Some time ago, the media reported that two of the most successful players on Jeopardy!, a popular quiz show in the US, fought it out and lost against a new contestant called Watson.

The newcomer, who was much faster at responding correctly, won the $1 million (€730,500) jackpot after a marathon lasting three nights.

Jeopardy! is a ‘reverse’ quiz program, in which contestants provide questions on a huge range of topics – from current affairs to the arts and sciences – based on clues that make up their answer.

Surprising was not only that Watson, a complete newcomer, easily managed to outplay his two experienced opponents, but that Watson is not a human being. Rather, Watson is an IBM computer.

Even more fascinating is the fact that not only did Watson respond correctly and faster than the other two, but it produced its responses in English.

Those familiar with Stanley Kubrick’s classic science fiction film 2001: A Space Odyssey will recall the flight computer Hal, who not only speaks perfect English, but also has feelings and ambitions.

Several decades after Kubrick articulated his science fiction fantasy, we have not managed to create a machine that gets anywhere near Hal’s linguistic skills, let alone his emotional maturity. However, the reality is in some ways more interesting than fiction.

The dream of creating machines that can communicate with people using natural language has been around for a while. Imagine being able to have a sensible conversation with your laptop or iPhone.

However, creating machines that can use natural language has always been a huge challenge for researchers.

Great progress has been made, but we still have a very long way to go.

For example, machines like Watson are very good at what they do – in this case guessing the right response based on a number of clues and a huge knowledge database.

However, if you asked Watson a simple question like “What colour are my eyes?” or “What’s the weather like outside today?”, or “What do you think of so-and-so’s latest song?”, it would be stumped.

Assuming a machine can somehow possess knowledge that, at least in restricted areas, compares well with human knowledge, what does it need to be able to talk about it?

Natural language is itself a very complex system that works by combining small pieces into larger pieces, at many different levels.

Thus, meaningless sounds combine to form words; words and morphemes (little bits of content, like the plural ‘s’ in English, or the bound pronoun ‘ha’ in Maltese) combine to form new words; words themselves combine to form phrases; and phrases merge into larger phrases and sentences.

And it’s all done in context, often with a particular listener in mind to whom we somehow manage to convey what we mean, while also understanding what they’re saying to us – well, most of the time anyway.

Getting a computer to do all that is a tall order, and it’s no surprise that the computational treatment of human language – the study of Human Language Technology (HLT) – involves the contribution of a variety of disciplines – from psychology and linguistics to computer science and artificial intelligence.

Indeed, HLT has become home to many researchers who come from either of these disciplines. The good news is that the University’s Institute of Linguistics is offering a new undergraduate B.Sc. (Hons) course in HLT starting in October.

Most of us use language-sensitive computational tools all the time, even if we don’t realise it. Over the years, the area of HLT has spawned new and exciting fields of research.

Information retrieval is a part of HLT and it is what underlies most contemporary search engine technology, but also many sophisticated data mining techniques, used by organisations from companies to political parties.

These days, anything you Google is bound to return a huge number of hits. That’s just one example of the problem of information overload.

Automatic summarisation systems process multiple documents and select the important information from them, combining them into a single, coherent document. These systems are studied under the umbrella of HLT.

Users of Yahoo!’s Babelfish, or Google Translate, will know just how poor an automatic translation can be. Nevertheless, these and similar systems have proven useful for those who need to quickly process text in a foreign language.

Meanwhile, the dream of having machines that can translate fluently from one language to another remains as compelling today as it ever was, and is researched through HLT.

Another area of HLT is question answering: rather than typing the usual queries into search engines, it sometimes more useful to get concise answers to specific questions (try asking Google “Who is the rector of the University of Malta?”). Question answering systems seek to do just that.

Speech recognition and synthesis is behind many of the systems we periodically get to talk to (and sigh exasperatedly at) over the phone.

Understanding and automatically synthesising human speech has huge benefit potential for many people – from visually impaired readers to customers who need quick answers over the phone.

The European Commission has identified HLT as one of the key research areas for the near future.

However, fascinating though HLT is, it is still under-represented in the Maltese context, though many researchers at the University and elsewhere have been actively working to improve this state of affairs.

The development of an HLT programme is expected to bring with it new research opportunities and to meet the needs of Maltese industry and society.

However, HLT technology cannot flourish without basic research which aims to: first further our understanding of natural language with the aid of computational tools; and incorporate natural language as the primary communicative medium in intelligent systems which are designed to interact and/or provide a service to human users.

The new B.Sc. (Hons) course in HLT provides students with the basic skills required in the sector, including programming, problem solving, project management and knowledge of IT, language and language applications, which should allow them to seek opportunities of employment both in the IT sector, the language sector and academia.

The new course is open both to students who have chosen arts subjects, especially languages and who have no mathematical background, and also to students who have a science, mathematics or IT background, but none in languages or linguistics.

The course kicks off with a foundation year which provides the basic knowledge to students from both backgrounds, subsequently focusing on specifically HLT-oriented areas, including those described above, as well as subjects that focus on language from a cognitive and psychological angle.

The course also includes credit for work experience through short placements with industry.

More information can be obtained from www.um.edu.mt/linguistics/human_language_technology.

Have your say

If you wish to contribute an article or would like a particular subject tackled in the Education section, call Davinia Hamilton on 2559 4513 or e-mail dhamilton@timesofmalta.com.