Automatic interpretation of neologisms the case of blends

📄 Project Abstract

The aim of this project was to use a combination of non-context and in context features to solve the problem of automatically identifying the source words of lexical blends, in particular, those blends which consist of a combination of two source words that occur sequentially within the blend and contribute a prefix and a suffix respectively. Lexical blends are neologisms formed from the combination of two or more pre-existing words. Previous research by Cook and Stevenson (2010) identified twenty-two features that take into account the frequencies and co-occurrence , the length, contribution and phonology and the sematic relatedness of the words. These features were used to generate values for candidate source words, which were generated through splitting the blend into prefix and suffix pairs. The values were used to rank the candidate source word pairs to help determine the correct source words for the blend. Through the collection of a corpus and the re-implementation of the twenty...