akoúste

akoúste

Client-side listening experiments in the browser

The akoúste project hosts a number of simple, adaptable listening experiments aimed at situations where the use of a dynamic host server storing and serving the speech and collecting the responses over the internet is unpractical or unwanted, e.g., due to privacy concerns. Akoúste pages can be used both on desk/laptops and on mobile devices.

Akoúste experiments can as easily be run off thumbdrives, or local hard drives, with all data kept locally, as they can be run from a web server. Submission of results is under the control of the subject doing the experiment. The whole experiment is run inside the browser and intermediate results are stored in the browser's local storage. After completing the experiment, the subject can download the results in text format which can be submitted to the experimenter.

List of example experiments
ABXIdentify an unknown stimulus, X, as one of two choices, A or B
VASRating a stimulus on one or more computerized Visual Analogue Scales
Likert 5 pointRating a stimulus on one or more 5 point Likert scales
Likert 3 pointRating a stimulus on one or more 3 point Likert scales
AB experimentSelect best out of two (this is an adapted settings of a Likert 3 experiment)

Example experiments are derived from:
Audio: Ribeiro, Manuel Sam. (2018). Parallel Audiobook Corpus, [dataset]. University of Edinburgh. School of Informatics. https://doi.org/10.7488/ds/2468
Pseudonymization: Pseudonymize Speech

Inspect and process results of akoúste listening experiments

Load the results of a listening experiment into a special page for inspection and processing to a "database friendly" format.

Inspect & processthe answers

Under the hood

The akoúste experiments are constructed from self-contained plain HTML, CSS, and Javascript pages. Each experiment is controlled by a settings, or initialisation, file and a file listing the stimuli. All user visible texts are listed in the settings file by language. Adding new languages and translations amounts to copying the list and translating the items. User language detection can be automatic or fixed by the experimenter.

akoúste on Github

Loading local CSV table files is not always possible. Therefore, stimulus lists must be present as Javascript, .js files. There is a special 'page' that will convert CSV text tables into the required form.

ConvertConvert CSV tables into akoúste JSON stimulus lists

It is possible to construct a Read_<type>_Stimuluslist.js file containing several stimulus lists which are selected by adding: ?expnum=<n> to the URL (n is the list number counting from 1!). In the Read_<type>_Stimuluslist.js file replace var stimulusTableJSON = "<list>"; in the first line with:

		var stimulusTableJSON = new Array();
		stimulusTableJSON.push("<list>");
	

Add a new line with stimulusTableJSON.push("<list>"); for every stimulus list.


akoúste is licensed under the AGPL v3
akoúste icon Ear by Gregor Cresnar from the Noun Project