SpinningBytes – Text Analytics & Speech Processing

Our Areas of Expertise

SpinningBytes is a Swiss company that develops state-of-the-art solutions for Natural Language Processing (NLP). We started in 2015 when new methods in machine learning showed excellent results for several NLP tasks. Our CEO, Mark Cieliebak, is also Professor for Text Analytics and Dialogue Systems at the Zurich University of Applied Sciences (ZHAW), and we closely collaborate with his research team.

Our customers are companies and organizations who want to extract useful information from their text and audio data with state of the art technology, for example ChatGPT. Here is a selection of tasks for which we already implemented successful solutions:

Chatbots

Chatbots can talk to customers and help them solve their tasks.

Dialogue Systems

We have the know-how to build reliable dialogue systems at scale.

Sentiment Analysis

Is a text positive or negative? Our algorithm won SemEval, one of the most prestigeous international competitions.

Topic Categorization

Distinguish hundreds of pre-defined topics/categories and label incoming documents accordingly.

Automated Customer Support

Free up valuable resources without sacrificing the human bond to your customers.

Author Profiling

Segment your customer base with automatic age, gender, and geolocation detection from text messages.

Speech-to-Text

Automatically transcribe audio recordings into text.

Machine Translation

Tap into the power of machine learning to move seamlessly between different languages.

Headline Generation

Generate genuine headlines for a given news article.

Speaker Recognition

Use voice-specific features to identify speakers for security or forensics.

Summarization

Generate brief summaries of single documents or large document sets.

Text-to-Speech

Generate audio from a given text in differente voices and speaker styles.

What about ChatGPT?

Large Language Models such as ChatGPT or Falcon show an impressive performance for tasks such as summarization, re-phrasing, or ideation. We are working with large language models since 2019, when we implemented first projects using a GPT model. Since then, we have systematically evaluated the capabilities of ChatGPT and other Models such as Falcon, LLaMA, Vicuna etc. and implemented several prototypes and applications using these models (together with the Centre for AI at Zurich University of Applied Sciences).

If you are interested in using a Large Language Model, please get in touch!

Swiss Technology on the Inside

SpinningBytes is an accredited spin-off of the Swiss Federal Institute of Technology (ETH) and the Zurich University of Applied Sciences (ZHAW).

The technology of SpinningBytes was developed in three research projects funded by
Innosuisse/CTI, the Innovation Agency of the Swiss Government.

Listen to our Clients

Marcel WengerInnovation Manager

University Hospital Zurich

“From the presentation, we got a good overview of what’s going on in machine learning for health.”

Irene AmstutzCEO

Swiss Economic Archive

“We were impressed by their expertise in machine learning.”

Janis GrunderCo-Founder

professional.ch

“The first project was a good first step—we already started a follow-up project with SpinningBytes.”

Award Winning Technologies

Winner of SemEval 2016

Winner of Evalita 2016: Sentiment Polarity Classification

Winner of GermEval 2020

2nd Place at WNUT-2017: Named Entity Recognition on Tweets

2nd Place at GermEval 2018: Hate Speech Detection on Tweets

2nd Place at VarDial 2018: German Dialect Identification

3rd Place at VarDial 2020

We Shape our Projects to your Needs

Custom Projects

Our team of experts can support you in your NLP project. We offer:

Brainstorming Workshops: We quickly assess if your problem can be solved and how much resources will be needed. Duration: 2-4 hours
Rapid Prototypes: We build a first prototype on your data, evaluate the resulting quality and give recommendations for implementing a productive solution. Duration: 5-15 days
Production-ready Solutions: We develop a software component that solves your text analytics task. This includes definition of architecture and interfaces (e.g. REST), performance optimization (quality and runtime) and deployment in your infrastructure. Time and cost depend on the size of the project.
Data Labeling: We organize the labeling process for you, from finding annotators to quality control.
Data Analysis and Reports: If you need one-time insights from your textual data, we can apply our tools to your data, extract useful information and generate a written report.
Training and Education: We offer trainings on the fundamentals, applications and methods of Natural Language Processing. Target audience ranges from novices to tech-experts.