Papers by Miriam Winkler
Slot and Intent Detection Resources for Bavarian and Lithuanian: Assessing Translations vs Natural Queries to Digital Assistants (2024.lrec-main)
Copied to clipboard
| Challenge: | xSID datasets for low-resource languages like English are lacking in translations of high-ressource languages . however, many native speakers of such languages may want to use virtual assistants in their mother tongue . |
| Approach: | They extend a dataset to include two underrepresented languages: Bavarian and Lithuanian . they provide "natural" queries to digital assistants generated by native speakers . |
| Outcome: | The proposed dataset includes two underrepresented languages: Bavarian and Lithuanian . the results show that translated data can produce overly optimistic scores . |
Standard-to-Dialect Transfer Trends Differ across Text and Speech: A Case Study on Intent and Topic Classification in German Dialects (2026.acl-long)
Copied to clipboard
| Challenge: | Research on cross-dialectal transfer from a standard to a non-standard dialect variety has typically focused on text data. |
| Approach: | They compare standard-to-dialect transfer in three settings: text models, speech models, and cascaded systems where speech first gets automatically transcribed and then further processed by a text model. |
| Outcome: | The proposed model performs best on German dialect data while the text-only model perform best on the standard data. |