Taalstatistiek

Discussion of the possibilities and limitations of Language Statistics. Sampling of “normaI” prose is difficult as the “universe of written Dutch” is undefined. What predictions can be made about the contents of tomorrow's newspaper? The letter frequencies are very stable. Frequencies ofbigrams and trigrams of letters bring us towards the word level about which something may be predicted. The sentence level can only be dealt with after a mechanical sentence analysis method is available. The highest level is still outside the scientific domain and belongs to the literary critic. Literary Statistics should be founded on a solid knowledge of Language Statistics. It is considered inappropriate to begin this field with difficult historical problems. The mutual distrust between linguist and statistician requires much tact on both sides. Eventually Language Statistics may become one of the bridges between the “two cultures”. The modern computer is an indispensable tool both for practical reasons (the giant mass of material) and for theoretical ones (the need to give unambiguous definitions of concepts like “sentence”, “word” and “syliable”). The use of Language Statistics in Mechanical Translation research is discussed. Review of the activities in the Netherlands. A one million word count is on its way. The speaker pleads for a National Center for Lexicology and Language Statistics.