Gartner Blog Network


Eight Factors Improve VPA Relevance

by Werner Goertz  |  October 24, 2017  |  Submit a Comment

Think that the bigger the knowledge base the happier the user? Think again! Why does my Virtual Personal Assistant (VPA) need to know Einstein’s shoe size or the number of atoms in the universe, when all I want to do is book an airline ticket? Here are eight factors that will make you either a happy camper or a dissatisfied VPA user:

modeling a user profile based on predictive analytics and machine learning (ML):
a VPA builds a “digital twin” of me in the Cloud and refines it with each interaction. (This type of training is exactly what my dog does to perfectly train me! Each time he observes reinforcable behavior, he rewards me with a lick). A positive outcome occurs when the model knows me well enough to deliver the right outcomes.

context and threaded conversations: well, that’s still nascent and pending because VPAs today have the attention span of a 3-year-old. They’re not good (yet) at “holding state” (remembering what I said two sentences ago), but if a VPA could improve that, it would lead to better outcomes.

video (YouTube), and personal photos: there’s an incredible amount of cultural, contextual and personal knowledge to be gleaned from visuals! Image recognition (or “Rekognition” as some industry players know it!) will deliver a source of personalization that can be turned into relevant outcomes!

text crawlers: capturing written knowledge from libraries of web sites, databases and publicly available repositories (see what folks like www.bitext.com are doing) is a great source of further context.

bots and specialist VPAs: the major VPAs (Alexa, Google Assistant, Cortana, Bixby, Siri,…) can’t possibly know all the domain-relevant details or geography-specific insights (hey, I was raised in Germany: I know more about local football than Alexa ever will!). So, they will invoke specialist VPAs to hear about what is relevant in your neck of the woods (like what Sherpa does: see www.sher.pa)

locational input, mapping: knowing where you are now, where you are headed and what you are doing is crucial to coming back with a relevant response. Getting hints from your GPS and knowing the surrounding map will make VPA responses more relevant.

emotion recognition: “hey, Bixby – don’t bore me with details! Can’t you tell I am angry? Tell me what’s relevant and we can have a longer chat when I feel better – ya feelin’ me?” Emotion detection is being used in e-commerce and call centers today. Why not use it to tailor the VPA experience to the mood of the speaker?

knowledge base and traditional search: ahhh … finally the point you’ve been waiting for. I am NOT saying that knowledge graph isn’t relevant. I am saying that it is one of several factors that lead to positive VPA experiences.

There you have it: your VPA experiences are still lacking because the Amazons, Googles, Apples et al. don’t have all their ducks in a row yet. The sooner they do, the quicker the VPA adoption, in connected home, enterprise, automotive, you name it…

Category: 

Werner Goertz
Research Director
1 years at Gartner
21 years IT Industry

Werner Goertz is a Research Director within Gartner's Personal Technnologies team, where he covers personal devices (smartphones, PCs/Ultrabooks, tablets/ultramobiles and wearables) and IoT. A special emphasis of his research lies in the Human Machine Interface (HMI) and multimodal I/O technologies: voice/speech processing and recognition, facial recognition and eye tracking, biometrics and motion/gesture control. Read Full Bio




Leave a Reply

Your email address will not be published. Required fields are marked *

Comments or opinions expressed on this blog are those of the individual contributors only, and do not necessarily represent the views of Gartner, Inc. or its management. Readers may copy and redistribute blog postings on other blogs, or otherwise for private, non-commercial or journalistic purposes, with attribution to Gartner. This content may not be used for any other purposes in any other formats or media. The content on this blog is provided on an "as-is" basis. Gartner shall not be liable for any damages whatsoever arising out of the content or use of this blog.