Neural MT and Blockchain Are About to Kill the Traditional Translation Market
Neural Machine Translation (NMT) systems are changing the world right now! Unlike machine translation technologies of the past, NMT is providing high quality translation and is improving fast, very fast.
Continuing at the current rate, NMT will drastically change the traditional, human based, translation industry in as little as 1–3 years, impacting 600,000 linguists and 21,000+ LSPs.
For customers, the age of NMT will lead to massive price reduction, and substantial speed improvements (even when humans are needed).
To ride the NMT wave, LSPs and businesses should be able to handle automatically a complex process involving simultaneous projects, high load, quality control etc. OHT and companies like Booking.com, are already doing that already using a hybrid (NMT+human) translation approach.
Interestingly, Blockchain, another fast-growing technology can assist NMT in reaching market domination.
Why Neural Networks (NN) matter
Neural based, deep learning, artificial intelligence systems are changing the world, it is happening right now and they expand quickly from one field to another.
Neural Networks (NN), the basic technology driving these systems, is enabling image recognition, autonomous driving, digital personal assistants, x-ray analysis, voice recognition, Go and Chess mastery and more.
A lot has been written about the field, in brief, NN are imitating the way the human brain is built and how it learns new skills/ information. From a technical perspective an AI system can be viewed as several matrices, layered one on top of the other. The cells in each matrix are representing neurons, which are connected to other cells in the next matrix, with varying degrees of “strength”.
The top matrix is fed some input, eg. the pixels of an image. The input “trickles” down to the bottom matrix, layer by layer, via the connections between the cells, and the output from the bottom matrix can be whatever the system was trained to produce, e.g. does the image contain a car or not.
Traditional information systems like SAP, Salesforce, Facebook and even the OHT platform, have millions of lines of code each! They take years to develop, and are subject to endless debugging and improvements so they can continue to function properly.
One of the amazing things about neural networks is that the software required to run them, is relatively simple, few thousands of lines of code, compared to the very long and complex code behind the traditional IT systems. This is a very important aspect of that technology, because making an NN based system better or “smarter” is much easier than improving traditional software.
What makes a Neural Network “smart” and useful is mainly the data it is “trained” with, and not the complexity of the software itself.
So unlike conventional software where the smarts is mostly in the software itself, with NNs, the value and the desired function of the system comes mainly from the nature, quality and quantity of the data the system is “trained” with. NN training is a relatively simple process where input data is fed to the system, and the result is examined vs. the desired result, a simple process adjusts the connections between the cells to get the result closer to the desired one. Doing this hundreds of thousands or millions of times produce a NN that can handle its designated task well. Clearly, the accuracy of the feedback provided to the NN is crucial for its training.
The exact same NN can be useless before it is trained,
and extremely valuable afterwards.
Why is this technical mumbo-jumbo important? Because it has one simple meaning. Making NN systems better is easy. All is needed is more compute power, and more quality input. Unlike traditional software, they do not require smart architects and product designers to think of new ways to improve the software. Nor they require brilliant engineers to work for years to make the software better and better.
Making NN systems better is relatively easy.
All is needed is more compute power, and more quality input.
Technology and Translation
In the past 50 years or so compute power has been growing exponentially. Gordon Moore (co-founder of Intel) predicted already in 1965, that the number of transistors on a given piece of silicon will double every 2 years (Moore’s law).
To better understand exponential growth, consider the following, doubling the length of a 1 meter stick 25 times, will make it +33,000Km long(!!), almost 3 times the diameter of the earth. Continuing to double it 25 times more (ie. 50 times in total), will make it over 1 Trillion and 125 Billion kilometers long!! For comparison the radius of the solar system (average distance between the Sun and Pluto) is “just” 5.9 Billion Km!.
Applying this to calculations, a computer that could run just 100 calculation per second 50 years ago, is able to process over 3.3 Billion (!) calculations per second today, and will do over 112,500 Trillion (!) calculations a second in 50 years (assuming the current rate of improvements continue), ie. 112,500,000,000,000,000 calculations per second!
This improvement in compute power is happening very fast, and, with time the performance increase is more and more substantial (e.g. doubling 10 is 20, ie. 10 more, doubling 20 is 40, which is 20 more and so forth). These improvement are happening regardless of what the compute power is used for.
The other important factor in making NN better is the quantity and the quality of the data. Consider the fact that 90%(!!) of the online material today (images, text in various languages etc.) is less than 2 years old. Ie more and more content is generated all the time.
More and more quality data is available online for NMT training.
How does this apply to translation? In the past 2–3 years several companies started using NN for translation. The results of these efforts are stunning, over this short period, the quality of NMT in some areas became human like, passing swiftly all previous translation technologies.
Moreover, NMT systems continue to improve fast, thanks to:
- Increase in compute power because of:
a. Exponential compute power improvements as described above
b. More compute power allocated to these systems as they demonstrate great results - More training material coming from
a. Crawling the web — source and translations available online
b. Translation memories created by business customers
c. Proactive translations done for training purposes - Improving NMT due to the ability to provide human feedback on a massive scale, e.g.
a. OHT is already running NMT rating and feedback projects (over million projects to date) for NMT vendors / users.
b. Facebook and Google encourage users to provide translation feedback etc.
The important thing to understand is the rate of change. Unlike past software technologies that improve slowly because traditional software improvements depend on human developers, thus slow. With NMT, the core technology is already working. Improvements happen quickly because compute power and input data are easy to add.
The Revolution is here!
In simple terms, NMT is a tsunami (!) approaching quickly, it is not “yet another” wave of technology advance. I expect that within 1 to 3 years, 30% to 50%+ of all translations in the world will be done using NMT! (with potentially some level of human intervention).
For comparison, the world’s biggest LSP, TransPerfect, has less than 2%(!) of the $40B+ global translation market. This means that NMT (with post-editing, quality control etc) has the potential to totally disrupt the market.
NMT has direct impact on over 600,000 linguists and over 21,000 translation agencies. Those who manage to leverage the technology will survive and the others will have difficult times. The same is already happening with drivers because of autonomous cars/ trucks (Tesla etc) and with office workers that are replaced by NN based automation (UiPath).
To ride the NMT wave (and avoid drowning), LSPs and major businesses that use NMT systems, should be able to handle a complex translation process. The NMT engine is not sufficient for real business use. Using an automotive analogy, NMT is like the car’s engine, while a business solution, like OHT’s Hybrid translation service, is the entire vehicle.
To work properly, a hybrid (NMT+human) translation service should handle hundreds of thousands of projects simultaneously, dynamically select the right NMT, decide what human intervention is needed and where, and make sure quality control is done properly and smoothly. This is a complex, multi-step process, alas, without it NMT engines are just like real car engines, heavy blocks of metal that are not very useful on their own.
NMT engines are just like car engines — not very useful without the rest of the car. Business customers need the entire vehicle,
ie. NMT engine + procedures around it, to benefit from NMT.
OHT made a substantial shift of focus to become the first hybrid translation agency. Using our Hybrid translation system (NMT+Humans), we are already providing high quality, low cost translations to business customers.
We recently released ONEs — the first independent, human based NMT evaluation score (https://slator.com/sponsored-content/make-neural-machine-translation-better-faster-a-new-way-to-measure-nmt-quality/ )
using ONES we are able to select “on the fly”, per project, the NMT engine that will provide the best results for the specific project, and will require the least human intervention.
We are also working with few of the largest NMT vendors to train their systems by providing high volumes of human translation, rating and evaluation of NMT results, human feedback and corrections etc. More importantly, we do that in 100 (!) languages, as one of the main issues with NMTs is having enough material for training in languages other than English.
I estimate that by the end of this year (2018) 80%(!) of our general translations will be done using NMT, providing human quality at a much lower price and much higher speed than traditional translation.
Bottom-line, NMT is already here, it is happening fast, very fast, and more and more business customers are beginning to enjoy the NMT benefits.
Where does Blockchain come in?
There is a huge pool of existing business translations that are saved as Translation Memories (TM). These TMs were built over many years and contain tons of data. The data is saved in a way that is perfect for NMT training. Using these TMs, NMT training can leap forward, producing engines which are perfect for business customers. Furthermore, the investment in these TMs was already made, so any future income they generate for their owners is pure profit.
So what is missing? To make these TMs available there is a need for an easily accessible repository or some sort of a marketplace. The minimal requirements are to make it easy to search the source while making sure the translation is delivered only upon payment, and to eliminate the need to trust a central vendor that will manage it all. Most users (businesses or customers) will not upload their TMs to some central system due to trust issues.
A blockchain based system can be an ideal solution! Using a blockchain based architecture, it is possible for Translation-Memory (TM) owners to earn some income without risking the TM being stolen. TM owners like companies, translators etc., will be able to share their TM in a way that will only expose it once a payment was made. Customers will be able to select the translation they want based on metadata (like rating), and translators will be able to upload their translations for common phrases, and get paid everytime someone uses them.
The financial incentive for such system is clear, powerful and already exist. Many of our customers will gladly recoup some of their past translation cost by selling their TMs.
Once such a blockchain based system is available,
NMT builders will be able to purchase the training data they need, in many languages, and make their NMTs better at business translations, instantly!
OHT is in the process of building such blockchain, and I will elaborate more about that in a separate post.
