Contextually Intelligent NLP Assistants – AI’s Next Big Technical Challenge

Author: William Vorhies

Summary:  Contextually intelligent, NLP-based interactive assistants are one of the next big things for AI/ML.  The tech is already here from recommendation engines.  The need to be more efficient and to become AI-augmented in our decision making is now.  Getting the contextual awareness is the hard part.

 

Last week we took the position that from a technical standpoint, ‘deeply inclusive and contextually sensitive’ AI is one of the two ‘next big things’ in AI.

In retrospect I wish there were a more concise agreed naming convention for this bit of technical legerdemain.  “Inclusive” and “contextually sensitive” are in the category of those ‘suitcase words’ Marvin Minsky called out as being so dependent on the user’s experience that agreement on meaning is difficult.

What we’re not talking about is the ability of NLP to hold a contextually appropriate conversation, such as making a reasonable response or request for clarification based on the topic at hand.  For the most part, short of performing psychoanalysis, chatbots can do pretty well with human ad hoc conversation.

Also, we’re not talking about being culturally inclusive as in detecting and eliminating bias.  Important, but not what we’re getting at.

What we’re describing is the next big step in NLP utility in which the NLP puts together facts it knows about us and proactively takes action or makes suggestions that make our life easier.

The example we gave in our previous article is about having the NLP assistant remind me of my mother’s upcoming birthday in a week or so without my having explicitly created a reminder.  More importantly my NLP assistant could make a recommendation for a present.  Presumably my past communications with her both in fact and tone contain some strong signals about my mom’s demographics and perhaps even her interests so why not predict a short list of appropriate gifts.  Now that would be valuable.

So perhaps a better description of this behavior then would be ‘contextually intelligent’.  We’ll stick with that.

 

Helping Make Decisions

The base technology for this advancement is the already well developed field of recommendation engines.  Thus far these have been the bread and butter of ecommerce whether recommending books, airplane flights, or love interests.  What is coming is the expansion of this tech from predicting things you might like, to actions you might take and then helping you make that decision.

One element of this problem is that we have to add information to make these more sophisticated recommendations.  The close-in sources are our calendars, email, and texts with other sources added as the field develops.  As it happens, calendar and email-aware intelligent assistants are in early research and development, making this a lead candidate for our next break through.

But beyond adding information sources, the challenge is how to integrate this into a useful tool and that requires a deeper understanding of how people make decisions.  For example, your intelligent assistant may be able to predict what your next action could be, but how comfortable will you be if the IA simply says ‘now do this’.

Our decision making is typically built on a filter of what’s possible.  That is we look at the options (like different airline flights) and understanding those options we make a decision.  That filtering of possibilities is a major challenge.  Present too much information about the how the recommended action was made and we bury the user in the paradox of too many options.  Present little or no underlying decision data and the user’s comfort and willingness to accept the action plummets.

 

Defining Context

This ability to help make decisions is clearly going to start small before it expands, and in terms of email and calendars the contextual information that can be extracted is most likely to be about what will happen, when will it happen, and which people will be involved.

A first step, knowing these three things, might be to figure out what you’re going to need in advance of that event to help you prepare and participate.  This begins to look like the behavior a really great personal assistant might display in helping you get ready without your really having to tell them explicitly what you need.

We all attend a lot of meetings and it’s no stretch to say that the content and goal of each calendar event is not on our top of mind.  Wouldn’t it be great if your IA could assemble the documents and information you need, anticipate how to manage your time to get ready, or even provide you with a list of information you might need on-the-instant during the meeting.  It might also provide you with a synopsis of previous meetings leading up to this one.

A more advance scenario might be where you are managing several parallel projects.  Since you are the common thread, it’s likely that these project schedules are interlinked at some point in the future that defines the critical path and might block one project from proceeding until other elements in a separate project are complete.  A contextually intelligent IA could be trained to foresee those scheduling conflicts even months ahead of time and alert you.

 

Efficiency First – Augmented Decision Making Follows

At the outset of contextually intelligent assistants, the most likely goal is to simply be more efficient with our time, to be able to focus on what’s important and let the IA coordinate the details.

But the magic may be in the way the IA can present information.  Once the IA has achieved a contextual understanding of your goals and actions, the next thing you’ll need is information.  Eventually IAs should be able to reach beyond the information you already have captured in your various devices and through communication to recommend other sources.

Think of a young programmer who brings a certain set of skills and knowledge with them.  Chances are this includes web sites or specific libraries or notebooks the person is used to relying on.  The future intelligent IA might propose solutions from different sources, accelerating the learning curve and success of that beginner.

 

Remembering What You Forgot

Beyond simple calendar reminders you might create, your advanced IA might detect when we’ve forgotten something.  For example, your wife emails that you should make restaurant reservations and arrange for a babysitter for your night out next Friday.  You make the restaurant reservations but get distracted by the next task at hand and forget about the babysitter.  Your IA might reasonably detect the omission of this action and remind you.  Disaster averted.

 

What About Privacy

This will be an important concern for next gen IAs.  Data gathering is likely to go well beyond your calendar and email to include even your eye movements, your cursor movements, and particularly outside data sources. 

For example in tuning models to discover words that are important to you, the model needs to train on your data but needs to be constrained so that it doesn’t give something away when generating communication.  If it saw that discussing your new job at a competitor was suddenly a common theme in your email, that’s not something you’d want your IA showing to anyone.  As with so many things AI, privacy will be an ongoing challenge.

The case is strong that contextually intelligent NLP interactive assistants can be the ‘next big thing’ in AI/ML.  The tech is here.  The development is underway.  Soon we may all be able to have our own super prescient Radar O’Reilly, or Tony Stark’s digital assistant Jarvis at our own beck and call.

 

 

Other articles by Bill Vorhies

 

About the author:  Bill is Contributing Editor for Data Science Central.  Bill is also President & Chief Data Scientist at Data-Magnum and has practiced as a data scientist since 2001.  His articles have been read more than 2 million times.

He can be reached at:

Bill@DataScienceCentral.com or Bill@Data-Magnum.com

Go to Source