SPARKLE's overall research plan is focussed on the development of tools and data intended to assist the creation of commercial language engineering applications with specific reference to multilingual information management. The project intends to explore the optimal functionality of robust phrasal parsing combined with classification techniques which utilise manageable linguistic knowledge and statistical data from large corpora to semiautomatically acquire a new spectrum of application-specific lexical resources. These will provide rich and reliable information about areas of word knowledge in which most extant conventional dictionaries, lexical databases and large scale NLP lexicons are demonstrably weak, e.g. predicate subcategorisation, argument structure and semantic preferences. This work will have immediate impact on the next generation of multilingual lexical resources and will therefore be a milestone in the creation of information management tools.