Clause Search - the results are rather baffling!

Richard Lyall

Clause searches return the following:

subject:Jesus verb:to weep returns John 11:35 - "Jesus wept"

subject:Jesus verb:to weep, to cry returns Luke 19:41-44 but not John 11:35

Example 2:

subject:Jesus verb:to rebuke, to reprove, to warn returns 0 results but

subject:Jesus verb:to rebuke, to reprove returns 13

subject:Jesus verb:to rebuke returns 0

I only installed L5 earlier this evening, but this is not at all intuitive!

Is there a more indepth guide to Clause Searches anywhere that might explain this? I thought the comma was operating as an OR operator, which should add results not remove them.

Thanks
Richard

Find more posts tagged with

Comments

fgh

Clause searches return the following:

subject:Jesus verb:to weep returns John 11:35 - "Jesus wept"

subject:Jesus verb:to weep, to cry returns Luke 19:41-44 but not John 11:35

From another thread:

even though the English labels for verbs overlap ("to say" vs "to say, to tell"), they're not searching the same thing: the underlying Greek verb is different. So you should expect to get different results.

Dave Hooton

subject:Jesus verb:to weep returns John 11:35 - "Jesus wept"

subject:Jesus verb:to weep, to cry returns Luke 19:41-44 but not John 11:35

Different lemmas, but I do agree that I would not have expected that differentiation.

subject:Jesus verb:to rebuke, to reprove, to warn returns 0 results but

subject:Jesus verb:to rebuke, to reprove returns 13

subject:Jesus verb:to rebuke returns 0

For the verb term alone (without subject I get results for the first two.

I don't see verb:to rebuke as a valid value! Notice the red border around the Find box if you try it.

Rick Brannan (Logos)

https://community.logos.com/discussion/comment/427010#Comment_427010

even though the English labels for verbs overlap ("to say" vs "to say, to tell"), they're not searching the same thing: the underlying Greek verb is different. So you should expect to get different results.

The values searched here are, as I understand it, the "lexical value" fields from the Lexham Greek NT Interlinear.

Those values are context sensitive; when the editor determined more than one sense might apply, he listed them all. It is not a simple concordant gloss.

It seems the expected behavior for clause search is more of a 1:1 correspondence between the lemma and the gloss. I believe (Sean can clarify as he sees fit) that we're discussing ways to do this internally.

I think one thing to realize through all of this is that clause search is really cool, but nothing like it has really ever been done before. We're learning, just like you, how to fine-tune things. Feedback like this helps in that process, so it is greatly appreciated.

Dave Hooton

https://community.logos.com/discussion/comment/427026#Comment_427026

The values searched here are, as I understand it, the "lexical value" fields from the Lexham Greek NT Interlinear.

Certainly appear to be. The Literal values are: Lk 19:41 he wept, Jn 11:35 wept

It seems the expected behavior for clause search is more of a 1:1 correspondence between the lemma and the gloss.

Not from observations in this thread. I did not expect the above two references to be differentiated. The Sense Lexicon doesn't help at all as weep produces "weeping person" whilst cry offers a number of alternatives; none of which have the lemmas in the above passages. But the fact that the lexical value offers "to weep" for both suggests they should not be differentiated.

MJ. Smith

https://community.logos.com/discussion/comment/427010#Comment_427010

From another thread:

even though the English labels for verbs overlap ("to say" vs "to say, to tell"), they're not searching the same thing: the underlying Greek verb is different. So you should expect to get different results.

Unfortunately the example did not support Sean's statement. The same lemma appeared in both subject:Moses verb:to say and subject:Moses verb:to say, to tell. The latter having fewer and different entries than the former. There may be a clear explanation in the semantics - but I haven't seen it yet.

MJ. Smith

https://community.logos.com/discussion/comment/427026#Comment_427026

It seems the expected behavior for clause search is more of a 1:1 correspondence between the lemma and the gloss. I believe (Sean can clarify as he sees fit) that we're discussing ways to do this internally.

This certainly is not true for myself - what I do expect is to be able to understand the results. I'm thinking that the problem may be that the same gloss is used for different senses. I would expect a many-to-many relationship between lemma and sense and a many-to-many relationship between sense and gloss. However, I would expect lemmas, senses, and glosses to all be unique i.e. no duplicates.

MJ. Smith

https://community.logos.com/discussion/comment/427026#Comment_427026

the "lexical value" fields from the Lexham Greek NT Interlinear.

Okay, I've identified one major error in my thought.

I am offered verb:to say --- it works as i expect
I am offered verb:to say, to tell --- I am expecting to tell OR to say 'cause that is what a comma means in a search argument NOPE Logos expected me to see "to say, to tell" is a single argument which contrasts with "to say"

"to say" is one verb sense; "to say, to tell" is another verb sense. The first has two lemmas that fall within it. The second has one lemma that falls within it - a different sense of that lemma falls within the first sense.

Okay, Logos, 'fess up - who didn't see that giving a comma a different meaning in the clause search than in other searches was a dumb idea? Sorry I hadn't sorted this out earlier.

DominicM

https://community.logos.com/discussion/comment/427069#Comment_427069

Okay, Logos, 'fess up - who didn't see that giving a comma a different meaning in the clause search than in other searches was a dumb idea?

Subtle.. real subtle.. but agree its "rather" confusing from what I have been able to follow

Rick Ausdahl

See the thread titled "BUG with Clause Search?" at http://community.logos.com/forums/p/59899/427032.aspx for additional examples of confusing/unexpected results and a reply from Logos about them.

In fairness to Logos and perhaps to other L5 users who find the clause search to work as expected, perhaps I should say, examples of results that are confusing/unexpected to me.

fgh

https://community.logos.com/discussion/comment/427064#Comment_427064

Unfortunately the example did not support Sean's statement.

Oops. Guess I shouldn't say anything when I don't even have L5 yet.[:)]

On the other hand: if Sean Boisen can't get it right, I refuse to [:$] because I didn't![:D]

Okay, I've identified one major error in my thought.

I am offered verb:to say --- it works as i expect

I am offered verb:to say, to tell --- I am expecting to tell OR to say 'cause that is what a comma means in a search argument NOPE Logos expected me to see "to say, to tell" is a single argument which contrasts with "to say"

"to say" is one verb sense; "to say, to tell" is another verb sense. The first has two lemmas that fall within it. The second has one lemma that falls within it - a different sense of that lemma falls within the first sense.

Okay, Logos, 'fess up - who didn't see that giving a comma a different meaning in the clause search than in other searches was a dumb idea?

OK, I haven't seen this feature in real life yet, but this seems perfectly logical to me. It's never occurred to me that verb:to say, to tell implied an OR (for that I would have expected verb:to say OR verb:to tell or verb:(to say,to tell). I've always associated to lexicons and the Search dropdown, where to x, to y would imply a different lemma than to x. But different sense makes sense as well (pun intended).

Ruben Gomez

https://community.logos.com/discussion/comment/427177#Comment_427177

I believe these results will become less baffling when Boolean operators are implemented as part of the Clause Search feature. I have no idea when that will happen, but I'm certainly looking forward to it.

Richard Lyall

https://community.logos.com/discussion/comment/427195#Comment_427195

This is not at all intuitive, and I wonder how useful it is as a feature in its current incarnation. It sounds like other search methods are more reliable in the meantime.

It seems to me that usability and intuitiveness are taking a back seat at the moment, because the feature is not properly thought through. To rely on the results at the moment requires an indepth knowledge of the process and the underlying data, which works against the whole purpose of it being an intuitive and easy to use feature it seems to me.

In the meantime is there ANY documentation about how to construct Clause Searches and how to understand the results?

Dave Hooton

https://community.logos.com/discussion/comment/427227#Comment_427227

In the meantime is there ANY documentation about how to construct Clause Searches and how to understand the results?

Nothing in the wiki.

MJ. Smith

https://community.logos.com/discussion/comment/427177#Comment_427177

but this seems perfectly logical to me.

The problem is that on one hand Logos tells us that they have assigned multiple senses when there are multiple senses that fit. However, when we go to create a search, the multiple senses suddenly become a single sense. Logos can't have it both ways - we need to be able to distinguish multiple senses from a compound name of a single sense. It had seemed fine to me until I starting trying to predict the results of small changes in the search argument.

MJ. Smith

https://community.logos.com/discussion/comment/427227#Comment_427227

In the meantime is there ANY documentation about how to construct Clause Searches and how to understand the results?

I can start some soon - after a Logos employee gives me a response to the one-to-one, one-to-many issue.