Morphology question

Question

A linguist in our area is wondering about the best way to use the morphology features of Paratext.
She is primarily using these features to help with spell checking, but also to confirm the kind of choices the translators have made. I don’t think she’s using the morphology to help with things like Biblical terms.

In trying to help her, I’m confronted with the fact that I do not know much about how Paratext is using the morphology parses, and I’m unsure about best practice in this area. I have already looked in the Paratext Help and some posts on the forum.

Here is the example from the linguist. In Mbugwe, she has been marking morphemes with a surface form, but in Rangi they are parsing with the underlying form. My intuition is that it would be best to mark the underlying forms. One wrinkle is that kee- is also another morpheme in Mbugwe.

orthographic word: kokeelaana ‘then we said goodbye to each other’
linguistic morphology: /ko-ka-e-laan-a/ 1PL-NARR-RECP-say.goodbye-FV
Paratext morphology in the Mbugwe way: ko-ke-e-laan-a
Paratext morphology in the Rangi way: ko-ka-e-laan-a

I think her primary question would be, what is best practice for marking morphemes in Paratext, so that it will make educated guesses about the morphology of other words (and possibly interlinear guesses)? How does it handle underlying morphemes that look quite different in the surface forms?

Answer 1

I have some suggestions about this topic. To fully use morphology in Paratext, then you must turn on the Matched based on stems option. Here is a bit from the help file:

image878×726 272 KB

Paratext will allow you to have more than one gloss for a morpheme if this option is turned on. This will allow you to handle homophoras morphemes,

Paratext parsing does not handle underlying forms. If you try to teach it underlying forms then you will quickly confuse it’s statistical guessing algorithm and it will produce nonsense. If you want to work with underlying forms then you have to use Fieldworks/Flex. Unfortunately, in the current state of how Paratext and Flex communicate, if one team member links their Paratext to Flex, then all team members must do it. If a single team member does not do this, then they cannot use the interlinearizer on their machine.

This video is a bit dated now, but it can give you some ideas about what Paratext can do with Biblical terms and morphology.

Comments

Thank you, anon044949. We have been using Match based on stems since before PT9. My problem is not how to do interlinearizing in PT; it is that PT now seems to be sharing its existing morphology and gloss data with FLEx—bidirectionally. Which is definitely not what we want, because we have been deliberately (and systematically) underspecifying morpheme breaks in PT to avoid bogging PT down, make it faster to do, and provide glossing that is more understandable by translation consultants, many of whom do not want to wade through a text with cryptic, linguistic paper-type morpheme glosses. We are now starting to enter our old Toolbox texts into FLEx and analyze them for linguistics work.

So now I seem to have a bunch of PT-style word morphology and gloss data in FLEx (only recent analyses from PT9 since the new PT–FW routines became active.) I had been assuming that the FW–>PT sharing was in one direction. I guess it makes sense to send PT analyses to FW, but it was added to PT9 without a warning to watch for unforeseen consequences. I have interacted with many other people who also “do” PT interlinearization like we do—targeting translation consultants rather than linguists, so I’m thinking there will be others interested in this topic as well.

Note for PT9 users: If you do not have a FLEx database attached to your Paratext project this won’t apply to you at this time.

So it would be helpful if someone could give us a description of what exactly happens during interlinearization in PT9, with an associated FLEx database. For example,

• Do parsing rules defined in FLEx apply in PT?

• Is it possible to turn off PT sending parsing and glossing info to FW?

• Is the PT interlinear updated on the fly when FLEx morpheme breaks/glosses have been edited?

• What happens when I edit some words in a PT interlinear verse (but ignore words that I’m not interested in today), and then re-approve the verse? Are old morpheme breaks/glosses sent to FW in this case?
  Thanks in advance for any insights.

---

KimB,

I think this should be another thread about Paratext-Flex integration. It might get noticed by people who know more about this if it is a separated threat. I will try and repost.

---

You’re right, anon044949. Thanks for reposting.

Answer 2

Related to this question: Can someone explain what is going on between Paratext 9 and FLEx (9, perhaps), please?

We have a lot of PT interlinearization in our project, done “roughly” over the years to benefit advisors and consultants, because we have known that “someday” PT would move to using FLEx-style interlinearization, which should be more precise and linguistically oriented. We have had our FLEx database associated with the PT project for several years, but we have not attempted to interlinearize Scripture in FLEx. I have seen evidence that PT9 is now finding FLEx entries, which is great.

Recently we have been working toward importing and interlinearizing non-Scripture texts in FLEx. This week while working with the many affix entries we see something weird and disturbing: It looks as though the Paratext morpheme breaks and glosses we create may be being transferred into FLEx automatically. Is this possible, or is something else going on–I’ve seen at least 150 FLEx lexical entries created over the past couple of years that look like our PT morpheme breaks and glossing style.

This automatic insertion into FLEx is not desired at this point in our project work, but if it is happening we need to know about it and modify our glossing procedures in PT. (During this time we have mostly been interlinearizing individual verses in order to get PT to recognize affixed Biblical Terms, which may explain the rather surprisingly low number of these FLEx entries.)

Can someone comment, please, or point me to an existing explanation of the changes to interlinearization in PT9?

Many thanks on the great development of both PT and FLEx and the integration features! We appreciate it very much; we just need to know how the changes are affecting our workflow.

Answer 3

Thank you for the feedback anon044949.

It seems like the Paratext morphology parsing isn’t really doing the kinds of things that we had perhaps envisioned. But knowing that is enough for now.
Thanks again.

Comments

In my use of morphology for checking spelling and verifying that the back translation adequately reflected the tense/aspect and other characteristics of the translation, I used the morphological parser in Fieldworks. Some languages may require that the phonological parser in Fieldworks be used. One such case is when tone carries a heavy grammatical load.