Significant Problems with the Bible Sense Lexicon

Page 1 of 1 (17 items)
This post has 16 Replies | 1 Follower

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Posted: Sun, Oct 13 2013 11:59 PM

I've been working with the BSL for a few days, and have come across several problems with the data, which has significantly hampered my study. My study involved looking at the claim that the ancients considered the 'bowels' as the seat of the emotions. Here are some of the problems I encountered:

  1. One of the senses show in a BWS for מֵעֶה ('bowels') is anguish. But if you open the BSL to that sense, the lemma מֵעֶה ('bowels') is not shown in the lemma section. It's Jeremiah 4:19 that's missing (but that's tagged with the sense entrails (compassion)). There needs to be consistency between the tagging and the BWS data.
  2. If you visit the BSL for internal feature, it shows a lemma of τὰ ἐνόντα, as the result of incorrect tagging.
  3. The BSL entry for to be deeply moved (bowels) is orphaned, and not part of the tree. The same is true of to be hard-hearted ⇔ shut one’s bowels.
  4. Likewise to be hard-hearted ⇔ shut one’s bowels gives the lemma of κλείω τὰ σπλάγχνα, which is fine (although it's not a lemma, but a phrase). But the English gloss is incomplete: "to shut".
  5. The entry for heart organs (internal feature) has no data at all.
  6. The entry for internal feature ⇔ bowels ought to have "internal feature" in it's relationships section. And given it's position in the tree, it's name should be reversed bowels ⇔ internal feature.

I understand the BSL is a work in progress, but the presence of so many small errors did significantly hamper my study.

I also identified a couple of other problems with the BWS:

  1. If I type מֵעֶה ('bowels') into a BWS reference box, the only lemma suggestion is מעה ("grains").
  2. If I run a BWS on מעה ("grains"), two of the lemmas show glosses for מֵעֶה ('bowels') - specifically GHCLOT and BYBHV. Those lexicons and  NASB Dictionaries and CDWGTHB also points to the wrong word.

This is with Logos 5.1b SR-2, with a BSL dated 2013-09-25T00:00:29Z.

Posts 21846
Forum MVP
Graham Criddle | Forum Activity | Replied: Mon, Oct 14 2013 12:04 AM

Mark Barnes:
This is with Logos 5.1b SR-2, with a BSL dated 2013-09-25T00:00:29Z.

I've just checked and - probably unsurprisingly - the same issues seem to be present with a BSL dated 2013-10-01T17:47:30Z on 5.2 Beta 1.

Posts 10107
Denise | Forum Activity | Replied: Mon, Oct 14 2013 8:00 AM

Wow. Thanks Mark!!! Another way to crash Logos5. I clicked on your reference logos4:Senses;KeyId=ws.internal_feature.n.01 and promptly crashed my L5. I have to refine it a bit, so that it will qualify for the crash-L5 competition, but I'm confident.  I'm guessing it doesn't buffer the memory correctly and so other competitors won't be able to replicate it easily.


Posts 227
LogosEmployee
Mike | Forum Activity | Replied: Mon, Oct 14 2013 10:01 AM

Hey Mark,

I'd be glad to look into this for you. I don't know much about Hebrew, but if I can get an understanding of where the issue is, maybe I can get someone involved who knows more on the topic. Please confirm if I'm following correctly:

I opened Jeremiah 4:19 in the ESV and generated a BWS for "anguish" (מֵעֶה - h:me'eh). One of the definitions shows up as "anguish" in the wheel, as translated from Je 4:19 and the sense, "entrails (compassion)" is listed as one of the possible senses for me'eh.

To my untrained eye, this all seems to be as it should so far.

I then attempted to generate a new BWS by typing "h:me'eh" in the reference box. There, I was given a suggested result for "grains" (מָעָה - h:ma'ah or h:m'h). I don't know why this one lists a vowelled and unvowelled lemma, but maybe that's why BWS suggests it and not the other.

So I am confused about whether there is some problem with "entrails (compassion)" as the sense for me'eh, or if the real concern here is that BWS doesn't suggest me'eh as a possible entry in the reference box. If you can clarify that for me, I'll submit a report and we'll see what we can find out for you.

Mike

Faithlife Tech Support
Win Logs|Mac Logs|Training

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Replied: Mon, Oct 14 2013 11:13 AM

One problem is that you can't generate a BWS for "anguish" (מֵעֶה - h:me'eh) by typing into the BWS box. You can only do it from the right-click menu.

The problem relating to Jeremiah 4:19 was that if you look on your screenshot above, the BWS gives the sense of 'anguish' for (מֵעֶה - h:me'eh), but the BSL doesn't give (מֵעֶה - h:me'eh) as a lemma for 'anguish'. I thought that Jer 4:19, might be the verse that should be tagged differently, as that's the only verse I know where (מֵעֶה - h:me'eh) might be argued to have that sense.

Posts 227
LogosEmployee
Mike | Forum Activity | Replied: Mon, Oct 14 2013 12:37 PM

I'll see what I can find out, but I think that according to BSL, that specific  "anguish" sense does not apply to Je 4:19. The sense "entrails (compassion)" applies to the Je 4:19 usage. This sense also applies to the same lemma in Je 31:20, Job 30:27, Ps 40:8, Song 5:4, Isa 16:11 and Lam 1:20 and 2:11.

I think the confusing thing is that BSL is almost reverse/counter-indexed. It's organized by the individual senses, not the lemmas or lexical definitions. So looking at the entry for "anguish" will only show occurrences where that lemma matches that particular sense. None of the verses I listed should show up for the "anguish" sense either, but they are the same lemma.

This gives us some scholarly input when trying to determine which sense the author intended when they used that particular word in that particular situation, rather than leaving us to pick from the list of possible definitions/senses for ourselves; like the Louw-Nida lexicon, but in an interactive word-tree graph format.

Mike

Faithlife Tech Support
Win Logs|Mac Logs|Training

Posts 24837
Forum MVP
Dave Hooton | Forum Activity | Replied: Mon, Oct 14 2013 2:02 PM

Mark Barnes:
If I type מֵעֶה ('bowels') into a BWS reference box, the only lemma suggestion is מעה ("grains").

That is precisely what I got trying to reproduce Issue 1 after pasting the lemma! I could only get the BWS with multiple senses after a BWS on bowels, clicking the ring in Hebrew Words & then clicking the lemma (doubtless you generated it from one of the references e.g. from Nu 5.22 in a bible).

Dave
===

Windows & Android

Posts 113
LogosEmployee

Hi Mark,

Thanks a lot for your feedback. We are glad to hear any time that someone is using the BSL, and hopefully we can help with or at least give some insight into some of these issues.

On the data end I can respond to a couple of these points (1, 3, 5, and 6), but we will need to look more closely at points 2 and 4. For points 2 and 4 part of the issue is that you are right that the Bible Sense Lexicon is not always dealing with "lemmas," but sometimes with larger units. This is causing a mismatch where the larger unit is getting only the gloss for the first item in the unit. As per 1, 3, 5, and 6:

1. We will be fixing this. I'll offer a bit of explanation as to why it currently looks like this. We had initially left lemmas attached to senses even if we did not use the senses in the annotation of the lemmas. We felt like this might alert users to the fact that there were other options in case they disagreed with our tagging (though I'm not certain why anyone would ever want to do that Big Smile). But, this seems to be causing some confusion.

3. There are sometimes orphans. This will often be the case for verbs with labels that begin with "to be ..." The reason for this is that we had too many stative verb senses (listed as "result" in the BSL tool) underneath the sense "to be" for it to seem like useful information to have them all listed there. We hope that some of the orphans will get connections when we add the adjective data because many of the orphaned senses are stative verbs that could have connections to adjective senses. But, if there are too many disconnected senses left we may have to revisit this.

5. This is similar to #1 above and should be fixed soon.

6. I'm not certain I understand this particular issue. In the BSL internal feature ⇔ bowels does have a connection to internal feature.

Our convention is to have the more concrete meaning on the right and the extended meaning on the left. Please let me know if I am misunderstanding.

I hope that some of this helps for the time being.

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Replied: Mon, Oct 14 2013 2:44 PM

Mike (Tech Support):
I'll see what I can find out,

<edit> I posted whilst Jeremy was answering the question. But as I spent a long time making a pretty picture, I'm leaving it up for posterity (and for others who be still be confused!). But Jeremy's answer to #1 answers this question.</edit>

I'm obviously not making myself clear Smile. How about a picture?

The same data is (supposed) to be the same in the two different places. Yes, it's organised differently — the BWS is organised first by lemma and then by sense, and the BSL is organised first by sense and the by lemma — but the specific sense/lemma combination should be the same in both places.

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Replied: Mon, Oct 14 2013 3:01 PM

Jeremy Thompson:
Thanks a lot for your feedback. We are glad to hear any time that someone is using the BSL, and hopefully we can help with or at least give some insight into some of these issues.

Thanks, Jeremy for your full response. I appreciate it.

(1) and (5). I can see what you were trying to do. It would be great if you could tag a specific occurrence of a lemma with multiple senses, but if you can't, then it does confuse people. Well, if confused me, anyway.

(3) Fair enough.

(6) What I wrote didn't make sense, sorry! What I was trying to say that internal feature <-> bowels is not a child of internal feature, but a sibling. And the point about order was probably caused by the fact that all the others have '(internal feature)' afterwards. I realize that's conveying different information, but it probably conditioned me to see the same word order applied here.

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Replied: Mon, Oct 14 2013 3:02 PM

Mike,

Thanks for your contributions. I'm happy now on the BSL questions, but can you chase up the two BWS questions, please? Thanks.

Posts 139
Sean Boisen | Forum Activity | Replied: Mon, Oct 14 2013 6:24 PM

Mark Barnes:

<snip/>

(6) What I wrote didn't make sense, sorry! What I was trying to say that internal feature <-> bowels is not a child of internal feature, but a sibling. And the point about order was probably caused by the fact that all the others have '(internal feature)' afterwards. I realize that's conveying different information, but it probably conditioned me to see the same word order applied here.

I think Mark has this now, but just in case it's not clear:

The convention for a sense S whose label is "X <=> Y" is that S is "a kind of" X, but its concrete expression or conceptualization is as Y. In the current data (I'm looking at 5.1b SR-2), the sense "internal feature <=> bowels" is indeed a child of "internal feature", along with "bosom (thoughts)", "congnition", "conscience", and other kinds of internal features.

When we write a label like "X (Y)", that's because the English word X is itself ambiguous: Y is supposed to help resolve the ambiguity and give you a clue as to the intended meaning. So we have "heart (internal feature)" because "heart" in English can mean several things: an internal bodily organ, or a location (the heart of the city), or a central point (the heart of the matter), or others. In this case there are several children of "internal feature" whose plain English label would be ambiguous (liver, loins, spirit), so we've tried to consistently label then with "(internal feature)". But this is more about providing helpful and consistent labeling.

Posts 139
Sean Boisen | Forum Activity | Replied: Mon, Oct 14 2013 6:36 PM

Denise:

Wow. Thanks Mark!!! Another way to crash Logos5. I clicked on your reference logos4:Senses;KeyId=ws.internal_feature.n.01 and promptly crashed my L5. I have to refine it a bit, so that it will qualify for the crash-L5 competition, but I'm confident.  I'm guessing it doesn't buffer the memory correctly and so other competitors won't be able to replicate it easily.

Hi Denise:

This doesn't crash for me (5.1b SR-2, with DB:WORD-SENSES as of 2013-09-25T00:00:29Z). If you can reproduce this crash (especially after updating), can you send in a log so we can diagnose the problem?

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Oct 15 2013 4:23 AM

Sean Boisen:
The convention for a sense S whose label is "X <=> Y" is that S is "a kind of" X, but its concrete expression or conceptualization is as Y. In the current data (I'm looking at 5.1b SR-2), the sense "internal feature <=> bowels" is indeed a child of "internal feature", along with "bosom (thoughts)", "congnition", "conscience", and other kinds of internal features.

Thanks for dropping by, Sean.

I understood X <=> Y to mean that X and Y are effectively equivalent — two different ways of expressing the same sense. So in this example, kidney (in the sense it's meant here) is effectively equivalent to conscience, because conscience  <=> kidneys.

You're saying that's not the case?

Posts 139
Sean Boisen | Forum Activity | Replied: Tue, Oct 15 2013 8:29 AM

Mark Barnes:

Sean Boisen:
The convention for a sense S whose label is "X <=> Y" is that S is "a kind of" X, but its concrete expression or conceptualization is as Y. In the current data (I'm looking at 5.1b SR-2), the sense "internal feature <=> bowels" is indeed a child of "internal feature", along with "bosom (thoughts)", "congnition", "conscience", and other kinds of internal features.

Thanks for dropping by, Sean.

I understood X <=> Y to mean that X and Y are effectively equivalent — two different ways of expressing the same sense. So in this example, kidney (in the sense it's meant here) is effectively equivalent to conscience, because conscience  <=> kidneys.

You're saying that's not the case?

It's a little tricky to talk about these senses and their relationships, but i don't think "equivalent" is helpful.

Meaning in this system is mostly provided by a definition and by structural relationships (the parent/child hierarchy). So the label "conscience  <=> kidneys" is.trying to convey that it's the sense of the parent 'conscience', but expressed as though it were kidneys. I usually read this notation as "conscience conceptualized as kidneys": we also talk about it as metaphorical extension. Usually the extension is lexical: it might be the same word that's normally used to concretely express kidneys as a bodily organ, but here's the word is instead expressing the notion of conscience.

But also keep in mind, "conscience  <=> kidneys" is its own sense, distinct from "conscience", and also distinct from some literal sense for 'kidneys". It's clearer to talk about sense by their IDs (though they're not user friendly): so logos4:Senses;KeyId=ws.kidney_psychological_feature.n.01 is a kind of logos4:Senses;KeyId=ws.conscience.n.01 ... oops! except that these are indeed siblings, i believe incorrectly so. Looks to me like (per our normal convention) kidney_psychological_feature.n.01 should be a child of conscience.n.01, not a sibling. Maybe that's what your earlier post was trying to point out and i didn't understand. But we'll take a closer look at the hierarchy here.

Posts 113
LogosEmployee

This is actually incorrect as it stands. I think this was an issue of merging two languages. Conscience only has a Greek lemma, whereas conscience as expressed by kidneys has both Hebrew and Greek lemmas, and the Greek lemma there probably took the Hebrew hierarchy at some point. We will get this fixed.

In general, we might say that conscience and conscience expressed as kidneys do "mean" roughly the same thing. However, with metaphors sometimes the range is not always the same, so for example, they might occur in different kinds of contexts or be slightly more restricted in meaning. So, rather than grouping the metaphors in with the more general synsets we thought it would be more valuable information to use the more general synsets like grouping mechanisms with the metaphors as children (though we failed in this case).  

In the case of something like "power," being able to tease out the imagery I think gives some really wonderful illustrative verses for a Bible study or sermon as opposed to just looking at verses with generic words for "power."  I think there is one particular verse in Luke where God's power is referred to in terms of his "finger" that I found exploring with the tool. And, I just thought that was really cool. So, that is partly our intention with the extended senses.

Posts 13342
Forum MVP
Mark Barnes | Forum Activity | Replied: Tue, Oct 15 2013 9:18 AM

Sean Boisen:
It's a little tricky to talk about these senses and their relationships, but i don't think "equivalent" is helpful.

I guess I assumed equivalent because the <=> symbol led me to the belief that whatever the relationship was between the two words, then the relationship was the same in both directions. If that relationship is not the same in both directions, then I believe a better symbol might be conscience => kidneys.

Sean Boisen:
Looks to me like (per our normal convention) kidney_psychological_feature.n.01 should be a child of conscience.n.01, not a sibling. Maybe that's what your earlier post was trying to point out and i didn't understand.

I perceived a difference from the way bowels and kidneys were being treated. I assumed bowels was wrong, but if kidneys is wrong, that would certainly explain my confusion.

Page 1 of 1 (17 items) | RSS