If you use electronics much, you have probably at some point come across a translation program. These are computer programs that take text (or voice) from one language and translate it into another. This is quite handy for someone who is traveling in a foreign country and can’t read the road signs, or language learners who can’t quite remember that obscure new word.
Are programs like these helpful in Bible translation? Could this language technology be used to speed up a translation project that would otherwise takes a decade or more to complete? The answer, unfortunately, is “not really.” But rather than asking you to just take my word for it, I’ll explain these programs and the limitations that make them unusable for Bible translation.
Traditional Translation Programs
There are two different types of translation software that exist. The first type I will call traditional translation programs, because they came first. Traditional translation programs are manually coded to translate words and phrases between two languages. To make one of these programs, programmers tell the program exactly how to translate between languages. For instance, the program creator will tell the program, “Whenever you encounter ‘dog’ in Language A, change it to ‘perro’ in Language B.” The creators of the translation program then code similar commands for every relevant word or phrase that exists in Language A. Additionally the program must be told what adjustments to make for syntactic differences between the two languages. If Language B has a different sentence word order than Language A, this must be accounted for in the program’s development.
The development of a traditional translation program is tedious and requires the talents of multiple linguists and computer programmers. Developers could easily spend several years or more developing one translation program, and that’s not counting the time they spend on the education necessary to undertake such a project. In addition, the development team would need at least one individual very fluent in Language B.
Traditional translation programs work the best between similar languages such as English and German. These two languages have similar features that work well with the “find-and-replace” approach of the traditional translation programs. However, it is much more difficult to make these programs accurate between dissimilar languages. The more dissimilar two languages are, the more time and talent will be required for the development of a translation program between the two, and the less accurate the resultant program will be. Most of the languages that still need Bibles are quite different from English (or any other Indo-European language), making a traditional translation program an impractical solution for Bible translation most (if not all) the time.
Artificial Neural Networks and Google Translate
Beginning in the 1990s, a new type of translation software has emerged that consistently out-performs traditional translation programs. This new method uses artificial neural networks (ANNs) programmed with self-learning algorithms to create the translation software. Rather than telling the translation program how to translate words from Language A to Language B, the program creators take a document which already exists in the two languages (translated by humans from Language A to Language B or vice versa) and feed the bilingual version into the ANN. The ANN analyzes the document in the two different languages and begins forming hypotheses on its own about how to translate between the two languages the words, phrases, and sentences contained in the document. The software developers then go on to feed hundreds, thousands, or even millions of documents into the ANN, all already translated into both Language A and Language B. As the network analyzes all these documents, it continually refines its hypotheses about how to translate between the two languages. Eventually, the ANN develops translation algorithms that work quite well, and the resultant software is accurate enough to perform many basic translation tasks.
Google Translate uses the neural machine translation approach described above, and its software has high levels of accuracy between English and French, German, Spanish, and Chinese (though it’s still not perfect). Google Translate supports over 100 other languages, but the translation accuracy for these languages is lower than the four pairs listed above. The reason for this is obvious, and it is also what makes the neural machine translation approach unusable for Bible translation. For a satisfactory level of translation accuracy to be developed by an ANN, hundreds (at a minimum) of documents need to be available in both the source language and the target language. These documents exist in the widely spoken and politically connected languages of English, French, German, Spanish, and Chinese, but are harder to find in languages spoken by fewer people or a less influential group.
Bible translation projects are happening today among the least-developed languages of the world. For many of these languages, the Bible is the very first book or document to ever be translated. Written material simply does not exist to support the development of neural machine translation.
Only a Rough Draft
To summarize, why is neither type of translation software helpful for Bible translation?
Traditional Software
- Requires extensive development for each language by a well-educated team.
- Does not translate well between dissimilar languages.
Google Translate-type Software
- Requires hundreds of documents already translated into the target language for software development.
- Not an option for undeveloped languages.
But even if a decent translation program could be miraculously created with minimal effort, it would still only accomplish part of the Bible translation process. The Bible contains a lot of word plays, poetry, and figurative speech, language features that are much more difficult to translate than standard prose or narrative. Thus, any Bible portions translated by computer software would be merely a “rough draft.” This rough draft would need to be meticulously edited in conjunction with the target language community, ensuring that every statement of the Bible will communicate its intended meaning clearly and correctly.
A Life Lived Among Them
So no, translation programs are currently not very helpful in Bible translation, and it is doubtful that they ever will be. But I don’t see this as a loss because I believe humans are much better suited for the job in other respects. At All-Nations, we like to talk about “incarnational ministry.” Jesus came to Earth and walked among men, preaching the Kingdom of Heaven, living out the will of the Father, and leaving us an example to follow. In the same way, our goal is to “incarnate” into communities of unreached people groups. We do not take the place of Jesus for them, but we live out the will of the Father as He did and teach them to do the same. The long years of struggling over the details of Bible translation alongside an unchurched people group provides the perfect opportunity for the human Bible translator to demonstrate a life lived for God, something a computer program could never do.
—LH, All-Nations field member