How translation can help to shine a light on the relationship between words and their meanings - Part 1

cartographie réseau sémantique 570

Scientist have discovered a universal model when dealing with the meaning of words, revealed by the way in which the sense of the word evolved.

Some words have radically changed meaning with the passage of time. Once, the word awful denoted impressed or filled with fear. Clue once meant a ball of thread. Being called “nice” wasn’t really a compliment; it meant stupid, idiotic, or ignorant.

But how does the definition of a word change so radically?

In an effort to try to understand how the meaning of words evolve, a team of scientists mapped out the relationships between different words and their meanings, trying to capture the process in motion.

If you have ever learned a new language or translated sentences from one language to another, you know that a single word will not always be directly translatable by another. Some terms even have multiple meanings, like the Hawaiian term aloha, which means both hello and goodbye.

The team of scientists concentrated on words with multiple meanings to create a semantic network linking words through multi-language translations. The linguistic map that resulted can be seen here, taken from an article published in the journal Proceedings of the National Academy of Sciences.

We didn’t know how to represent the relationship between words and their meaning, nor how the meanings of the words change,” said Tanmoy Bhattacharya, one of the researchers on the study form the Institute of Santa Fe.

The team searched for models between the words in different languages. First they chose no less than 81 languages from across the world. These languages were chosen across diverse historical, cultural, and geographical contexts.

Most linguistic studies are centred on languages in occidental, industrialised, rich, and democratic countries,” indicated the principal author of the study, Hyejin Youn, also from the Santa Fe Institute. “But, since the team has been chasing universal models, they have chosen languages that are not likely to have many points in common in their ancestry or by interaction.”

English was used as the meta-language. 22 simple nouns were selected, like the sun, the sea, dust, water, and stone, and used for simple translation.

Each word was translated from English to each of the 81 languages. Then, the word or words that were returned were translated back to English. Over the course of this re-translating, certain words started to take on multiple meanings.

Dr Bhattacharya described, “You can start with the word sun and you’ll end up with all kinds of words. When you take these same words and you translate them back to the source language, you get the sun and the moon in certain languages.”

This process of translation and re translation created a network of words more or less strongly related.

A web over the natural elements

When they looked at the data, the researchers realised that the strongest ties regrouped words in three large categories. These groupings, astonishingly, belonged to the classification of our world's natural elements. Words relating to water formed a sub class while those related to stone or mountains formed another. The third group was the largest and comprised of multiple elements connecting words like dust with smoke or the sun with fire and wind.

You can easily pass from a word meaning the sun to something describing a star,” said Bhatacharya, “but you will have a hard time arriving at the word river.

There seems to be a universal structure across different languages in the meanings of words at a certain level,” concluded Dr. Youn.

Read part 2

Category: News