Amazon's Alexa is getting a bit of help from the public to answer questions that the smart digital assistant cannot answer.
Alexa Asks The Internet For Help
On Thursday, Sept. 12, the e-commerce company opened Alexa Answers to everyone in the United States. The crowdsourcing platform works like Yahoo Answers and Quora; users can submit answers to questions posted on the website except, in this case, people are answering questions that Alexa does not have an answer to.
To sign up to Alexa Answers, people are asked to use their Amazon accounts to see a list of questions. They have the option to filter questions by specific topics, most frequently asked, and newest.
An example of a question posted on the crowdsourcing platform is, "what is the state snack of Texas?"
Alexa Answers was first launched last year, but access was per-invite only. According to Amazon, hundreds of thousands of questions answered by contributors have already been used by the digital assistant.
Whenever Alexa pulls an answer from the platform, it mentions that the information was submitted by an Amazon customer.
A Community To Make Alexa A Better Assistant
The trillion-dollar company is placing filters to remove profanity and offensive words from appearing on the platform and being used by the digital assistant. Users can also flag offensive or irrelevant answers as well as upvote/downvote submissions.
In addition, human editors as well as algorithms will be present to police content.
Bill Barton, vice president of Alexa Information, said that the smart digital assistant can answer most questions with data from hundreds of vetted sources. However, for the rest, he trusts Amazon customers who already help one another out by divulging product details that vendors do not disclose in their listings on the online shopping website.
"It's just part of Amazon's DNA to engage our community of users in helping to provide more information for customers who are asking questions," he told Fast Company.
Alexa Answers does not require citations so far, but Barton believes that the voting system will be good enough to identify when the information submitted by a user is accurate or not.
"We're certainly open to revisiting it if we get strong feedback from contributors or from end users if they find more attribution beneficial," he added, "but in general what we find from our community is that they prefer crispness and conciseness in their responses."