2016 11:58 AM

If I'm understanding correctly, in this revised model, after checking "Adverb (93)" the count next to Noun would change from 405 to 497 because 497 total lemmas will be displayed as a result of checking the "Noun" checkbox?

Posts 1896
Donnie Hale | Forum Activity | Replied: Tue, Aug 23 2016 12:10 PM

Bradley Grainger (Faithlife):

If I'm understanding correctly, in this revised model, after checking "Adverb (93)" the count next to Noun would change from 405 to 497 because 497 total lemmas will be displayed as a result of checking the "Noun" checkbox?

Yes, that is what I would expect, with one further comment. The number of actual results shown on the right represents the number of words which are "hits" against the lemmas resulting from the filters clicked on the left. That is typically a lot more than the number of lemmas (since I'm in "lemma" mode) that match the filters (e.g. 405+93 vs 3,539 in my 3rd screenshot). So it might be helpful to put that lemma count (maybe in parentheses) or something at the top of the right to cue the correspondence between filter hits vs. individual words that match against those lemmas. Hopefully that makes sense.

Thanks for helping me get my point across as well as for clarifying the OR vs AND behavior.

Donnie

Posts 27034
Forum MVP
MJ. Smith | Forum Activity | Replied: Tue, Aug 23 2016 1:00 PM

Bradley Grainger (Faithlife):
I can see how it might be more consistent to make all the checkboxes show the number of results that would be shown if they were checked, rather than providing information on what is currently being displayed.

Unfortunately, the current behavior is what I prefer ... it gives me the statistics of my currently selected data. I would hate to see it changed unless statistical were available in another form.

Orthodox Bishop Hilarion Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."

Posts 13369
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Aug 23 2016 2:14 PM

Bradley Grainger (Faithlife):
If I'm understanding correctly, in this revised model, after checking "Adverb (93)" the count next to Noun would change from 405 to 497 because 497 total lemmas will be displayed as a result of checking the "Noun" checkbox?

Surely it should say Adverb (93), and Noun (405) — although I agree with Donnie that it would be helpful to have the two totals (distinct lemmas and total lemmas) at the top of the right-hand panel.

Posts 8084
LogosEmployee

Mark Barnes:
Surely it should say Adverb (93), and Noun (405)

Yes, when nothing is checked.

Once Adverb is checked, Noun should change to 497 because there will be 497 total results shown if you check that checkbox. (Speaking hypothetically about this new UI, of course.)

Posts 27034
Forum MVP
MJ. Smith | Forum Activity | Replied: Tue, Aug 23 2016 2:43 PM

Please don't jump to a new UI that would screw my use up royally without giving it time for thought and discussion.

Orthodox Bishop Hilarion Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."

Posts 13369
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Aug 23 2016 2:49 PM

Bradley Grainger (Faithlife):
Once Adverb is checked, Noun should change to 497 because there will be 497 total results shown if you check that checkbox. (Speaking hypothetically about this new UI, of course.)

That makes no sense to me. I can't logically see how the number next to "noun" actually means "the total results shown if I check that box". Logically, I can only see it as meaning "the number of nouns that would be added to the filter if I click this box".

Posts 1896
Donnie Hale | Forum Activity | Replied: Tue, Aug 23 2016 3:07 PM

MJ. Smith:
Please don't jump to a new UI that would screw my use up royally without giving it time for thought and discussion.

My brain reads that as, "MJ will get the UI she wants because she's MJ."

It surprises me that you prefer this given that, if I recall correctly, you've been a pretty vocal advocate of a faceted UI. The current UI triggers the "oh, these are facets" expectations but then goes against them in practice. Do you think you would prefer the current behavior if you had never seen this feature before and were using it for the first time?

I can certainly see an argument for providing both "currently displayed" stats as well as "if you click this stats." How best to do that is the question, I think.

Donnie

Page 1 of 2 (22 items) 1 2 Next > | RSS