1.  Reducing the length One version of the Edinburgh syntactic component used pair-wise transition probabilities between words.

2.  TRACE II does not have a syntactic component, but has a mechanism for such information to be used.

3.  For example, suppose the syntactic component used the pair-wise parsing process mentioned above to decide between different lexical hypotheses.

4.  A more complex syntactic component might need both more grammatical information and more time to make a decision.

5.  Even the perfect fine-class transcriptions resulted in extremely large numbers of paths, making it very difficult for the syntactic component to distinguish the correct interpretation.

6.  One version of the Edinburgh syntactic component used pair-wise transition probabilities between words.

7.  Even with fine-class input the syntactic component was unable to discriminate between the intended utterance and various homophonic utterances produced by lack of word boundary information.