Analysing speech and visualising opinions
Russia and Germany. From Confrontation to Cooperation
We created the media installation for the exhibition “Russia and Germany. From Confrontation to Cooperation” which was on display in 2015 at Berlin’s Martin-Gropius-Bau and presented by the German-Russian Museum Berlin-Karlshorst in cooperation with the State Historical Museum of Moscow.
STIMMEN is a participative installation which gives the visitors a voice in the ongoing discussion around german-russian relationships of the past and the present. It was developed for the exhibition “RUSSIA AND GERMANY. FROM CONFRONTATION TO COOPERATION” at Martin Gropius Bau.
The speech is analyzed and visualized in various information graphics that give visitors an oversight of the diversity of opinions and showed coherences and tendencies in the discussion. The spoken words are played back as an atmospheric soundscape, that let the visitors engage with the subject and each other on a personal level.
Through the course of the exhibition, the visualized content was growing into a broad catalog, that was filled by the visitors themselves, creating a platform for discussion and exchange.
Visitors were given a spoken statement or answer to one of five questions about German-Russian relations in politics, sports, art and culture. The speech was then recognized and analyzed by an automated system.
The grammar and semantics of the words were identified by the system, using several analysis tools based on machine learning. The algorithms were customized to recognize special words and phrases relevant to the exhibition, enabling us to visualize the quantity and relationships between relevant politicians, organizations, places. This gave us new insights: Which verbs and adjectives are most commonly used to describe a given topic? Who are the most important actors and which properties are assigned to them? What are the overall tendencies of a certain question?
Additional information was gathered from the internet in real time and integrated into the visualizations to give context to the discussion. For example, all mentioned locations were placed on a map to place the relationships in a global context.
SETUP / TOOLS
The interface was built as a custom web app powered by modern HTML5 features and the speech recognition was realised using Google’s Web Speech API.
After being collected, the speech data was sent over the network to a VVVV application, which handled the database connection, stored and retrieved the data, and served as the central hub to integrate the different text analysis tools like Text Razor and Tree Tagger. All visualizations were generated in real time with VVVV.
GALLERY OF ENCOUTERS
The entrance-installation welcomed visitors in the form of a large wall covered in photographs of iconic moments in German-Russian relations. A projection map highlights each photograph in a logical sequence.
Through this introduction, visitors were given a taste of the visual and audio exhibition that was to follow. The photographs in the entrance hall later became jumping off points for the discussions and topics they would dive into.
Media Concept / Design
Little Adén, Hanna Adén
Design + Programming
Google Web Speech
Creative Direction / Animation
Design of Installation
Valentin von Lindenau
kling klang klong