In some systems the semantic analysis is preceeded by a syntactic analysis. In those systems the grammar constrains the sentences that are allowed.
An augmented transition network (ATN) can be used for extending the syntactic analysis to enforce grammatical agreement between phrases, while doing the syntactic analysis [27]. An augmented transition network is a state machine with nodes that represent states (in the parsing process) and arcs connecting the words with legal paths [16]. (See Fig. 4.1.) The grammar of the system is represented by this state machine, allowing only meaningful sentences. Each arc is labeled by a condition that needs to be satisfied in order to traverse it; also with each arc there is an associated action. These actions can be implemented by procedures that modify the system's data base, and they enforce grammatical agreements over a sentence. The conditions can be a word, or another ATN that needs to be satisfied. The ATN may be called recursively in one of these conditions. The input to the ATN is a stream of written words. Each word will lead the state machine to different nodes in the ATN, producing different actions during this process.
One way to represent the meaning contained in a sentence is through relationships of objects. During this process the main event describing the sentence and participants are found, determining the roles they play in the event, and under which conditions the events took place. Sometimes finding the verb sets the possible role and conditions in which the actions may occur. Then it is possible to associate some verb frames that need to be filled; these may represent the participants and their relationships. Conceptual Dependency (CD) uses this type of technique to represent the meaning contained in a sentence.