lexicalised grammar formalisms such as lexicalized tree adjoining grammar (ltag) and com binatory categorial grammar (ccg) assign one or more syntactic structures to each word in a sentencewhich are then manipulated by the parser. supertag ging was introduced for ltag as a way of increasingparsing efficiency by reducing the number of struc tures assigned to each word (bangalore and joshi, 1999). supertagging has more recently been applied to ccg (clark, 2002; curran and clark, 2003).supertagging accuracy is relatively high for man ually constructed ltags (bangalore and joshi,1999). however, for ltags extracted automati cally from the penn treebank, performance is much lower (chen et al, 1999; chen et al, 2002). in fact, performance for such grammars is below that needed for successful integration into a full parser (sarkar et al, 2000). in this paper we demonstratethat ccg supertagging accuracy is not only sufficient for accurate and robust parsing using an auto matically extracted grammar, but also offers several practical advantages. our wide-coverage ccg parser uses a log-linear model to select an analysis. the model paramaters are estimated using a discriminative method, that is,one which requires all incorrect parses for a sentence as well as the correct parse. since an auto matically extracted ccg grammar can produce anextremely large number of parses, the use of a su pertagger is crucial in limiting the total number of parses for the training data to a computationally manageable number. the supertagger is also crucial for increasing thespeed of the parser. we show that spectacular in creases in speed can be obtained, without affectingaccuracy or coverage, by tightly integrating the su pertagger with the ccg grammar and parser. to achieve maximum speed, the supertagger initially assigns only a small number of ccg categories toeach word, and the parser only requests more cate gories from the supertagger if it cannot provide an analysis. we also demonstrate how extra constraints on the category combinations, and the application of beam search using the parsing model, can further increase parsing speed.this is the first work we are aware of to succes fully integrate a supertagger with a full parser which uses a lexicalised grammar automatically extractedfrom the penn treebank. we also report signifi cantly higher parsing speeds on newspaper text than any previously reported for a full wide-coverage parser. our results confirm that wide-coverage ccg parsing is feasible for many large-scale nlp tasks.this research was supported by epsrc grant gr/m96889, and a commonwealth scholarship and a sydney university travelling scholarship to the second author. this paper has shown that by tightly integrating a supertagger with a ccg parser, very fast parse times can be achieved for penn treebank wsj text. our results confirm that wide-coverage ccg parsing is feasible for many large-scale nlp tasks. lexicalised grammar formalisms such as lexicalized tree adjoining grammar (ltag) and com binatory categorial grammar (ccg) assign one or more syntactic structures to each word in a sentencewhich are then manipulated by the parser. supertag ging was introduced for ltag as a way of increasingparsing efficiency by reducing the number of struc tures assigned to each word (bangalore and joshi, 1999). the best speeds we have reported for the ccg parser are an order of magnitude faster. to give one example, the number of categories in the tag dictionary?s entry for the wordis is 45 (only considering categories which have appeared at least 10 times in the training data). we also report signifi cantly higher parsing speeds on newspaper text than any previously reported for a full wide-coverage parser. however, in the sentence mr. vinken is chairman of elsevier n.v., the dutch publishing group., the supertag ger correctly assigns 1 category to is for ? = 0.1, and 3 categories for ? = 0.01.