Voice First Systems Do Not Need To Answer Every Question. These 100 Are The Most Important.


What Was The Question? Let Me Google That.

The most popular functional Google search terms (search terms that are not about companies, people or events) show a deep insight about the most popular things people look for on a recurring basis.  This dataset can have a great effect on understanding the needs for future Voice First app (or Alexa Skills).  I have used the search term list for almost two decades to aid in companies I have built and in advising hundreds of startup and legacy companies.

Over the last two years I have held open “Office Hours” with over 300 Voice First companies/founders.  I use the popular search engine keyword lists to aid in showing just how people interact with Q & A technology.  I also use questions and answers presented at Quora for over 6 years as a benchmark on what people are asking.  Additional I use a number of other tools that I hope to share in a future issue of Multiplex Magazine (subscribe!).

I spoke to the I call the log2(n) – n paradox (or the Evan’s paradox) in a recent posting [1].  The premise of the paradox is the log2(n) – n number of potential questions that a Voice First system would need to answer and the log2(n) – n of potential answers.  This centers around intent extraction after natural language recognition has decode the actual words. I presented one of many novel ways to deal with a potential infinity.   To ad to the solutions is to some degree “know” the probability of the top of questions statically and Google has that information for free.





This data also informs the basis for an entire startup to be formed to deal with the “short tail” and “long tail” requests.

What keywords on Google are searched the most?

I have removed branded terms like “Gmail” and “Craigslist” that are frequently searched, and are a common point of reference for users. The graphic list is of the 100 most searched for non-branded keywords on Google, in the United States region. The list (with help from Ross Hudgens) was determined by manually sifting through the most popular search terms overall to find keywords that were not associated with a brand. In addition porn-related keywords (a big number) were removed for many reasons. Thus removed from the top five was:

  1. Gmail
  2. Craigslist
  3. Amazon
  4. Yahoo
  5. Porn

SM-Most-Popular-Keywords (2)We can see that at first look there are some big surprises. The top 20 represent some new and under represented Voice Apps and Skills.  Clearly there are some questions that does not lend itself to a Voice First system, however a majority does.  I can say this with clear certainty, one or more startups will find a way to go after just one of these search terms and build a multi-billion dollar business around it in a Voice First setting.

Thus when we are faced with the quagmire of the log2(n) – n paradox, perhaps if we solve the top 5,000-25,000 typical questions to a deep degree, the statically insignificant questions can be answered by one of the solutions I have presented in the Special Edition 1 Issue of Multiplex Magazine.  And if that is not good enough there are other ways. Stay tuned for more in future posting and issues.


[1] http://readmultiplex.com/2017/03/12/voice-first-cant-possibly-answer-every-question/






Any reproduction, copying, or redistribution, in whole or in part, is prohibited without written permission from the publisher. Information contained herein is obtained from sources believed to be reliable, but its accuracy cannot be guaranteed. We are not financial advisors nor do we give personalized financial advice. The opinions expressed herein are those of the publisher and are subject to change without notice. It may become outdated and there is no obligation to update any such information. Recommendations should be made only after consulting with your advisor and only after reviewing the prospectus or financial statements of the any company in question. You shouldn’t make any decision based solely on what you read here.