'Key words' could help signpost document translation

‘Key words’ could help signpost document translation

What does document translation have in common with, say, reading a story, or getting a web page into a search engine’s results?

The answer is that all can begin with key words – the most important words or phrases in a text that carry the most meaning.

Online, search engines look for repeated phrases and use these to help decide what subject a page has been written about.

Likewise, when you’re reading a book, the themes that come up again and again are likely to stick in your mind the most.

But researchers at the Max Planck Institute for the Physics of Complex Systems have found that it’s not just the individual words in a text that matter – there are also relationships that exist over lengthy passages of text.

For instance, when one page of a book like War and Peace mentions the word ‘war’, there is a higher chance of seeing more uses of that word – and even of the letter ‘w’ – elsewhere in the surrounding pages.

It’s a principle that many intermediate speakers of foreign languages put into effect during document translation when looking for the main words in a text as a starting point.

However, researcher Eduardo Altmann reveals that the effect is universal, and remains even if numbers are substituted for the letters.

“It is absolutely irrelevant which language the text is written in,” he says. “The only thing that matters is the story, and not language-specific rules.”

Source: AlphaGalileo Foundation