“Chaos was the law of nature; Order was the dream of man,” said Henry Brooks Adams, an American historian over a century ago. Rules and order are our tools to explain the world. We like to see cause and effect, and languages are no exceptions. Grammar is a way to find rules in a language.
Dictionary and grammar aren’t enough
Rules and machines are inseparable. Programming has traditionally been a way to teach a machine to follow the rules. It’s no surprise that rule-based machine translation was one of the first approaches to automate translation.
The idea was pretty clear. If you give a dictionary and a grammar to a machine, it can translate. The machine used the words in a dictionary based on the rules of grammar. The outcome was a translation from one language to another. However, the quality of a rule-based machine translation isn’t as good as one could expect.
Languages are more than grammar
The problem with rule-based machine translation is that it doesn’t fit into the world of languages. It is an attempt to find rules that don’t exist. Languages are complex, irregular and constantly evolving. Of course, there are some consistency and repetitiveness in every language, but rules and grammar explain a language only partly.
If you have ever studied a foreign language, you know that exceptions and irregularities are extremely common. For example the use of verbs, most verbs might be used according to some sort of regular form. Yet, the most used verbs tend to have irregular inflections. People want to find logic. Rules help to explain the world and gives the feeling of control and security. But that’s not how languages work. And thus, if learning a language is a challenge for a human, it will be difficult for a machine.
Learning a language means hard work
All this might seem strange if you only know one language. We have a blind spot in our mother languages. When a child learns his or her first language, it goes with small steps and takes a long time. Although the learning never stops, after a certain level the use of mother language becomes somewhat automatic. You don’t have to think about the words and you can concentrate on your message.
Even if you master your mother language, you may not know how it works. It is a different thing to learn how to express your thoughts and feelings than it is to explain the language. For example, Finnish speaking Finns don’t think that it is the most difficult language to learn in the world. And if you first learn to speak Finnish, the next language won’t be much easier to adopt.
Languages are complex and fantastic
If you developed a language in a controlled way, it would be well structured and logical. Languages are means of communication and it has taken centuries to get to this point. Languages aren’t tools, they have a history and they tell a story of their own. Every word has an etymology and trajectory.
We would be absolutely bored to speak, write and read an unsurprising language. Languages are a matter of the heart rather than brains. Still, we need a great deal of our amazing brain capacity to use our languages. Rule-based machine translation is far too simple as compared to our minds.