Is a Universal Translator possible?

-
This short presentation provides an excellent video introduction to the basic principles and difficulties of machine translation. Two main methods of machine translation are discussed: Rule-based Translation and and Statistical Machine Translation. Each method poses particular difficulties.

1. Rule-based Translation uses lexical databases (dictionaries) and sets of linguistic rules to parse syntax, morphology and semantics. It's the semantic aspect that poses the greatest challenge.

2. Statistical Machine Translation analyses databases of books, documents and articles that have already been translated by humans, and uses these to identify correspondent language patterns. The quality of this method depends on the size of the database and the availability of samples for certain languages or styles of writing. The difficulties arise primarily with linguistic irregularities and shades of meaning.

It seems that understanding language may be a unique facet of the human brain structure, and some researchers now believe that the universal translator is not possible.

For now, studying language the old-fashioned way will give better results.

Watch the video for the full explanation. Alternatively visit the full online lesson at TEDEd.

 

Comments

Post a Comment

Popular posts from this blog

Computers have learned to make us jump through hoops

-
Image
John Naughton , writing in the Observer (18th November 2018) explains that the acronym CAPTCHA means 'Completely Automated Public Turing test to tell Computers and Humans Apart'.  In the Turing Test , a human interrogates two interlocutors - one human, one machine - and based on the interlocutors' responses, the human interrogator has to determine whether she is conversing with a machine or a human.  In other words, to what extent can a machine understand and utter human language patterns? Naughton argues, however, that the Turing Test has been turned on its head, and it is now machines asking we humans to prove that we are not machines.   He further explains that image-based CAPTCHA, like the one shown below, are being used to train machine-learning algorithms for driver-less cars. So we humans are training the machines, and as such we are being used as unpaid labour. Read the article online here
Read more

Social Networking under Attack

-
'The Guardian' recently ran an interesting article about a current trend of criticism of social networking technologies (e.g. websites such as Twitter and Facebook).  Read the article here The arguments include claims that social networking makes us lazier, less able to digest large quantities of information, and less human. An example: we go into a cafe and see people sitting at their laptops or tapping away on their mobile devices, whilst not actually talking to the people around them. So in effect, ignoring the real people around them, and focusing instead on communicating with others through electronic devices. Those who disagree with these arguments state that social networking actually increases communication and helps to bridge geographical distances. They also argue that it is simply not true that people used to chat with those around them more before these new technologies existed. So the question is: did people really used to chat to each other more (on buses an
Read more

Can you tell whether you are talking to a human or a chatbot?

-
Image
In ' The charge of the chatbots: how do you tell who’s human online? ', published in the Observer newspaper November 18th 2018, Tim Adams reports on advances in artificial intelligence in the field of human-computer conversation. Some of the main points he makes are: 1. Advances in technology are creating chatbots that mimic human behaviour to a sophisticated extent:  “The ability of bots – a term which can describe any automated process present in a computer network – to mimic human online behaviour and language has developed sharply in the past three years.” 2. One dimension of the effectiveness of these entities is the human tendency to put our trust in other human voices: "these creations can exploit our tendency to ascribe trusted human characteristics to voices even if, on a rational level, we suspect that they are artificial. That psychology is as old as electronic communication itself." 3. The first such chatbot was developed in the USA in
Read more