#3: Natural Language Processing − Cocktail Session

This is a past event

48 people went

Location visible to members


The third Natural Language Processing & Text Analytics Meetup will take place at die Mobiliar. Thanks to Mirco Rossi and die Mobiliar for co-organizing and hosting this session.

The Smart Analytics Team at die Mobiliar presents how NLP helps automating text driven processes in the insurance industry. Noëmi Aepli and Nora Hollenstein talk about the challenges while building a Swiss German corpus. Igor Nozhov introduces us to the state-of-the-art speech technologies through use cases. Each presentation is 20−25 minutes followed by a short discussion time. Please RSVP on the meetup site.


18:00 Welcome

18:10 Smart Analytics Team (die Mobiliar): How NLP is changing the Insurance Business

18:40 Noëmi Aepli (UZH) & Nora Hollenstein (IBM Research):NLP for Swiss German

19:10 Igor Nozhov (Spitch): Speech technology and its applications

19:00 Networking


Smart Analytics Team (die Mobiliar): How NLP is changing the Insurance Business

Description: Insurance companies are storing a lot of free text data, including text data generated by communication with customers, claim descriptions or underwriting. NLP can be used to automate the processing of such data, to extract additional information or to improve processes with techniques such as entity extraction, clustering and classification. In this talk we will give you some insights how and for which scenarios we are using NLP at Die Mobiliar.

About the Team: The Smart Analytics team of die Mobiliar is a group of 5 data scientists aimed at improving process efficiency and generating insight for new business opportunities from internal and external data. Mirco Rossi obtained a Dr. sc. ETH Zurich degree. In his Ph.D. he focused on machine learning for wearable computing. Teresa Kubacka obtained a PhD degree in experimental physics from ETH Zurich, with a special focus on designing proof-of-concept experiments and signal processing. Matyas Filep obtained a MSc in Mathematics at the University of Basel, gathered experience in the actuarial department at several insurance companies. All three are working now on evaluation and implementation of cognitive technologies in data-driven tools, mainly focusing on the text processing domain.


Noëmi Aepli (UZH) & Nora Hollenstein (IBM Research): NLP for Swiss German

Abstract: Natural Language Processing is challenging for any language, yet for dialects there are even more aspects to consider. This talk shall give an introduction to the challenges of developing NLP methods for Swiss German dialects, highlight the differences to Standard German and provide an overview of the research that has been done in this area in recent years. For this purpose, we will present our own contributions including the creation of a Swiss German corpus, part-of-speech tagging, dialect identification, dependency parsing and some observations regarding the phonetical aspects.

About: Noëmi is currently finishing her Computational Linguistics studies at the University of Zurich, developing a parser for Swiss German within the scope of her Master’s thesis. As CorpusLab assistant (@ URPP Language and Space at UZH) she has been involved in Swiss German NLP projects in recent years.

Nora studied Computational Linguistics at the University of Zurich and obtained a M.Sc. in Artificial Intelligence from the University of Edinburgh. After graduating she worked as a technical consultant for Watson projects at IBM before starting her PhD at IBM Research.


Igor Nozhov (Spitch): Speech technology and its applications

Description: This talk covers the technology stack developed at Spitch (www.spitch.ch (http://www.spitch.ch/)) as well as the real-life applications for Swiss German. The first part of the presentation introduces the state-of-the-art implementation of speech-to-text, voice biometrics including speaker verification and identification, semantic modeling for IVR system, and sentiment analysis of a live conversation. A data augmentation approach for acoustic and language modeling is considered to avoid sparsity problem.

The second part of the presentation is dedicated to the voice search in railway timetable scheduling, automatic subtitling for TV weather forecasts, and sentiment analysis of the live dialogues recorded in the call centers. The applications for the corresponding use cases will be shown during the demonstration.

About: Igor is the co-founder and CTO at Spitch AG. In the past he worked with Inxight, SVOX, Nuance and Yandex. Igor studied computational linguistics and artificial intelligence in Moscow and received his PhD in the field of information retrieval.


Die Mobiliar, Zurich 8050, Lamprechtweg 2


Die Mobiliar, Swiss Re