Skip to content

Latest commit

 

History

History
71 lines (61 loc) · 7.77 KB

HELP_PREREQUISITES.md

File metadata and controls

71 lines (61 loc) · 7.77 KB
Raw

Prerequisites for Consultancy Help

I want to help as many people as I can get the most out of A-Z+T, and move forward the collection and analysis of consonant, vowel and tone data across Africa. To facilitate this goal, if you want help from me, I have drawn up a list of things you should do, below. These are not required to work with A-Z+T; you can do all kinds of work with A-Z+T under different terms, and I hope you do, if that's what you want. But I must keep my work organized, or I will have to help fewer people. So, if you take this list seriously, it will help me take you seriously. So here goes:

Get set up on languagedepot.org Set up a collaboration scheme that works for your team

  • Git is preferred, but Mercurial is also supported.
  • plug in a USB drive and tell A-Z+T you want to use it for backup.
  • use that USB drive to set up the project on other computers
  • do this first, as it may take awhile to finish
  • You will need an account (username and password) for yourself first, which you can sign up for here.
  • Once you have an account, sign in, and you will see instructions on the main page to ask for an repository for your language data. Basically, you will write an Email to [email protected], requesting an account with five lines of information. You want a Dictionary repository to work with A-Z+T; everything else should be straightforward.
  • Wait for a response from [email protected], which will give you your project id. It should be something like <ethnologue_code>-dictionary. If it isn't, please forward the Email to me and I'll advise you how to proceed.
  • While you wait, install Mercurial (Hg) so A-Z+T can make changes to your repository for you.

Prepare your LIFT database

  • If you have a database but not in LIFT (i.e., in FLEx, or a spreadsheet, or an electronic or paper document), then you must get help converting it. Some conversions are easier than other (e.g., from FLEx); others are very difficult (e.g., typing from a handwritten notebook). Of particular interest:

    • use Unicode only (no legacy fonts anywhere)
    • NFD (decomposed characters, but let's talk, if you must use NFC for some reason)
    • Check for consistency of part of speech marking
      • Think in terms of distribution, what syntactic frames/environments a morpheme goes in.
      • Check smaller parts of speech, to confirm that each is a paradigm, and everything you will want to compare is in the same ps.
      • Separate things like subj and obj pronouns, if they don't fit in the same syntactic space. Subcategorize, if you want to think of it that way.
    • Convert definition only senses to have glosses, as well.
    • Make citation forms, even if the same as lexeme forms. This should be pronounceable, and the right place to put a recorded form. Also, if you analyze a form later, you keep the full, pronounceable form (with recording) undisturbed.
    • If you have examples which are marked for a location field, not which locations you are using, so you don't reuse them as tone frame names. Otherwise, you could end up overwriting them (unlikely, I hope, but possible)
    • If you have a sense/field@type=tone currently in use, move or rename it (must!) This is where tone ur draft info is kept.

  • If you don't already have a database, and don't already use FLEx, download WeSay and follow the WeSay from Scratch directions to set up your project in WeSay set up your project there. start a new project in A-Z+T (you will need your language's ISO code)

  • Take a look at the practical prerequisites for your database, and see that yours has what it needs.

    • If you followed the WeSay from Scratch directions started your project in A-Z+T as above, you should be OK.

Install A-Z+T

  • Run this file (or use the simple instructions here). If you have any problems with the batch file, please let me know, and consider moving to the simple instructions. If you use the simple instructions, please work all the way down this page, and do everything on it. Don't stop unless you really can't continue. If you get stuck, please Email me with specific questions or problems.

#Develop and Define Tone Frames in process (to do)

Collect data in A-Z+T

  • Organize a meeting of more (3+) speakers, for them to sort and record data together in A-Z+T. You can organize this as you like, but expect at least two weeks of full time work for everyone.
  • At the meeting:

    • Sort data in A-Z+T:

      • As you sort,
        • each surface group should have just one pitch melody for the whole utterance, for each word in the group.
          • if not, remove the different ones on the "Verify" page.
        • each surface group should have a different pitch melody than every other group.
          • Take the time to compare each group with every other group!
          • if some are the same, join them on the "Review Groups" page.
        • keep going, around and around, until the each group is just one thing, and each is different from each other.
      • Start with:
        • largest 2-3 syllable profiles for Nouns in each of four tonally different frames
        • largest 2-3 syllable profiles for Verbs in each of four tonally different frames
        • There is a lot more sorting you can do, but this should be the right place to start, and will be a good beginning to understand your tone system.
        • Tonally different: if you don't know what this means, plan on getting help from a linguist near you, or sorting on at least eight frames (You should have this sorted out above, before starting a workshop).

    • Record all sorted data in A-Z+T

      • Please pay attention to
        • speak as naturally as possible, without unnecessary pauses. Pauses may break your recording into multiple phonological phrases, making your recordings useless for tone study.
          • if the syntactic context of a frame requires a significant pause (>150ms), you should probably use a different frame.
        • include the whole utterance in the recording (don't clip the first or last syllable)
        • not include extra space before or after the utterance
          • In case this last one sounds petty, extra space in the recording wastes computer space, internet bandwidth, and time in listening. For instance, if you have utterances that are one (1) second long, but you record three (3) seconds for each, then you are taking up three times as much space as you need, all of which needs to be stored on a computer, and shared across the internet. And you are asking your listener (me, other collaborators, a dictionary user) to spend that extra time listening to each file without reason. Two seconds wasted may not seem like much, but across thousands of recordings, it adds up.

    • Record citation forms for at least the above (largest 2-3 syllable profiles for Nouns and Verbs)

Collaborate and Get Help

  • Read bugs.MD, and be ready to send in reports
  • Once you have data you want feedback on, give me permission to read your languagedepot.org repository (Settings/Members/New Member), and send me your questions!