greek syntax Potential Bug in HITS command? Or user error?

[Norman Meyer]

Short Background: I'm relatively new to Accordance searching. I've watched several videos and read quite a few of the help articles. I'm trying to construct a more complex search, but I'm starting with sub-queries that I hope to eventually group together to get the results desired. The query below may seem very sophomoric, but I'm just trying to understand the software and I'm getting very unexpected results.

 

The Query: I am doing a simple Greek construct search in which I place a clause at the top of the first column. It's attributes are: independent clause, any clause type, ignore speech, default depth to infinity, and I check both "allow hits within sub-clauses" and "show hit antecedent or elision" (I'll provide screen shots below). I'm using the GNT28-T. Then below the clause block I put a lexeme block and have it search for the greek word “εἰ" (i.e. if, whether). I do the search and get 506 hits in 455 verses. The analysis window has some interesting results:

 

Note that the first analysis tab shown below is for the search described below. The tab to the right has the analysis from the search described above.

 

why are there other lexemes beside εἰ included in the results? I did a little investigation as to why this might be myself. I provide screenshots of 4 examples below, but it seems that if "εἰ" follows a null (i.e. "—") item (e.g. an implied subject) and that null item is associated with another lexeme (pictorially it shows an arc on the syntax diagram), then that lexeme is included in the results. This must be a bug, correct? I'm under the impression that Accordance 14 has some stability issues; is it possible that Accordance 13 wouldn't present this bug? Or am I misunderstanding something and this is not a bug?

 

 

I eventually want to use this query in another query so, to test hings out, I open up another GNT28-T query and type in [HITS NA28 Greek NT] and press enter. The results from this search, which I expect to be the same as the first, are quite different. This search returns 4190 hits in 2779 verses. The lexemes included in the hits now includes κάγω and it has many more occurrences of each lexeme: 

Quote

        διακονέω    (διά, κόνις)    to serve, wait on = 37
        ἐγώ    I (sg.); we (pl.) = 2581
        εἰ    if; whether = 502
        κἀγώ    (καί, ἐγώ)    and I (crasis) = 83
        λαλέω    (λάλος)    to speak = 296
        μένω    to remain, abide, stay = 118
        μετέχω    (μετά, ἔχω)    to partake = 8
        νίπτω    to wash = 17
        τις    a certain, a, one = 525
        χαρίζομαι    (χάρις)    to give; favor; forgive = 23

 

Why does this happen? From my understanding it should return the same results as the first search, i.e. 506 hits in 455 verses. I’m using the latest release of Accordance 14. Can someone please help me understand the results or if this is a bug in the software. Thank you very much

 

[Tim Planche]

Hi Norman

 

Interesting post - I First to say that I exactly reproduce what you find. I think what is being seen here is likely to be a feature rather than a bug. 

This is a syntactical search and not a lexical search and not having a background in linguistics I do find syntactical searches challenging and often a matter of trial and error.

As a very simple rule of thumb I find a lexical search easiest if I am looking for frequency of words, their usage or correspondences between Hebrew and Greek. Syntactic searches I tend to use for answering questions such as when would  "εἰ"  and "δε" occur in the same sentence or similar questions. 

 

The first thing i would say is if you do a simple lexical search for  "εἰ"  - without specifying a syntactic search in an independent clause the you get the very consistent search you are looking for with 502 occurrences of  "εἰ" occurring in 460 verses.  This gives an identical answer when using HITS search.

 

In terms of this syntactic search the easiest feature  to explain is results of words other than  "εἰ"   -this is because you checked the "show hit antecedent or elision" - this picks up phrases coded as being presumed antecedent  - this shows the word tagged as such - if you eliminate this option you then only hits of  "εἰ"  are returned.

 

Consequently the results returned by the  [HITS NA28 Greek NT] -  506  results refers to the 9 words  found in the  construct search and not to the numbers of occurrences of  "εἰ" returned by  the search. So when you use HITS it picks up there are 2779 occurrences of these words .  You can see this is  comparing the analysis of both the construct search and the HITS search  - I assume "καγω" appears as it is tagged as "εγω". 

 

The other options to examine, which I understand less,  would be the behaviour if you change the  depth of search and whether to look in sub-causes or not. -I tend to do this by trial and error. 

 

I am sure that there are others who do more syntactic searches than I and look forward to hearing from them,

[Tim Planche]

Dear Norman

 

sorry forgot to add - but these kind of enquiries are often better in the "original Languages" or even "Syntax Modules" sections 

 

though I am afraid I am not sure how to move this topic.

 

 

[Dr. Nathan Parker]

I can move this to one of those sections if need be. Just let me know where.

 

If there turns out to be a bug, let me know and I'll look into it.

[Norman Meyer]

@Tim Planche Thanks for the reply! I found your comments to be very helpful! However, in regard to the "show hit antecedent or elision" the extra lexical items are not antecedents of "εἰ", rather they are verbs connected to a NULL element, which have no syntactical relationship to "εἰ"; especially note 1 Cor 7:8-9, where the "εἰ" is in verse 9 but the null character, which directly proceeds it, is from verse 8 and is in a different sentence. Since the "show hit antecedent or elision" check box is not in version 13, I assume this is a new feature. I think the behavior exhibited on this search is bug; I cannot fathom an instance where someone would expect or want the results returned as shown above.

 

Regarding my other inquiry as to the appropriateness of the "HITS" query; I did find documentation which suggested it merely takes the lexemes from the original query and re-searches for those; although this behavior wasn't explicitly described. Thus, my expectations of the HITS command seem wrong and the software appears to be functioning appropriately. Unfortunately, this means Accordance does not appear to be able to perform the search I desire. I would like to find all instances of "εἰ" which are in an independent clause but not also hierarchically in a dependent clause. If anyone knows how to perform such a search, please let me know (I'll probably post this as a separate question in the Greek Syntax sub-forum). The usefulness of this search is to investigate the use of "εἰ" outside of the apodosis protasis construction. Thanks again for all the input!

[Norman Meyer]

I should also comment that I looked for documentation on the "show hit antecedent or elision" check box but was unable to find any information.

[Tim Planche]

Hi Norman,

 

glad I was of some help - I can see what your aim is and the search you want looks quite syntactical and probably somewhat beyond my limited linguistics skills. I think posting on the original languages or syntax forums may get you help from users with more experience. But I will have a try to see if I can see a solution 

 

[Tim Planche]

Dear Norman

 

sorry forgot to add - but these kind of enquiries are often better in the "original Languages" or even "Syntax Modules" sections 

 

though I am afraid I am not sure how to move this topic.

 

 

[Tim Planche]

Does this work?

 

I had a quick look and this return 228 words in 207 verses (I left the "show hit antecedent or elision" option unchecked)

seems to be getting what you were looking for - but i may be wrong - you may also need to tweak the search depth options

 

apologies if Ihave got the wrong end of the stick - I find these syntax searches a bit challenging