A systematic approach : MPC service study

 

This study was made in association with the software firm Hyptique, Radio France and France Telecom.
It is part of an online thematic radio project, named 'RadioThem', of which here is the temporary interface.


It is based on 48 samples that were designed to convey the following meanings :

- in message - in mail - out mail - in radio link - out radio link - feedback yes - feedback no - alarm clock -

The in mail and out mail categories : mean "you've got a new mail" and "your mail has been successfully sent"
The feedback yes category : a traditional "feedback degree +2 : positive feedback to an explicit question"
The feedback no category : a traditional "feedback degree -2 : negative feedback to an explicit question"
The alarm clock category : a schedule reminder
The in message category : something close to the in mail category, but meant to be less urgent and/or important

The radio link categories :
These radio links are specific to vocal navigation utilities in a streaming radio context ; they can be explained the following way.

A user is listening to a radio broadcast, and other "audio articles" about the current subject are available on another stream : an out radio link sample warns the user that other streams are available about this subject.

A user is listening to a radio broadcast, and something important, that isn't necesserarily related with the current subject, is suddenly available from another stream - a goal in a soccer match or anything else : an in radio link warns the user.


Once these samples were made, they were proposed to 18 people who were asked specific questions about these samples.

* Phase 1 - the question was :
Here are 48 samples, proposed in random order. You know nothing about them. Put together the samples that seem to convey the same meaning.

* Phase 2 - the question was :
Now that you've made these groups - categories - of samples, describe these groups in your own words.

* Phase 3 - the question was :
OK, here are the 6 categories the sound designers proposed for these samples : in mail - out mail - feedback yes - feedback no - alarm clock - radio link
Now take your own categories, and name them with these words.

Now we take 18 other people and we propose them the same samples, asking different questions.

* Phase 4 - the question was :
Here are 48 samples, and here are 6 categories : in mail - out mail - feedback yes - feedback no - alarm clock - radio link
Please put each sample in one of the categories.
Also, chose in each category the sample that seems to you the more representative of its group ( prototype )

* Phase 5 - the question was :
Choose the samples that you find unpleasant or offensive.


Once these experiments are done, the results are analysed. Here they are :

* phase 1&2 (free categorisation & description)

1. cluster analysis results

This analysis shows which samples were grouped with which.
The more nodes between two samples, the less often these samples were associated together.

From this analysis, we were able to distinguish five classes. Here is a detailed study of the samples in each class :

2. class by class sample properties : class 1 - class 2 - class 3 - class 4 - class 5

By studying the samples in each class, we were able to draw some conclusions - here are these conclusions, summed up :

3. class by class sample properties : conclusion sum-up


* phase 4 (hinted categorisation)

1. cluster analysis results

This analysis shows which samples were more often put in which category.
The more nodes between two samples, the less often these samples were associated together inside this category - for instance, feedback no 1 & 2 were always put in the feedback no category.

Now we take the samples that were more often put in each category, and have a close look at them :

2. category by category sample properties : alarm - radio link - mail in - mail out - feedback no - feedback yes

3. category by category sample properties : conclusion sum-up

Now the prototype choices : let's see which samples were most often taken for being a prototype for each category :

4. prototype choices : alarm - radio link - mail in - mail out - feedback no - feedback yes

5. prototype choices : conclusion sum-up


* phase 5 ( unpleasant samples )

Five samples were found unpleasant by more than 50% of the listeners.
Here are the concerned samples.

The conclusion that can be drawn listening to these samples is that one must not be too heavy on hi-medium frequencies : be smooth !


And now, the general conclusion sum up : all the conclusions put together end ordered.

* Category properties

The terms, properties or profiles that were associated to some categories.

* important aspects of the sound, perceived as such by the listeners.

These aspects were sufficient to create a "perceptive" category by themselves, thus overwriting the "functional" categorisation.

* classification changes

Conclusions that lead us to modify the classification system proposed in chapter IV.

 

back