Alexa ask SAGE


We wanted to see whether it was difficult to build a simple tool for academics that integrates with Amazon's cloud-based voice service, Alexa. We thought about social science terms and concepts, and tried to teach Alexa some definitions from SAGE Publishing and use these to answer questions when prompted. 


We built a simple tool with Amazon Alexa that listened to the phrase 'ask SAGE to define [insert term]', then retrieved the definition from the SAGE Research Methods ontology and read it out to the user. We used the standard Amazon Development Console. 

We needed to do two things - one is to teach Alexa to respond to 'ask SAGE', and two—show Alexa where SAGE keeps the answers. We built part one (or the request-response workflow) in the Amazon Developer Console. The request was then processed in an AWS Lambda function in Python, leading Alexa to the database of knowledge (i.e. the definitions from SAGE). We used a database called DynamoDB to store these. The set up was such that when you ask Alexa to talk to SAGE, it would look for answers in this non-exhaustive database, as a proof of concept.

'Alexa ask SAGE to define [insert term from social science]' is now live and you can try it out by following this linkPlease note, it is currently supported in UK English only.


Amazon's Alexa provides a simple, quick and effective way of creating skills, via the Development Console. A big advantage is that Alexa has built-in synonym disambiguation, making it easy to handle any vague utterances. 

We learned that all the voice-driven interfaces, as they stand today, are best suited to problems that arise when your hands are full, and for conversational interactions rather than simple queries. 

The big positive, is that you can create your own Alexa Skill definitions relatively easily and on your local machine in a JSON format. This can be really helpful, especially if you have a lot of phrases and terms to register. For anyone interested in trying this out, Amazon has some guidance on building an Alexa Skill in under 5 minutes and the code for this experiment is available on GitHub.


Ian Mulvany, Alan Maloney, Zoe Webster, Razvan Telitoiu