Language and innovation are inseparable. Language conveys ideas which are essential in innovation, establishes the most immediate connections with our conceptualisation of the outside world, and provides the building blocks for communication. Every linguistic choice is necessarily meaningful, and it involves the parallel construction of form and meaning. From this perspective, language is a dynamic knowledge construction process. Emphasis is laid on investigating how words are used to describe innovation, and how innovation topics can influence word usage and collocational behaviour. The lexical representation of innovative knowledge in a context-based approach is closely related to the representation of knowledge itself, and gives the opportunity to reduce the gap between knowledge representation and knowledge understanding. This will bring into focus the dynamic interplay between lexical creativity and innovative pragmatic contexts, and the necessity for a dynamic semantic shift from context-driven vagueness to domain-driven specialisation.

Methodology and experimental evidence - Method and materials: the challenge of identifying changes in word sense has only recently been considered in Computational Linguistics. To investigate the themes discussed in the previous sections genre-oriented and stylistically heterogeneous English texts are analysed, with the support of SKETCH ENGINE (Kilgarriff et al., 2004), which is a corpus query tool, based on a distributed infrastructure, that generates word sketches and thesauri which specify similarities and differences between near-synonyms. By selecting a collocate of interest in a sketched word, the user is taken to a concordance of the corpus evidence giving rise to that collocate. Ambiguous and polysemous words have been selected with particular reference to innovative domains, and their collocations are analysed. In particular, we considered the domain of brain sciences and new technologies of brain functional imaging, the domain of knowledge management processes, and the field of information technologies, by mainly focusing on the following test words: IMAGING, RETENTION, STORAGE, CORPUS, NETWORK, GRID. The selected words present a potentially high degree of semantic ambiguity or polysemy and different degrees of semantic specialisation, which can be analysed objectively by studying their context collocations. For a terminology exploration, both domain-specific and general-purpose texts materials are selected by using generic search web engine queries (www.google.com by using seed words), domain-specific databases and type coherent multidisciplinary large corpora (e.g. www.opengrey.eu, www.ncbi.nlm.nih.gov/pubmed by selecting the domain). Collocations and concordances are then compared with large balanced corpora (e.g. the British National Corpus, British Academic Written English, New Model Corpus, and the like, whose size ranges between 8 M and 12 G tokens).