Phrase/Word Classification

[rwrobinson88]

I have a question about when constituents are considered phrases or words. 

 

For example, in Philippians 2:15 (ἵνα γένησθε ἄμεμπτοι καὶ ἀκέραιοι, τέκνα θεοῦ ἄμωμα μέσον γενεᾶς σκολιᾶς καὶ διεστραμμένης).

 

I was looking to search for an appositive phrase with two adjuncts. Here is the syntax graph:

 

 

I figured I'd put θεοῦ ἄμωμα both as adjunct words within the appositive phrase. I didn't get this passage back. I scratched my head. Then I tried two adjunct phrases instead. Got the hit.

 

Why would these be marked as phrases? Also, how is one to tell if it is a phrase versus a word?

 

Thanks!

[Λύχνις Δαν]

I had the same experience in a different part of syntax the other day so I would be interested in more clarity on this point too.

[Robert Holmstedt]

edit-- this is all obviously from the Hebrew database perspective

 

I was going to answer this right away, but thought it'd be good to test things out a bit. I've run into a strange tagging bug that kept me from getting back to this. And even now I'm not sure I can explain the hits numbers.

 

In any case, in linguistic terms, there really is no difference between appositive word and appositive phrase. With that said, Hebrew morphology is fusional, which means that a single "word" may have more than one syntactic constituent (e.g, when a complement is attached to the verb as a clitic pronoun, or a possessive pronoun is attached to a noun). So it could be a useful distinction, though I can't think of a really good contrastive example at the moment. 

 

As far as Accordance searches go, I *think* that an appositive word search will hit any word that is tagged an appositive regardless what comes after it, while an appositive phrase looks for the entire appositive, whether it is a single word or a very long phrase. What I went searching for was to see if the appositive word search simply looked for the first word after the appositive tag, or if there was some other discernible criteria. And that's when I hit my distracting tagging puzzle, so I don't have an answer yet.

 

An appositive clause is obviously an appositive with its own predication (e..g., Gen 1.4 'And God saw the light, [that (it) (was) good]', where I take the KY/'that' clause to be in apposition to the NP complement 'the light')

Edited November 29, 2016 by Robert Holmstedt

[rwrobinson88]

Thanks for expounding. 

 

Just to be clear. In the NT Database, specifically with the reference above, I only get a hit on this if I put θεοῦ ἄμωμα as two adjunct phrases. If the theory that backs up the database is cool with one word being phrase, I don't have quibbles. The issue becomes when does one know if it's a phrase or a word? Especially when it's only one word.

 

​It seems like you believe that it should come up whether I mark it as an adjunct phrase or word? If I'm reading you correctly and that is the case, there is a bug. 

 

​Please confirm. I'll want to submit it as a bug if that is the case.

 

Thank you!

[Robert Holmstedt]

An appositive word search should hit the first word, τέκνα. An appositive phrase search should include the whole phrase, τέκνα θεοῦ ἄμωμα. 

 

If not, it's probably a minor bug.

[Λύχνις Δαν]

Hi Robert,

 

Thanx for investigating.

 

The problem is not so much the Appositive searches which do indeed function as you describe. The problem is with the Adjuncts. If one searches for an Appositive phrase with two subordinate Adjuncts (words) Phil 2:15 is not returned. If you use to Adjunct phrases it is. If you use Adjunct phrases each with an Adjunct word beneath it also works. If the search engine always treats a single word constituent (whether adjunct or other type) as a phrase when searching that's probably ok. If it does not then one must know a priori whether constituents like this are phrases or words, or one must try both.

 

Another way of looking at it is that I had always assumed that the last constituent shown in the graph was the syntax classification of the actual word, in this case Adjunct words. What this query would seem to suggest though is that the last constituent in the graph is not the actual word's classification but rather that the two are distinct, the A's in the chart being phrases and the words (θεοῦ and ἄμωμα) being Adjunct words.

 

I should add this is all with depth=0 searches. If one increase the App Phrase search depth to 1 and searches for App with two subordinate Adjunct words then one finds Phil 2:15 also. Again this tends to support the above, that the A's are phrases and the words themselves Adjunct words.

 

Thx

D

[Jordan S]

Perhaps this is a bug. I'm running 12.0.2b1 and I'm having no issues finding these syntax combinations in Phil 2:15.

 

 

 

 

 

 

[Λύχνις Δαν]

Hi Jordan,

 

  Try reducing the depth in query two to 0. That should fail to find Philipians 2:15.

  So long as it does then I think I may understand this, though any confirmation or correction of my speculations above would be welcome too.

 

Thx

D

Edited November 30, 2016 by דָנִיאֶל

[Jordan S]

A search depth of 0 provides the same results as my first image. I need to set the depth to 1 to find the results in the second and third images.

[rwrobinson88]

It fits with what Robert says that the words themselves are tagged as adjuncts on the word level (the depth gives us the hits because the words are 1 deep). 

 

That makes sense to me. So, what I'm concluding is that an Adjunct Word is at the word level. If I see an "A" then I can know that is an adjunct phrase.

Edited November 30, 2016 by RyanWRobinson

[Λύχνις Δαν]

A search depth of 0 provides the same results as my first image. I need to set the depth to 1 to find the results in the second and third images.

 

Initially I didn't quite follow your response here but yes in the first search depth does not matter but in the other two it certainly does. Thanx for the confirmation.

 

 

It fits with what Robert says that the words themselves are tagged as adjuncts on the word level (the depth gives us the hits because the words are 1 deep). 

 

That makes sense to me. So, what I'm concluding is that an Adjunct Word is at the word level. If I see an "A" then I can know that is an adjunct phrase.

 

Agreed. One could argue that the syntax charts would be clearer if they were to annotate the words themselves with the word level syntax but in any case we know now. Perhaps the doc could clarify this point.

 

Thx

D

[Robert Holmstedt]

Yes, the last tag in the graph gives the word's syntactic category. Thus, θεοῦ and ἄμωμα are both adjuncts and would be in the hit results for an adjunct search. But as adjuncts, they modify a higher constituent, τέκνα, which is tagged as an appositive. Thus, if one searched for appositive phrases, all three should be highlighted by virtue of the two adjuncts being within the domain of the appositive they modify. This is the hierarchy of syntax.

 

The depth should not be an issue here. An appositive phrase search should result in the entire phrase, including the hierarchically lower elements relating to an appositive noun. 

Edited November 30, 2016 by Robert Holmstedt

[Robert Holmstedt]

The line from the last tag/label to the word itself (and that fact that the word does not have a label beside it) is not clear? This is fairly conventional.

Labelling the words themselves might be seen as a confusion of syntax (which is about function relative to other words) and the word category (noun, verb, preposition) or even morphology. I'm not sure I think this would end up any clearer.

[rwrobinson88]

I'm fine without it being marked on the word level myself. It's clear to me now how it works. The "A" before the line to θεοῦ for example is a phrase. Or to state it another way, word categories (complement, adjunct, etc) are only the word level. 

 

That seems stupid to say now that I'm saying it that way, I typically have thought of phrases being more than one word within syntax. But, hey, I'm not a linguist. (I'm just starting to Van Valin these days...sorry Robert...RRG ) 

[Robert Holmstedt]

RRG? Oh dear.  ;-)

[Λύχνις Δαν]

The line from the last tag/label to the word itself (and that fact that the word does not have a label beside it) is not clear? This is fairly conventional.

Labelling the words themselves might be seen as a confusion of syntax (which is about function relative to other words) and the word category (noun, verb, preposition) or even morphology. I'm not sure I think this would end up any clearer.

 

The point I was trying to get at with my comment was whether the word itself had a syntax tag which was distinct from that of last tag in the chart before the word. The syntax charts don't show one, and you indicate here that there isn't one, but the searches appear to behave as though there is. If there is not then I agree there would be no value in affixing a tag to the word.

 

As to whether it would be appropriate linguistically to show it, if it exists then it makes sense to show it, if not then not. The syntax charts overlay the text describing a model that the text fits into. If the words in that model have another label (a syntactic one in this case) I have no issue with it. As it is, instant details reports a syntactic label for each word along with other annotations like morphology.

 

Ok so I've gone over this in both Greek and Hebrew and it is the same in both. I understand that this is how it works and it appears that there is no problem once one understands, except that one cannot assume that putting an ADJUNCT tag in your search will find a hit even if it is the last tag before the word itself. One must use an ADJUNCT PHRASE, and if one wants, then an ADJUNCT.

 

I've attached an example workspace with four queries that indicate the different results and highlight what I'm describing. Of specific interest is the difference between "App Phr w Adj word" and "App Phr w Adj phrase w Adj word". This appears to imply distinct phrase and word level tags, though I that it need not to be so in the database itself.

 

I'm pretty sure that we are merely describing this in different ways.

 

Oops : in the Hebrew examples this is the test verse that I was working with, the second half having a similar structure to that of the Greek example.

 

”וּלְשֵׁ֥ם יֻלַּ֖ד גַּם־ה֑וּא אֲבִי֙ כָּל־בְּנֵי־עֵ֔בֶר אֲחִ֖י יֶ֥פֶת הַגָּדֽוֹל׃“

(בראשׁית י׃כא Hebrew Masoretic Text with Westminster Hebrew Morphology)

accord://read/HMT-W4#Gen._10:21

 

Thx

D

Greek Phrase vs Word.accord.zip

Hebrew Phrase vs Word.accord.zip

Edited December 1, 2016 by דָנִיאֶל

[rwrobinson88]

I thought that since we can see what the word is tagged for with instant details that sufficed. 

 

I can say that my original question has been answered though. So, I'm happy.  

[Robert Holmstedt]

Daniel,

 

I think that in order to understand the answer, you're going to have to shift from thinking about syntax as word-based to thinking about it as constituent based and hierarchically phrasal. Some Hebrew "words" have more than one syntax label (e.g., when there is a cliticized pronoun) and some words actually share a syntax label (e.g., when they constitute a compound), though in the latter case the trees duplicate the label as a convenience. And one word may have a label (e.g., appositive) but the phrase indicated by the label doesn't end with the word but includes a few more (e.g., when the appositive has its own adjuncts). 

 

The searches don't behave any differently than the trees, as long as one understands the nature of the syntax -- hierarchical and phrasal. This is true whether a phrase consists of one word or forty. And perhaps that's where you should begin by thinking -- in syntax, there is no such thing as a word. There are constituents, which may be simple or complex. They are all phrasal, even if the phrase consists of one item.

 

The way the syntax was really built was to use the phrases in searching. The non-phrasal options, which use "word" only as a convenience (what other label would users understand as indicating a simple constituent without any modifiers?), are necessary in order to drill down within the phrasal structure to identify the syntactic constituents (smaller phrases) within a larger phrase. Since it is obviously impractical to avoid providing a simple syntax label to insert when looking for constituents that are simple enough that a single morph is co-extensive with a syntactic role, we provided the non-phrasal "word" options. 

 

Unless this has been confusing across-the-board even though few have articulated it as you have here, I would venture to say that this approach can be taken as fairly intuitive. In fact, from working with a variety of users, I suspect that the way it operates appears simpler than even my description here, which is probably going to seem a bit tortured for most.

 

But if this is not correct and there has been general confusion about the words versus phrases, others need to chime in so we can re-evaluate the searching interface (well, at least the labels of the items in the interface).

Edited December 1, 2016 by Robert Holmstedt

[Λύχνις Δαν]

Hi Robert,

 

I'm fine with "word" for "constituent". A simple label that is a little imprecise is fine. But where I started this was a little different. Anyhow, the system is perfectly usable I believe and I do understand that when one wants to construct a search for a multi-level structure all the way down to final constituents (words, clitics, what have you) that you cannot represent the last tag in the chart before the constituent itself in the search by a "word" element - at least not where "word" and phrase versions of the same element exist. One must first insert a phrase element and then the relevant "word" element. That's what my examples really show. And that's fine.

 

I think what I am asking goes to implementation and representation of the data in the database and how it's processed. But once the above is understood then this is not really necessary for a user to know - but my com-sci data processing background .... Anyhow if the database can be used there isn't really a problem. And it can.

 

I suspect that the way it operates appears simpler than even my description here, which is probably going to seem a bit tortured for most.

 

Agreed. Describing all this in detail in text is way more convoluted than it actually is, even in the code I suspect. If we were standing in the same office with a white board, or pencil and paper, I think this conversation would have been much simpler

 

Thanx for taking the time.

 

Thx

D

[rwrobinson88]

Thanx for taking the time.

 

 

Yes. Thank you! Getting the clarification cleared things up for me. For most, intuitive. For the slower (I am reading RRG after all ), the details helped to clarify.