summetix uses Argument Mining and large language models to discover hidden insights in your customers' feedback and to monitor and discover tech and innovation trends.
Our argument mining based technology discovers truly actionable insights.
We reduce the amount of text you need to screen by 90%.
We read at least a 1000x faster than humans, but have human-level accuracy.
In the first step, customer feedback is reduced to the core information, i.e. meaningful statements about specific error patterns, but also the strengths of a product.
In the next step, similar groups of such statements are combined into clusters containing repeatedly mentioned product problems and strengths.
Clusters are then correlated with customer feedback metadata like a production date, a product line, or sales market. Based on this, our customers can prepare or adapt specific actions for product development and improvement or communication campaigns up to 6 months earlier.
Davide Cadamuro, AI Architekt at BMW Group IT
"With the help of the argument mining solution from the deep learning start-up summetix, we can extract essential information from social media. Our quality assurance department as well as brand and product development can use that information to derive specific instructions for optimizing our products and customer experience."
Tobias Kehl, Project Lead AI Startup Rising at hessian.ai
"Working with summetix, we wanted to make a unique contribution to the 2021 federal elections in Germany. Using the company's AI technology, we were able to analyze all press coverage surrounding the elections and present viewpoints and arguments of political actors on any topic. The freely accessible tool has helped many voters to get independent and diverse information based on their own priorities. The cooperation with summetix was a lot of fun. All our requirements were met or exceeded despite a tight timeline."
Oliver Heyden, Chief Strategy Officer at pressrelations
"We are fully convinced that summetix' technology clearly improves the way we satisfy our customers' information needs. summetix' explorative approach to uncover the tone and key argumentation in online news generates results very close to our analysis approach as well as very fast, in a matter of minutes instead of days. With summetix' analytics technology we will enlarge our current product portfolio to fulfill our customers' needs to gain deep textual insights and react to a dynamic market environment much faster than currently possible."
summetix GmbH (formerly known as ArgumenText) is a Spin-Off of the Ubiquitous Knowledge Processing (UKP) Lab at the Technische Universität Darmstadt. The technology behind summetix is based on several years of intense research in multiple areas of Natural Language Processing, Deep Learning and Argument Mining. The results have been published in top-tier scientific venues. Below is an overview of the most important components of our system.
Advanced machine learning techniques ensure high precision in your applications. Details in:
Argument similarity controls a precise and scalable grouping of statements.
Multilingual models are available for your language-dependent document collections. Details in:
summetix currently has multiple job openings, among others we are looking for fullstack developers, sales assisstance and working students (IT and business support). If you're interesting in working with us, get in touch via email@example.com.
Most tasks in NLP require labeled data. In this recently published paper (at the 17th Linguistic Annotation Workshop held in conjunction with ACL 2023), we investigate how removing personally identifiable information as well as applying differential privacy rewriting enables using crowdsourcing for text with privacy-relevant information.
summetix will be at this year's Hessian AICon in Darmstadt! On July 6th we will show you how you can combine Argument Mining and Chat-GPT to avoid hallucinations and how summetix uses AI technology to inform world-leading companies about product problems six months earlier than before.