Worum es bei uns geht
Bevorstehende Events (1)
Hey folks, In this meetup, we will explore the intersection of behavioural economics and data, i.e. nudging. We will also dive into Name Entity Recognition and the implementation of BERT in this context. You can find the details about the topics and the speakers below. I hope to see as many of you as possible. Let's push this topic and make data great (again?)! If you have any question, idea or suggestion, write me an email at: [masked] Since then, have a restful Christmas time PS: Data Science meetups will now take place on a regular 4-weeks basis! So save the date for February, 21st! ------------- Topic 1: NUDGING WITH DATA SCIENCE Speaker, Nathan Maddix, Behavioral Economist at Max Planck Institute Summary In the last 20 years, tremendous gains have been made in both cognitive and computer science. These improvements in data quality and collection now allow researchers to study how small changes -- called "nudges" -- can alter human behavior, especially when applied to user-centered applications and programs to help individuals to make better decisions, make interactions more enjoyable, and improve experiences in general. This talk covers the basics of nudging and behavioral change efforts with the goal of making use of data science to study human behavior. Common challenges and approaches will be discussed when it comes to combining human behavior with tech-based applications in the burgeoning field called "behavioral design." The Speaker Nathan Maddix is an expert in nudging and behavioral science, working on numerous projects that involve big data, behavioral interventions, and technological applications. He will share his experience with user-centered design in nudging, and discuss how data science can increase both impact and scale of results. Nathan previously worked and studied at Harvard University and now runs behavioral and economics experiments at the Max Planck Institute in Bonn. ------------- Topic 2: UNDERSTANDING TEXT WITH CUTTING-EDGE DEEP LEARNING METHODS Speaker: Timo Möller, Co-founder deepset Content One key component for understanding text is a method called Named Entity Recognition (NER) which has a history dating back to the 90s. Entities are concepts like persons, places, organisations, etc. With this method you can resolve spelling errors or synonyms (apple = fruit or organisation), you can also detect previously unseen concepts only from their character pattern and surrounding context. In this talk we will focus on modern ways to do NER. We will introduce a very deep Neural Network (BiLSTM + CRF) and a recent method from Google called BERT, a model that learns a general understanding of language and is then fine-tuned for tasks like NER. We will show a real world application of NER that we have implemented for a large scientific publishing company and show how to use BERT for NER but also other interesting use-cases. The talk is directed to people interested in applying Machine Learning on text on a real-world use case - but also a more technical audience familiar with Neural Networks, to get a deeper understanding of RNNs and the mechanisms underlying BERT. About us Deepset is a tech startup focused on Natural Language Processing (NLP). By utilizing cutting edge Machine Learning models and applying them to customer-specific use cases, we deliver value into various industries, like manufacturing, aerospace or financial services. Our customers are ranging from startups to multinational corporates (DAX30, NASDAQ etc.) and trust us with access to their most sensitive data. Out of Germany we are bringing NLP to work all over Europe and Asia. The speaker Timo has more than a decade of experience in ML, has studied Data Science abroad and theoretical Neuroscience in Berlin. There he also worked in a startup as a ML engineer, developing Deep Learning solutions on text data. Together with two friends he founded the startup deepset in Mai 2018.