Still trying to understand how Case Frames are presented and how they relate to Clause Participants
I'm looking at a number of verbs and trying to understand how the BWS sections on Clause Participants and Case Frames represent them and their associated arguments / adjuncts.
The example in this post is καταλύω as used in the Gospels but I am seeing similar things with lots of different verbs so think there is something I am not understanding - or there are some bugs!
The screenshot shows that Clause Participants just have three "Agent" entries (1) and all associated with Jesus - but the Clause Search keyed off that entry returns four results (2).
But the Case Frames section shows a total of 7 Agent instances (5 with Patients, one with a missing Patient and one with a missing Experience). I don't understand why this doesn't match the number shown in Clause Participants.
And in some cases the Agent is not visually identified in the Case Frame section:
- Mt 5:17 with Patient (3)
- Mt 5:17 with missing Patient (4)
- Lk 19:7 with missing Experiencer (5)
For the same verb (but not shown on the screenshot), there are 12 Patient instances associated with this verb but only 8 are recognised in the Case Frame section. Yes another seeming discrepancy that I don't understand.
Am I misunderstanding how this works, are there some issues with tagging, some bugs in the display engine or something else?
Thanks, Graham
Comments
-
Hi Graham:
Thanks for continuing to interact with us on the semantic roles. There are a couple of things going on here:We do have a bug here with Luke 19:7, I think, due to the way we have that particular clause structured syntactically. I'll be looking into that.
Another part of the issue is with how we are handling infinitives. There is no noun or pronoun as the subject of the infinitive in our syntax to hang the Agent on. So, this is often responsible for a discrepancy in counts. In other words, the infinitive may still be an Agent-Patient frame, but the Agent not be explicit or be elsewhere in the verse associated primarily with another verb. One instance of this is "I am able to destroy ..." There the "I" goes with "am able" and "to destroy" is the infinitive. I'll be thinking about whether or not there may be a better way to handle the infinitives, but I'm not sure.
I can think of other cases where the number off participants that are Agent or Patient (or any given role) may be fewer than the frames that show up in the case-frames section. For example, one of the roles may be filled by an entire phrase that doesn't correspond to any one participant that we have in one of our datasets. So, it's sometimes better to think of the Agents (or any other role) that you will find in the clause participants section as a subset of the Agents that you would find in the case-frames section.
Other times the counts may be higher in the participants section, as you noted with Patients here. I think that may be primarily due to "Herod's Temple" and "Jewish Temple" providing the same results in the participants section.
All that to say, I think some of the discrepancy in counts is normal behavior.
0 -
Hi Graham,
Mt 5:17 should have been showing two instances of Jesus as Agent in the Clause Participants section. That was a bug that I fixed. This is what it looks like on my machine now.
Graham Criddle said:For the same verb (but not shown on the screenshot), there are 12 Patient instances associated with this verb but only 8 are recognised in the Case Frame section. Yes another seeming discrepancy that I don't understand.
Clause participants shows the biblical people, places and things associated with each role. When words are tagged with multiple entities they will show up twice in the clause participants section. Here, Herod's Temple and Jewish temple are tagged for the same words.
In general, the Case Frames section shows all of the semantic role analyses, whereas Clause Participants shows those analyses filtered through the biblical referent data, which isn't as comprehensive and precise.
Scott
0 -
Hi Jeremy
Thanks for continuing to interact with us on the semantic roles
Thanks for the responses - very helpful as I try and understand how this all works[:)]
We do have a bug here with Luke 19:7, I think, due to the way we have that particular clause structured syntactically. I'll be looking into that.
Fair enough.
Another part of the issue is with how we are handling infinitives. There is no noun or pronoun as the subject of the infinitive in our syntax to hang the Agent on. So, this is often responsible for a discrepancy in counts. In other words, the infinitive may still be an Agent-Patient frame, but the Agent not be explicit or be elsewhere in the verse associated primarily with another verb. One instance of this is "I am able to destroy ..." There the "I" goes with "am able" and "to destroy" is the infinitive. I'll be thinking about whether or not there may be a better way to handle the infinitives, but I'm not sure.
That explains a lot and I now understand something of the problem - if you can figure out some way to represent this it would be great but I see it could be a challenge.
I can think of other cases where the number off participants that are Agent or Patient (or any given role) may be fewer than the frames that show up in the case-frames section. For example, one of the roles may be filled by an entire phrase that doesn't correspond to any one participant that we have in one of our datasets. So, it's sometimes better to think of the Agents (or any other role) that you will find in the clause participants section as a subset of the Agents that you would find in the case-frames section.
Other times the counts may be higher in the participants section, as you noted with Patients here. I think that may be primarily due to "Herod's Temple" and "Jewish Temple" providing the same results in the participants section.
This is also very helpful insight.
All that to say, I think some of the discrepancy in counts is normal behavior.
And I understand better now what that is the case.
Thank you, Graham
0 -
Hi Scott
Scott Fleischman said:Mt 5:17 should have been showing two instances of Jesus as Agent in the Clause Participants section. That was a bug that I fixed. This is what it looks like on my machine now.
That looks a lot better[:)]
Scott Fleischman said:Clause participants shows the biblical people, places and things associated with each role. When words are tagged with multiple entities they will show up twice in the clause participants section
Scott Fleischman said:In general, the Case Frames section shows all of the semantic role analyses, whereas Clause Participants shows those analyses filtered through the biblical referent data, which isn't as comprehensive and precise.
That makes things much clearer to me
Many thanks, Graham
0 -
0
-
0
-
Hi gents
Just looking at this further and I have seen something I don't understand in the Case Frame section - specifically the first "Agent-Patient" case
Why is the Patient shown as "Prophets" as opposed to "the Law of the Prophets"?
The object of the verb seems to include them both
Thanks, Graham
0 -
Graham Criddle said:
Why is the Patient shown as "Prophets" as opposed to "the Law of the Prophets"?
This was caused by a tagging omission combined with a bug in the way our automated clause processing aggregated conjunctions. The next release of clause data should fix it.
0 -
Thanks Peter
Appreciated, Graham
0
