MCML Industry Pitchtalks Munich NLP

18:00 - 18:15

Open doors and welcome by MCML and Munich NLP

18:15 - 18:45

Verena Blaschke, PhD Student, LMU & MCML

Bavarian NLP: How (and why) to process dialect data

Natural language processing (NLP) has improved by leaps and bounds when it comes to processing data from standardized languages with plenty of available data, like German. However, NLP lags behind when closely related non-standard varieties are concerned, such as Bavarian dialects. I will give a brief overview of some of the challenges we face in dialect NLP: To start with, there is a general lack of high-quality data. On the technical side, subword tokenization methods sometimes struggle with ad-hoc pronunciation spellings, and evaluation metrics do not necessarily take into account linguistic variation. Furthermore, I argue that we should not only consider how to tackle dialectal variation in NLP, but also why, as the technology needs of dialect speakers can be different from those of standard language speakers.

18:45 - 19:15

Jenya Sukhodolskaya, Developer Advocate, Qdrant

miniCOIL: Sparse Neural Retrieval Done Right

I’ll present miniCOIL, an attempt to make a sparse neural retriever as it should be: a lightweight exact matches-based retrieval model with semantical understanding, that is performant on out-of-domain data.

19:15

Networking with Food and Drinks