Help: Syntax search for adjacent matching forms

Most of the time, I thoroughly enjoy using Logos. On occasion, however, I want to throw my computer through a wall. Last night was one of those occasions. What those occasions have in common is that while Logos makes something possible, it doesn’t always make it easy or even discoverable.

My objective: find all locations in Johannine literature where the sequence “near demonstrative pronoun” + “article” + “noun” appears, consecutively, each agreeing in form. So I tried to construct a syntax search. None of the searches I attempted found any results, though I know they exist (e.g. John 8:23 - τούτου τοῦ κόσμου). One of the challenges is that the videos I found seemed to be mostly from L4, some from L5, and best I can tell L6 has changed the UI for this. Beyond that, they still have the problem of being more tutorial-oriented than principal oriented. In other words…

Which kind of syntax search should I use (opentext, lexham, …)? And why? Where are the guidelines for making that decision? Why are there no templates for syntax searches in the lexham syntactic GNT?

Once I’ve picked a type of search, which kind of search term do I start with (sentence, group, clause, phrase)? What’s the difference between adding something to that “term” to its right (connected by an arrow) versus below it (unconnected)?

The query I want to create seems to be about as simple a query as someone might want to run. That I have to search the wiki hoping for help and eventually post to the forums asking for help seems to be a comment on the usability and documentation of the software.

I’ve put screenshots of a couple of my attempts. Help!!!??? Thanks for bearing with my frustration.

Donnie

Find more posts tagged with

Comments

Fr Devin Roza

My objective: find all locations in Johannine literature where the sequence “near demonstrative pronoun” + “article” + “noun” appears, consecutively, each agreeing in form. So I tried to construct a syntax search. None of the searches I attempted found any results, though I know they exist (e.g. John 8:23 - τούτου τοῦ κόσμου).

Not to worry! Here is the search:

Here are some of the results from the search:

If you haven't already, be sure to check out the Faithlife group to learn Syntax Searching. There are loads of examples you can hook up to. I will upload this search as well.

https://faithlife.com/logos-syntax-searching/activity

Donnie Hale

https://community.logos.com/discussion/comment/682796#Comment_682796

Thanks for the prompt reply. It was good for me to work through recreating your query. Interestingly, the word "Dataset" does not appear at the end of the phrase "Cascadia Syntax Graphs..." for me. Of course, seeing the result doesn't answer some of my fundamental questions: Why Cascadia? Why a terminal node? Why couldn't the agreement have been for the "Instance" vs each individual item (case, etc.)? And how do I know that 3 terminal nodes automatically means adjacent / in succession in the text?

I have a variation on this that I'm trying. It works right up until I try to add agreement in the 3rd terminal node. I'm getting zero results when I add that. I'll post a screenshot shortly.

Thanks again,

Donnie

Donnie Hale

https://community.logos.com/discussion/comment/682830#Comment_682830

Here's the other query I'm trying to run:

This works as shown in the screenshot. It finds 68 results. What I really want is for the 3rd word to agree with the article-noun. However, as soon as I add even a single item of agreement (any of the four - I've tried all variations), I get zero results. I know there are matches, John 8:23 has τοῦ κόσμου τούτου in it.

Any thoughts on why this search isn't working? Also, I think that ideally the article portion of the search would be optional (present or not present is ok, as long as it has noun + agreeing near demonstrative).

Thanks again,

Donnie

Dave Hooton

https://community.logos.com/discussion/comment/682838#Comment_682838

Any thoughts on why this search isn't working?

Try

The dotted lines = "Matching skips Levels" (in right click box of the Nodes). Needed because there could be a another phrase before the Terminal Node (as in Jn 8:23).

The containing Phrase ensures the 3 words (Nodes) that follow are consecutive words in the text. Your construction allows the three words to be arbitrarily separated.

EDIT: Search range should be Johannine Literature, and results can be improved by specifying a Nominal Phrase (Phrase 1).

Donnie Hale

https://community.logos.com/discussion/comment/682908#Comment_682908

The dotted lines = "Matching skips Levels" (in right click box of the term). Needed because there could be a another phrase before the Terminal Node (as in Jn 8:23).

Thanks for the reply, Dave. I have some questions about your comment on matching skip levels. I clearly see that if I don't have that checked for any of the terminal nodes, I get zero results. First, can you elaborate on why it gets zero results if any of the nodes don't have that checked? Does "phrase" have a technical definition here? Second, when it is checked for all of them, why do I get multiple hits in the same verse for the same sets of consecutive words (many of those hits for much longer than 3 words, e.g. John 7:36). Third, what I'm looking for would be exactly three consecutive words: article+noun+pronoun (near demonstrative), all agreeing. Is that what you think your search should yield?

Again, thanks.

Donnie

Fr Devin Roza

https://community.logos.com/discussion/comment/682838#Comment_682838

Any thoughts on why this search isn't working? Also, I think that ideally the article portion of the search would be optional (present or not present is ok, as long as it has noun + agreeing near demonstrative).

That search wasn't working because in the Cascadia Syntax Graphs the different Terminal Nodes aren't displayed in the same "column" when this type of construction is used. When you don't group items under a header like "Phrase" or "Clause" they have to be in the same column.

So, when Dave suggests grouping them under Phrase, and then selecting "Match skipping levels" this overcomes this problem. Match skipping levels tells the Syntax Search to not care how "deep" under the Phrase marker your "stuff" is (to use the technical term [:)]). Just look for it, in the order you specify.

It's not correct that the way you have the search above allows for word gaps or word order variation - that would only be the case if you included things like "Anything" or "Gap" or "Unordered Group". Rather, by default the Syntax Search by default always looks for things in exactly the order you put them, with no spaces in between, unless you tell it you don't want it to do that.

Note that in the Cascadia Syntax Graphs, for a search as simple as yours its actually possible to just use the "Word" column, which is what you were trying to do originally. It works very well. Note that it actually provides MORE "real" hits than when you use the option to group the words under a "Phrase", which provides more search results hits, but with lots of duplicate. The reason that the "Word" version of this search provides more "real" hits is that there are some occurrences that actually occur between two different phrases - like Jn 10:5-6! This is interesting, as the punctuation was added later, of course. These occurrences that cross "phrases" won't be returned if you tell the Syntax Search you want all three words grouped under the same phrase, so Jn 10:5-6 is not returned in the Phrase version.

Here is what the Word version looks like:

Note as well that you could replicate the original search you asked for with this style as well, using Words instead of Terminal Nodes.

Donnie Hale

https://community.logos.com/discussion/comment/682930#Comment_682930

Here is how that looks like:

I'm working on replicating this now. Out of curiosity, how do you get the "part-of-speech" entry on a Word node? I don't see anything like that in the options for a Word node in L6?

Donnie

Donnie Hale

https://community.logos.com/discussion/comment/682931#Comment_682931

how do you get the "part-of-speech" entry on a Word node?

Nevermind, I figured it out. It's called "Morphology" when you add it, but "part-of-speech" in the node. Not helpful...

-Donnie

Fr Devin Roza

https://community.logos.com/discussion/comment/682931#Comment_682931

Use the @ in the Morphology field, and just select "Article" or "Noun" or "Pronoun" and leave it at that.

Fr Devin Roza

https://community.logos.com/discussion/comment/682933#Comment_682933

Not helpful...

Yep, welcome to Syntax Search! [;)][B]

Luckily it seems Faithlife little by little is replacing all the functionality of syntax search with easier to use things like Clause Search, search extensions like Speaker, etc. It's getting harder and harder to find searches that require syntax search, and hopefully within a few years there won't be any need to use Syntax Search at all, except for incredibly specialized cases that are not really that useful anyway and just kept on for backward compatibility.

Dave Hooton

https://community.logos.com/discussion/comment/682929#Comment_682929

First, can you elaborate on why it gets zero results if any of the nodes don't have that checked?

By definition, the terms you specify are consecutive so that is what the search looks for. So it will miss this database construction

(p) ---> (p) ---> Terminal Node

and only find:

(p) ---> Terminal Node

Most syntax constructions belong to the first type, so we put "matching skips levels" to skip the second phrase (p) or any other intervening term in the syntax database. Else we would have to specify every possible intervening term when we just want to get to the last term.

Does "phrase" have a technical definition here?

Cascadia overuses the term (technical) but the 'phrase' of more than 1 word is what we would use.

More later ...

Donnie Hale

https://community.logos.com/discussion/comment/682934#Comment_682934

Use the @ in the Morphology field, and just select "Article" or "Noun" or "Pronoun" and leave it at that.

The search worked. The results looked reasonable.

Next question. For Word1, the article, under "General" I changed it to "May or may not appear" (displayed as: "Repeats = Optional"), the goal being to find anarthrous nouns followed by the pronoun. (I don't know whether any such constructions appear in John's writings.) When I ran that search, I got zero results. That makes no sense to me. If it's "optional," then it at least ought to find all of the results it found when it was "required" (appears at least once).

Thoughts?

Donnie

Fr Devin Roza

https://community.logos.com/discussion/comment/682944#Comment_682944

Next question. For Word1, the article, under "General" I changed it to "May or may not appear" (displayed as: "Repeats = Optional"), the goal being to find anarthrous nouns followed by the pronoun. (I don't know whether any such constructions appear in John's writings.) When I ran that search, I got zero results. That makes no sense to me. If it's "optional," then it at least ought to find all of the results it found when it was "required" (appears at least once).

Looks like if you are using the "Agreement" section, setting the item that is set to agree with another item to "May or May Not Appear" breaks the search and always returns 0 hits. Ouch.

I would think this is a bug. Any thoughts about this, Dave or others?

As a workaround, you can always add a "Group" to the left of the Words, then add an "OR" under the "Group", and then add another "Group" under the "OR" and then under that, the version without the definite article.

Donnie Hale

https://community.logos.com/discussion/comment/682969#Comment_682969

Looks like if you are using the "Agreement" section, setting the item that is set to agree with another item to "May or May Not Appear" breaks the search and always returns 0 hits.

Interesting. I tried a different workaround. I took the agreement rules off of Word2 and then set rules on Word1 and Word3 to agree with Word2. With just that change, the search worked as before. But when I set Word1 to optional, it was back to zero results. Seems like a bug to me, but what do I know?

As a workaround, you can always add a "Group" to the left of the Words, then add an "OR" under the "Group", and then add another "Group" under the "OR" and then under that, the version without the definite article.

Is there a way to do this without deleting the words that are already in the query? It seems very inflexible in working with nodes (words, phrases, terminal nodes, etc.) that are already on the query - connecting them, or changing what they're connected to (e.g. if I want to go from a phrase to a group).

Donnie

Donnie Hale

https://community.logos.com/discussion/comment/682969#Comment_682969

As a workaround, you can always add a "Group" to the left of the Words, then add an "OR" under the "Group", and then add another "Group" under the "OR" and then under that, the version without the definite article.

OK. I was able to accomplish this as you described. (Why is it "Group" when you add it, but "Wrap" once it's displayed???) However, there is a slight problem with the results. It finds the first set of results in the "OR" results, too - which makes sense. Ideally I wouldn't see the same set of words twice in the results (again, John 8:23 is a good example). Is there a way to say: Don't include results from group 2 whose words are contained within a result from group 1?

I appreciate the help and patience.

Donnie

Dave Hooton

https://community.logos.com/discussion/comment/682930#Comment_682930

It's not correct that the way you have the search above allows for word gaps or word order variation - that would only be the case if you included things like "Anything" or "Gap" or "Unordered Group".

My statement above - "Your construction allows the three words to be arbitrarily separated." - was incorrect!

Note that in the Cascadia Syntax Graphs, for a search as simple as yours its actually possible to just use the "Word" column, which is what you were trying to do originally. It works very well. Note that it actually provides MORE "real" hits than when you use the option to group the words under a "Phrase", which provides more search results hits, but with lots of duplicate. The reason that the "Word" version of this search provides more "real" hits is that there are some occurrences that actually occur between two different phrases - like Jn 10:5-6! This is interesting, as the punctuation was added later, of course. These occurrences that cross "phrases" won't be returned if you tell the Syntax Search you want all three words grouped under the same phrase, so Jn 10:5-6 is not returned in the Phrase version.

My real intent was that results are not bounded/constrained to a grammatical unit like phrase/clause with a "simple" word search as you get spurious results like Jn 1:1-2 and Jn 10:5-6! It has nothing to do with crossing verse boundaries, but that they cross phrase or clause boundaries and hence provide results that not meaningfully related (as is often the case with text Bible/Morph Searches with an arbitrary proximity value).

The value of the "real" hits in this case is that you omit the extra results showing the different possible phrase boundaries within which the "real" result can fit. But it will give spurious results. When using Syntax search you are trusting the analysis of the text (Cascadia vs. OpenText vs. Lexham) knowing that a Morph Search can't provide the "accuracy" you require, so it is best to avoid a "simple" Syntax word search even if you want to use the Agreement operator.

Fr Devin Roza

https://community.logos.com/discussion/comment/682994#Comment_682994

As a workaround, you can always add a "Group" to the left of the Words, then add an "OR" under the "Group", and then add another "Group" under the "OR" and then under that, the version without the definite article.

Yes, it's actually quite easy. Just hover over the left side of the Word box, and click on the + sign. A box will be added to the left of the Word box. Then, you can drag the other words under the first one and they will all fall under the same Group.

To remove them from a group it's similar - if you just delete the group, they will all be deleted, but if you drag them outside of the group, you can then delete the Group without affecting the Words.

TIP: Note as well - you can create one "Group" with the three "Words" underneath it. Then, just right-click on the first Group box, and select Copy. Then you can just paste a second Group, with the three Words underneath it, and then add in the Or box.

Fr Devin Roza

https://community.logos.com/discussion/comment/683003#Comment_683003

OK. I was able to accomplish this as you described. (Why is it "Group" when you add it, but "Wrap" once it's displayed???)

Syntax Search is full of little quirks like this. It's part of having a way-over-the-top-powerful search engine that is so complicated that it is somewhat abandoned even by its creators. [8-|] Oh well.

However, there is a slight problem with the results. It finds the first set of results in the "OR" results, too - which makes sense. Ideally I wouldn't see the same set of words twice in the results (again, John 8:23 is a good example). Is there a way to say: Don't include results from group 2 whose words are contained within a result from group 1?

I wish there were a way - at the very least, it would be nice if the Syntax Search returned both the "number of hits" and the "number of verses", as other searches do. It is so common that you get multiple hits per verse that this would really help make the search results more usable.

Fr Devin Roza

https://community.logos.com/discussion/comment/683285#Comment_683285

My real intent was that results are not bounded/constrained to a grammatical unit like phrase/clause with a "simple" word search as you get spuriousresults like Jn 1:1-2 and Jn 10:5-6!

Good point. In most cases grouping under a Phrase does in fact limit the search results to what we are really interested in....

Donnie Hale

https://community.logos.com/discussion/comment/683384#Comment_683384

It is so common that you get multiple hits per verse that this would really help make the search results more usable.

Which is why I've asked, for many years now, for a SQL-like query syntax in Logos. If they gave me a dataset, I'd be willing to write a prototype.

Donnie