Towards the Automatic Identification of Adjectival Scales: Clustering Adjectives According to Meaning