1 of 14

MARGO BLEVINS

mblevins@utexas.edu

The Texas German Dialect Project

The University of Texas at Austin

Systematic Normalization of

Spoken Mixed-Language,

Mixed-Dialect Data

Palma de Mallorca

16 May 2026

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

2 of 14

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

NLP

tools

standard(near)

3 of 14

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

NLP

tools

language�contact

NLP

tools

natural speech

NLP

tools

dialect

4 of 14

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

NLP

tools

language�contact

NLP

tools

natural speech

NLP

tools

dialect

5 of 14

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

tok

ham

ham

lang

deu

eng

norm

haben

ham

1

2

Annotate transcribed forms with their standard language equivalents

6 of 14

Language-tagging

  • Form
  • Semantics
  • Part-of-speech

​

​

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

standard language reference work

🡪 deu, eng, spa, ces … (ISO 639-2)

7 of 14

Tag

Explanation

Example

deu.txg

No lexical entry

Stinkkatze ‘skunk’

Non-standard � meaning

Hochschule

‘high school’

Non-standard

POS

mexikaner ‘Mexican’ (adjective)

mix

Morphological mixing

gejumpt ‘jumped’

🡪 mix.deu+eng

8 of 14

NORM – standard languages

  • Stay as close to the original utterance as possible
  • Be in reference work (form + intended meaning + intended part-of-speech)
  • No changes to case, gender, tense, etc.

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

9 of 14

NORM – non-standard material

  • Morphologically normalized according to donor language
  • jeschukt 🡪 geshuckt ‘shucked (corn)’

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

10 of 14

Additional guidelines for …

  • Ambiguity,
  • Cut-off words,
  • Backchanneling,
  • Vocalized hesitations, …

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

11 of 14

Summary & Conclusion

  • Standardized, flexible system for (mixed-language, mixed-dialect) spoken data,
  • Explicit, linguistically grounded basis 🡪 transparent and reproducible
  • Publicly accessible

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

12 of 14

Summary & Conclusion

  • Not a one-size-fits-all solution
  • A starting point for other projects
  • Transparency of pre-processing decisions
    • Facilitates downstream processing
    • Improves accuracy of future research

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

13 of 14

Texas German Sample Corpus

  • Annotated transcripts of spoken Texas German
  • ~13.5 hours
  • 75,000+ tokens

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

https://doi.org/10.18738/T8/IOX9ZA

14 of 14

References

TEXAS GERMAN

DIALECT PROJECT

Dedicated to the Preservation of Texas German

Blevins, Margaret. 2022. The language-tagging & orthographic normalization of spoken mixed-language data, with a focus on Texas German. PhD Dissertation. The University of Texas at Austin. Austin, Texas. doi: 10.26153/tsw/43598

Blevins, Margaret, 2022, "Texas German Sample Corpus", https://doi.org/10.18738/T8/IOX9ZA, Texas Data Repository, V1