Page Comparison

•  Andrew Collins list of options of drop downs for Best estimate dialect, here is a draft: https://utexas.box.com/s/ktgzmnrte2t4uujmpg0uq479ew58mqk8 after consensus in next meeting, Sonia will add it on Smartsheet - add a column “Best estimate dialect” with all dialect options, drop down options specially for acoustic analysis, make list, consensus and create

• Castilian dialect - regional variation

• Mexico dialect

• Puerto Rico dialect

• Catalan - regional variation, català-central, etc.,

Rule add dialect of participant to the drop down

other info:

• Hometown (where did they grew up)- sociodemographic questionnaire

Next tasks pending to assign:

• to move MINT samples here: https://utexas.box.com/s/00jk6h0p5tw1900bcyuanr1sziqmsywe MINT samples are saved here: https://utexas.box.com/s/6dh0k3zje1xencqh5nh1mbdj8l8rttpx and check it here: https://app.smartsheet.com/reports/3Hg9hM3vq6QQcc7hQ7Cxv7rr79r5c8F6cx793j61 (important task!)

• Meetings needed, when

  • Connected Speech planning (structure) - tasks pending - pre-r01 data, MINT.. see previous notes, no meeting needed, just more support, clipping students can help

  • Connected speech and VISTA transcription progress + Connected Reliability process

  • Connected Speech Clipping (Shayla lead)

  • Connected Speech Whisper (Whendy lead) Alejandra, Alexa - once trained just check every 2 weeks how they are doing

  • Connected Speech data analysis

• we have noticed that fo Spanish samples code switches are transcribed sometimes like (difficult on how similar languages are)

  • paraphasias, fragments, xxx

  • SOLUTION:

    • PROPS1: include it within the reliability, Rater 2 should be Helena/Núria/Sonia

    • PROPS2 (CWR): Having a code switching checker, if she notices a code switch she checks it in the smartsheet -

      • Alyne (brasilian spa-cat, recognizing code switching) non native speaker

      • Do a test (5 samples), if she misses one code switch is out - influence of catalan of samples, take notes, grammar, vocabulary, - 90% accuracy, (blinded) - transcripts+audios

    • PROPS3 (SMK): Helena/Núria/SMK listen to all WAB picnic and change if there are code switches - mark it in column in Smartsheet - Decide task by task – maybe Núria first

•  create a column in CATALAN, SPANISH, ENGLISH smartsheet to check if there is a code-switch

recommendation to use chat to ask questions in case there is a code-switch

Training -CATALAN transcriptions - Helena

Helena training complete, high reliability

• next steps - Diana has to give access to all systems

• which samples assign?

  • 1. CS picnic scene reliability? - 1 more sample

  • 2. VISTA samples?

• •  Ana Elisa Quinonez Camacho ask Lucian status of transcription 27 Nov 2024
  •  Ana Elisa Quinonez Camacho 02 Dec 2024 monday check rely with Lucian’s transcriptions

Chat preguntas transcripción

1. Cuanto tiempo toma hacer preguntas-respuestas

2. sistema mejor?

3. doubts

Discussion - effect on reliability, faster-more efficient

Have same for Helena

check by Mor doesn’t work

• Whendy update on students -

• Solution to not asking Kesha every time for access - do not log out

• Ana Elisa added to chat

  • after they run whisper, Whendy will meet with them and explain them where to find the audios and where to find them.

    •  Sonia Marqués create 3 report with whisper transcription columns (SPA, CAT, ENG)
    •  Whendy Avila Motta add hierarchy in whisper guide https://cloud.wikis.utexas.edu/wiki/x/twc-Dg https://cloud.wikis.utexas.edu/wiki/x/UYBZAw

• Instructions added for Rely: https://cloud.wikis.utexas.edu/wiki/x/jg9rDQ

• Connected Speech Reliability planning in Marqués, Sonia K: Connected Speech Reliability planning - updated after SM+AE meeting channel 

•  documentation
•  9. Connected Speech Reliability
•  https://cloud.wikis.utexas.edu/wiki/x/dI08Dw (In progress, need to discuss in meeting)
•  Folder structure
•  CS ok!

• Ana Elisa created the CS folder structure for Spanish for Kesha’s project and for Dr. Santos' project during the week of 11/25. Note, since this process/structure was developed after reliability was run, the transcriber 1 samples are the final transcripts with consensus.

•  VISTA probes - https://utexas.box.com/s/krxn8jx3vza7rwy98bziou0e0wfgnrdh In progress, need to discuss in meeting.

• Ana Elisa added the PicnicScene_WAB_PostTX_Coded_Rely samples and the PicnicScene_WAB_PreTX_Coded_Rely samples to this folder structure.

•  Reliability management
•  Smartsheet Catalan reliability report: https://app.smartsheet.com/reports/96Xjp4G54hwCXM3pVVHP5xcrvVFfPXgh9wq49g81 
•  Spanish reliability report, Ana Elisa creates it

•  VISTA reliability report - agree https://app.smartsheet.com/reports/GWGRp9wPqwgG8C52724F7jFqH3q83rV55gVHrX91

• teams work well, Jan excel

• ask student to create a deposit of doubts using https://utexas.box.com/s/t7zdxjtgspsbk2ofqvhlda85u6zx104b as template 2 tabs, questions of utterance segmentation, questions about coding, and a column specifying the coding using the wiki guide (ej. Punctuation, Capitalization, Numbers, Titles)- this way we can filter it and consult previous decisions before asking a question and also responding it appropriately - same way + having a corpus of decisions

Words not in CLAN library

example: ma (madre) detected as copula

mamí is not correct - correct in library

• issues with adjectives in MOR

https://utexas.box.com/s/7cl2a29xk7ddvc2yn0woa6cvthcefyat

Decision: code as paraphasia or keep as-is

• últimamente, los últimos quince años fue<eh director &+d general de una compañía inglesa>

  • sample: https://utexas.box.com/s/tr8t0iixuu5ki4xc3ztrfwruqcxjtflf

  • audio: https://utexas.box.com/s/5u1nxwn808mkskpm9yed9dontrr0sbpi

• siembra/siembren

• Decision 20241119 Code as phonological paraphasias

Sonia Marqués add to the guide