The stakes are large in the Natural Language Processing (NLP) market: It’s the high ground in the battle for control of the data economy and the key to turning silicon into gold, according to a report issued this quarter from market intelligence firm Tractica.
Report authors Bruce Daley, the Principal Analyst, and Clint Wheelock, the Managing Director, cite as an example the upcoming Facebook virtual digital assistant Moneypenny. Noting that Facebook already turns an average daily revenue of $11.96 per active user, they surmise that if the same number of active daily users adopt Moneypenny and generate just an additional $1 in additional ad revenue, the program would add $1 billion to the company’s annual top line and almost as much to its bottom line.
With these and other tantalizing economic prospects at hand, is it any wonder that big names from Amazon to Apple and from Google to Microsoft – not to mention IBM, Nuance, AT&T, and others – are driving further research into NLP and acquiring companies specializing in the space? This month saw another Natural Language Processing acquisition take place, as well-known CRM vendor SugarCRM purchased Contastic’s NLP technology. The platform analyzes communications in emails, LinkedIn, and other sources between salespeople and their contacts to keep their relationships on track, helping the former send the latter appropriate content. “SugarCRM will use Contastic’s NLP technology to analyze data within the Sugar platform so users can automatically send personalized content [to customers],” CEO Larry Augustin said in a statement about the acquisition.
Businesses with NLP in Their Sights
Natural Language Processing, the Tractica report notes, is arguably the most leveraged technology in Artificial Intelligence (AI), and that it is improving rapidly thanks to advances in related technologies, such as Deep Learning and Cognitive Computing. It is already providing a competitive advantage to businesses in the fields of digital ad services, legal, and media. Other areas – automotive, healthcare, education and retail – are likely to become invested in the technology as new business models in those sectors take shape and as – or if – NLP continues to evolve “to correctly interpret and adapt to the wide variety of human language and become engaging in the process” as virtual digital assistants of one type or another.
The report’s authors envision that the day may come when “applications that are both completely personalized and yet generic to produce” will result in shoppers turning to Amazon Alexa to answer product and inventory questions; individuals reaching out to Microsoft Cortana provides to gain investment advice; students being tutored by Apple Siri or schizophrenia being diagnosed by IBM Watson.
Indeed, a report published by MarketsandMarkets last year positions the global healthcare and life sciences NLP market to see a CAGR of close to 20 percent between 2015 and 2020. Evidence of traction this year alone in the healthcare area, for example, comes by way of eviCore healthcare’s acquisition of QPID Health. It is technology that includes NLP, clinical logic, and Machine Learning for generating patient facts from information found in any records in any format – including unstructured notes – to improve understanding of a patient’s history. IBM Watson Health also took another step forward with the announcement it’s acquiring Truven Health Analytics to add to its data set on patients and health that Watson can ingest to help brings medical insights to physicians.
A Take on the NLP Movement
Tractica forecasts that annual revenue of NLP software bought by enterprises for their own internal applications will increase from less than $30 million worldwide in 2015 to over $200 million in 2024, and that total NLP hardware, software, and services will amount to $2.1 billion by 2024.
DATAVERISTY® (DV) had an opportunity to discuss trends in the NLP market and the value the technology is producing with Tractica analyst Bruce Daley:
DV: Could you point to the top three outputs of the Natural Language Processing research that has been undertaken by big brand companies like Apple, Google, and others?
Tractica: It would be hard to point to three since the goal of most research has been incremental improvement in some narrow aspect of NLP. The progress has been significant enough to convince executives at some very tech-savvy companies to place enormous bets on developing NLP that interacts with people well enough to allow them to form social bonds.
The big step forward in NLP, if it happens, will be devices that you don’t type on, but talk to. They will still make mistakes, just as other people do in your social circle, but you will share with them your thoughts, needs, desires, etc. If this happens, it will create a whole new interface on top of the one we have now and will require an extensive business ecosystem to maintain it. Amazon, for example, is offering $100 million to developers to build on its Amazon Echo/Alexa platform.
DV: Discussions about NLP disrupting business models aren’t new, but this report does seem to indicate that those disruptions will affect not just hourly wage workers but the professional classes. What are the potential consequences and are the industries and the professionals in it cognizant of that?
Tractica: Exactly. Education [for example] is based on a business model developed in the Middle Ages, when information was so scarce and expensive to transfer that you would travel many miles to hear wisdom from the mouths of the learned.
In the future when college students have a question, they may not ask the professor. They may ask their cell phone instead because the phone will presumably do a better job of answering the question for that individual student.
Having mass tutoring will certainly change the role of teachers and professors, although I am not sure it will entirely eliminate it.
[Editor’s note: The same may be said of other professions, as least for now. For example, in this interview about IBM Watson Health’s acquisition of Truven given by Anil Jain, VP of Watson Health, Jain explained that doctors will use Watson to gather data about patients with similar symptoms to an individual they are treating; they’ll review the patient’s condition in the context of related information on everything from procedures to outcomes to medical journal data to better understand their own patient’s problem in order to themselves offer improved diagnoses and treatment.]
DV: Do the positives of growing NLP influence in the world and the digital economy outweigh those negatives?
Tractica: Yes, the positives will outweigh the negatives, but it won’t be an unalloyed benefit. When the car replaced the horse, there were lots of advantages, and some disadvantages like rush hour and highway fatalities. It will be the same this time, too.
DV: The report also discusses that there is work left to do in NLP to interpret and adapt to human language. What’s the big challenge there?
Tractica: One of the biggest challenges is the need for clean and accurate data. That is such a challenge that I wrote a book, Where Data is Wealth, to examine some of these issues in detail.
DV: Deep Learning models are now commercially practical to be employed in driving NLP that adapts to new conditions, we’re told in the report. How does Deep Learning provide this benefit?
Tractica: Deep Learning is an enabling technology. It does not necessarily do anything that can’t be done in other ways; it just allows it to be done faster, more accurately, and with fewer lines of code.
DV: As you allude to, we’ve seen the AI winter before and there are risks that if NLP doesn’t meet expectations this time around by surmounting its challenges, we may see one again. How big a risk is that?
Tractica: Being old enough to have lived through a couple of AI winters myself already, I would say the risk that it will stop new research entirely is low, but the risk that AI will not meet expectations is very high. That said, it seems to be good enough to be used profitably in such commercial enterprises, as Google, Bridgewater, and Apple are demonstrating. The trick is in picking the right application to use it with. It’s not going to work well with everything.