ALEX has always had speech up his sleeve, after all, he is a patient communication simulator, right?!

Apr 17, 2019 2:23:11 PM · 2 min read

ALEX has always had speech up his sleeve, after all, he is a patient communication simulator, right?!

Well, we’ve been working up some magic on his conversation. Come join us for a behind the scenes sneak peek into the secrets of the magician.  

ALEX’s speech has been something you could customize for a while now so many of you will know, all you need to do is tell him the questions you’ll ask him and the answer you want to hear, then follow the recipe word by word to have a conversation with your patient.

Recently though ALEX has been doing some heavy learning. With the help of our first magic trick, he’s gone through over 3 billion words of context to understand how they relate to each other, when and how they appear in conversations. All this to learn the concept of synonyms and understand your questions even if they’re not exactly following the recipe under his conversation settings.

You’ll notice he’s not only ready to respond to questions you can ask more flexibly now, but based on his studies – and not to spook you here, but – he may even know what you want before you finish your question. Thanks to our second magic trick he’s become so confident in fact, that he’s pretty much ready to answer you on the fly. So don’t be surprised if his responses arrive faster and more fluidly than before.

Finally, to top it all off, he’s expanded this conversation abilities with new topics and answers to cover more interview questions and to give you more precise answers.

Now to let you in on a bit of the magic. (This is going to get rather technical, so feel free to skip this paragraph at any point. Seriously.) Imagine a multi-dimensional vector space where each word occupies a point closely clustered with other words that are often found in similar context. Additionally, to make word representation more accurate, every word gets weighted based on how important it is in the sentence. When ALEX is asked a question, vectors are formed both from the words of the interviewer and from the many thousands questions ALEX knows. By calculating distances in this sentence-vector space the most similar question can be identified from ALEX’s knowledge base. Mind you, the similarity does not simply rely on the exact wording of the question but tries to capture how close the meaning is to the predefined ones. Amazing, isn’t it?

That’s all we can share in our sneak peek today, as true magicians never reveal all of their secrets.  We hope you still found it intriguing.

ALEX is excited to give you a fuller interview experience going forward, so give it a try and let us know your experiences. We can’t wait to hear your feedback.

Tagged under , Post by