An overview of the Machine Translation market & what it can, cannot, and should do

By Sean SangSoo Jun, Product Manager, XL8 Inc.

Sep 7, 2022

In case some of you are not familiar with the history of Machine Translation, the history of Machine Translation (MT) is almost as old as computers themselves. In 1954, Georgetown University and IBM were able to successfully translate more than 60 Russian sentences into English. This is known as the Georgetown Experiment, and it is one of the earliest recorded MT projects. Although researchers at this time were not able to take things further, due to the enormous data processing power and storage it required, we can see that the idea of computers being able to translate human languages is far from being new.

However, recent advances in computer software, data, and hardware make it possible for computers to handle the high complexity involved in language translation. As a result, the MT market is expected to surpass 7.5 billion USD by 2030, as reported in a research study by Global Market Insights [1].

So by now you must be asking what exactly is Machine Translation, and what does that have to do with me? In this post, we aim to explore these questions by covering the current MT market, its demands, and how XL8 is addressing them.

For those who are totally new to this field, MT can be defined as ‘The process of automatically translating content from one language to another without any human input’. However, modern MT goes beyond just simply translating text word for word. It can deliver the full meaning of the source language by analyzing all elements and their relationships within the input text.

If you are thinking ‘Okay I get it. But I’m not a translator and have no plans to be one. So what does this have to do with me?’. Well, MT can go far beyond just translation, and can be applied to many other diverse use cases as well. Here are just a few examples:

If your organization is doing business with countries that operate in an unfamiliar language, MT can help. MT can not only just translate the language in the tickets submitted by your customer, but can also sort and categorize (or sometimes even respond to) them by understanding the context of each submitted ticket.

Your legal department can also utilize MT by using it to translate legal documents into various languages without needing to hire additional translators that are proficient in this domain. With MT a large amount of legal content can quickly become available for analysis in multiple languages that would have taken very long to manually translate and process.

I’m sure a lot of you are already a member of at least one OTT platform, such as Netflix, Disney+, Hulu, and so on. This means you have access to entertainment media generated from every corner of the world. MT can help OTT platforms efficiently localize their media and provide them to users in various formats, whether it's subtitling, captioning, dubbing, and so on.

These are only some of the examples where MT can be used. So basically, if you are currently being overloaded with data that needs to be translated into another language, MT is here to make your life easier.

We are living in a very interesting time where everyone can be a content producer, whether it’s by uploading a video on Youtube, writing an article on Linkedin, or posting something interesting on a blog (just like what I am doing right now). Because of this increase in volume, the MT market is also seeing its golden age where a lot of leading companies are getting into the field.

This is probably the most widely used MT tool out there, and I’m sure you have also used it more than once. Google Translate is one of the forefathers of MT platforms, and it remains strong to this day. However, despite its position as the leading MT platform, it sometimes shows poor accuracy especially when it comes to some Asian languages such as Korean, Chinese, etc.

(Here is an example. ‘Baegjo’ in Korean means swan, but also ‘one trillion’ as well. In the above context, the correct translation will be ‘Once upon a time, there was a swan’ or ‘Long ago, there lived a swan’)

Amazon Translate works under its Amazon Web Services (AWS) cloud platform. It’s said to be more accurate than Google Translate for certain language pairs, like English-Chinese [2], but it is also said that it shows relatively poor performance for certain language pairs such as English-Russian, and English-Spanish [3].

DeepL is relatively unknown compared to the above two. But it’s MT engine is believed to produce more natural translations thanks to its proprietary neural AI [2]. DeepL started with offering translations between English, German, French, Spanish, Italian, Polish and Dutch, and has recently added support for Chinese (Simplified) and Japanese as well.

Papercup is a little different compared to the above examples. Papercup uses MT to automate video translation, and uses AI to produce synthetic voices that sounds life-like and identical to the original speaker’s voice. Besides dubbing and voice-overs, you can also use Papercup’s subtitle feature as an add-on.

Obviously there are a lot more players in the MT market, with each company or product having their own unique set of capabilities. However, even though all of them are extremely helpful in improving translation quality and productivity, it cannot be said they are flawless. In the next section, we will go over them in-depth, and explore how XL8 tackles these flaws and how it aims to improve the overall translation experience.

So by now you can see the increasing popularity that MT is gaining from business of all sorts. And with this increase in popularity, so did the demand for accuracy and use cases. But unfortunately, despite the ongoing advancements and refinements, there are still demands that need to be addressed.

Accurately translating the text ‘word for word’ is somewhat of a must-have for MT technology these days. But being able to deliver the accurate meaning of that text depending on the context, and actually understanding the cultural aspect of the original text, these are the kinds of things that can play as a differentiator.

As we have seen from the above swan example, the importance of being able to understand the context behind the text cannot be stressed enough. So MT AIs must be able to extract, interpret, and process data contextually in real time in order to deliver the full, accurate meaning of the original text.

XL8’s translation technology is unique in this sense as it uses Context Awareness models to better translate colloquial phrases. This allows XL8’s MT engine to fully understand the circumstances behind the original text and therefore allows it to accurately translate with those circumstances in mind. Thereby actually “localizing” content instead of simply translating it “word for word.”

Another thing that makes XL8’s translation technology stand out is that it can apply ‘formalities’ for specific languages. Imagine you want to translate a situation with multiple people with all of them having different relationships between one another. It is highly unlikely for all of these people to be talking to one another using the same formality. For the AI engine to recognize this relationship and apply different forms of formality is also very important in terms of accuracy as it will help the users understand the relationship between these people and therefore being able to accurately understand the overall situation.

As we have seen in the previous section, companies are largely focused on one aspect of the user’s translation experience: Google Translate and Amazon Translate are focused on providing text translation, and Papercup is focused on providing translated voices, and so on, However, in reality, all of these aspects are used to form a one cohesive experience for the user.

This is especially true when it comes to translating media and entertainment content. For example, if you are watching foreign content on Netflix, you can view the translated subtitles in multiple languages as well as dubbed voices for that content in multiple languages as well. So it will become very troublesome for companies if they need to separately purchase platforms for each aspect, and integrate them with one another.

XL8 is unique in this sense, because it can provide a holistic service across all aspects of localization. XL8’s platform can take its users from synchronizing a subtitle file with a media file, to translating that subtitle file to a target language, and dubbing that media file with the subtitle file that has just been translated. Users can finish this entire process in a single flow within XL8’s platform. So coming up with diverse types of localization content has never been easier. Not only this, the user can translate their text just like Google Translate (but not to mention it can get more accurate results) and also instantly create a subtitle for live streaming sites just by typing in that site’s URL.

According to a recent study conducted by Demadsage, Netflix has around 74.5 million subscribers in the US and Canada, while having 33.7 million subscribers in the Asia Pacific region [4]. However, we have seen Asian content making a big splash recently not just in North America, but all over the world as well. And of course, content made in the West has always been a huge hit in Asia (If you don’t believe me, just check out Conan O’brien’s visit to South Korea).

So when translating content there are no ‘intended users’, because you don’t know who will watch and enjoy this content. But instead, it should translate to as many languages as possible. That is why XL8 put huge emphasis on providing languages across as many regions and countries as possible. XL8 currently supports more than 60 language pairs with 20+ source languages, which covers widely used languages like English, Chinese, and Spanish, but also other languages as well.

And also, just providing a translation of a language, especially a major language like English or Spanish, is sometimes not enough. In order for its users to actually ‘enjoy’ the content that is generated from a specific country, the platform needs to translate a variation of a language which is spoken in that country. For example, Narcos was an American television series that was aired on Netflix, which was based on the story of a Columbian drug kingpin Pablo Escobar. So the show takes place mainly in Colombia, and most of the characters speak using the distinctive local brand of Spanish. Yes, Spanish speakers can ‘understand’ the show even without translations, but are they able to ‘enjoy’ the show when the characters speak regional slangs, dialects, puns, or insider jokes? Not exactly. That is why XL8 provides translations for major language variations, such as Latin American Spanish, Brazilian Portuguese, Taiwanese Chinese, etc, and is working hard to branch out again from these variations as well, such as Colombian Spanish, Argentinian Spanish, and so on. This will allow our platform to provide an enjoyable experience for not only users who don't know the language, but also those who know the language but don't have insight into its cultural background.

So to put in a nutshell, if you are looking to provide your content to the rest of the world, XL8 is the only platform that you will need to give your users an accurate, well-rounded localization experience!

Discover more here

Reference

[1] Preeti Wadhwani, Saloni Gankar (June 2022), “Machine Translation Market size worth $7.5 Bn by 2030”, Global Market Insights

[2] SmartCat (May 2022), “What is machine translation and how does it work?”, SmartCat

[3] Philip Kiely (Sep 2019), “Amazon versus Google Translate”, Wonderproxy

[4] Daniel Ruby (Jul 2022), “Netflix Subscribers 2022 — How Many Subscribers Does Netflix Have” Demandsage