Personal tools
You are here: Home Projects humaine Workshops Summer School 1



Programme and selected slides available here

The first HUMAINE summer school was held at Queen’s University Belfast from 31st August to 3rd September 2004 on the topic of Data and Databases. The programme was organised by Ellen Douglas-Cowie (leader of Workpackage 5 on Data and Databases) and Roddy Cowie (HUMAINE Co-Ordinator) with input from the Project Steering Board. There were 42 attendees representing 17 of the partner institutions in HUMAINE. There were also 4 representatives from other institutions including representation from the AMI 6th Framework NOE.


The summer school was designed to give participants a
hands-on feel for the range of techniques that might be used to address the aims of HUMAINE both in data collection/induction, and in data measurement/labelling. It departed from the traditional image of summer schools in one important way. It was not a dedicated training school aimed at mastery of a particular technique: rather it introduced participants to a range of techniques that might be relevant in the very new field of multimodal emotion databases, and invited constructive discussion and interaction on the issues and problems raised by the techniques. The techniques introduced came from a range of partners in the network – Queen’s Belfast, FAU Erlangen, LIMSI, ICCS-NTUA, TEL AVIV, UNIGE-GENEVA, KTH, DFKI and BARI.


The Co-ordinator of the network (Roddy Cowie) opened proceedings by setting the summer school in the context of HUMAINE. He reminded participants of the kind of database work that was relevant to HUMAINE. HUMAINE is concerned with emotion in a broad sense and with emotion as a factor in action and interaction, not emotions as isolated episodes. He suggested that emotion in the broad sense included both bounded emotions (specific kinds of mental state where a strong rush of feeling dominates a person’s awareness for a relatively short time) and pervasive emotion (a quality that ‘colours’ almost any mental state to a greater or lesser extent and therefore influences a wide range of actions & interactions). HUMAINE is committed to address both –and therefore, so is its work on databases. Within this context he suggested that database work would have two roles –

(i)                  Supportive  - where the role of databases is to let people fill in or check details of processes they understand in general terms,

(ii)                Provocative – where the role of databases is to collect examples that help people to expand and restructure their thinking about the area.

To understand pervasive emotion in action & interaction, certainly the database strand would have to play a provocative role.


To encourage participants to think about emotion in the broad and pervasive sense, he asked participants to respond to a list of words (based on Cowie et al 1999 and extended in the light of Scherer 2004 – see Deliverable D5c for further information) describing  emotions and emotion-related states. Participants were asked to rate the emotion words in terms of their relevance to emotion sensitive interfaces of the future (in the year 2054). The results of the exercise can be viewed here.


The core part of the summer school was two days of hands on sessions in which different relevant techniques were introduced. Many hands on sessions were led by two presenters (sometimes from different labs), and participants attended these in small groups. These sessions were preceded by introductions which set the practical sessions in context.


The first set of sessions was broadly concerned with data collection/induction and the nature of the data collected. Sessions covered the Belfast Naturalistic Database, the current work at LIMSI on a television database, the ERLANGEN AIBO robot database, the SAL (Sensitive Artificial Listener) data induction technique at Belfast, the Bad Mouth data induction technique at KTH, and a range of induction techniques used in Geneva. Core themes that emerged were the problems of assigning emotional labels to naturalistic data (much of the data showed mixed emotion or fluctuated over time) and the problems involved in inducing emotion in a way that was a satisfactory compromise between acting and totally naturalistic. The use of ANVIL for annotation purposes was also demonstrated.


The second set of sessions was focused on tools for annotation and measurement. Sessions covered FEELTRACE, ASSESS, Physiological measurement, Mark up languages, the STISIM driving simulator (as a tool for studying emotion in action and interaction) and FACS. Some of these sessions showed how the tools could be used to go some way towards dealing with the annotation and measurement of naturalistic data.


The final session of the summer school gave a practical introduction to ethical issues in data collection and storage. Roddy Cowie summarised the underlying issues and produced sample procedures and forms for ethical committees.


The closing session of the summer school included an evaluation questionnaire. The feedback was excellent as the attached summary indicates. There was discussion of the timetable for the next deliverable for WP5 on Data and Databases and on the agenda for the December Workshop in Belfast on the same theme. It was noted that there had been excellent interaction among the participants at the summer school and that interesting ideas for collaborative mini projects had emerged.







Document Actions
Powered by Plone

Portal usage statistics