No language problem as computers condense texts
Today´s extreme abundance of digital information makes it impossible to select the right one for ourselves, from among it all, by traditional means. The major world languages can now offer computer programs for making summaries – but in the case of minor languages the cost of this has so far been prohibitive. However, KTH scientist Martin Hassel has now developed a language-independent text summariser.
Automatic text summarising is performed by a computer that edits a longer text mass into a shorter one, making it free from irrelevant details. In his dissertation Martin Hassel presents a model of how to build a language-independent text summariser by combining a set of fundamental language adaptation tools. This makes it possible to devise automatic text summarising programs, also for minor languages, at a fairly modest cost.
– With minor languages the major obstacle is that they often lack large amounts of text that was collected for language research, says Martin Hassel. Also, resources for producing just that are in short supply, too! Besides, this is time-consuming and usually involves a lot of manual labour.
Still, also minor languages do need this facility for taming the steadily exploding volumes of electronic text. Martin Hassel has put his research focus on devising automatic condensation of texts with a minimum of human effort. In other words, resources for this should preferably consist of materials that are available today – these need not be produced specifically for text summarising. The ideal here is if one can find copy with a literary content, or written as part of a literary process. Martin Hassel continues:
– The summarising system is fairly easily put together from a small number of basic, elementary language tools. This results in a summariser that is very nearly language-independent! Thus it may easily be moved from one language to another.
Magnus Myrén