Researchers have developed an automated method that allows to understand the degree of difficulty of any individual sentence. This is the first technique of these characteristics, since the ones used so far only work for full documents.
Among possible applications, it would be specially useful for the purpose of knowing if the general public understands or is struck by a given commercial slogan; it would also help to adapt movie scripts so that they are easily understandable by everyone, or to evaluate the understandability of legal texts or political speeches.
"In the field of marketing, for instance, it could help a company to decide between two or three given slogans. This method would score each one of them, and this would allow them to be ranked according to their understandability and to choose which one will be best understood and will reach more people," point out Luis Leiva and Germán Sanchis-Trilles, researchers at the Pattern Recognition and Human Language Technology (PRHLT) center of the Universitat Politècnica de València (Spain).
The method takes into account how memorable and representative is a sentence, both in general linguistic terms and in a specific context. For this purpose, it relies on a statistical model that is independent of the language, but that implicitly gathers its peculiarities.
The researchers have successfully evaluated the method in a sampling experiment, for which they used a set of two million sentences from the proceedings of the European Parliament, and selected the 500 sentences which are most adequate according to the method, and also according to two other similar strategies. The experiment was carried out both in English and in Spanish, and the three techniques were evaluated with a group of native speakers of each language. In both cases, users were able to remember with more ease the sentences selected by the new method.
The method will be presented at the Human Factors in Computing Systems (CHI) conference in Toronto in April 2014.
Cite This Page: