Distinguished Speaker, Professor Antoinette Renouf

Language at Leeds is pleased to announce that we will be joined by Professor Antoinette Renouf (Birmingham City University), who will speak on "Semantic Neology: the challenges for automatic identification"

More information on Antoinette and her research can be found at: http://www.bcu.ac.uk/english/staff/antoinette-renouf

In this talk, the term ‘semantic neology’ is used to denote the emergence of a new sense of an existing word in news text. The identification and tracking of new word senses is vital to recorders of the language: lexicologists, lexicographers, translators, terminologists, NLP specialists; but while automated systems are now increasingly used to identify new words in electronic text across time, researchers are still struggling to find an automatic way to pinpoint new senses.         What is needed is an automated filtering system for words passing through a diachronic corpus, combined with a means of identifying a candidate new sense and monitoring its evolution towards a threshold of significance.         I shall outline such a prototype system, developed at Birmingham University in 1990-1993 and refined in 2006-2008 at Birmingham City University. The design is simple. Lexis is precisely patterned in text, and the collocational pairings of words is statistically analysable. A small theoretical step needs to be taken, to accept that the collocates of a word reveal (part of) its meaning, and a further step deems a change in collocates   to reveal a potential change in meaning. With this theoretical framework in place, an automated system is able to identify the emergence of semantic neology across time.         However, though this approach works in principle, there are constraints posed by the language, in relation to the discarding of an old sense and the accruing of a new one, processes which are not always clear-cut. In this talk, I shall outline the system and then illustrate some of the challenges posed by the evolving language of news text.