next up previous contents
Next: The realization rules Up: The Mini-grammar Previous: The Mini-grammar   Contents


The grammar network

The mini-grammar is a fragment of the full Genesys grammar described in Section 2.3.1, extracted and modified to be free-standing. Probability values are expressed in a simplified scale from 0 to 10. Some features which would lead into other areas of the full grammar are inhibited in the mini-grammar by being assigned zero probability, while other features are assigned absolute preference (i.e. 10).

There are two entry points for the mini-grammar network. The main entry point, which is the starting point for every generation with the network, has the entry condition situation and two output features, as shown in Figure 5.1. Congruent leads into the rest of the network for generation of a full clause, via 5 simultaneous systems, TRANSITIVITY, CONTINUING PERIOD, MOOD(1), POLARITY and INFORMATION FOCUS, which are described below. In a fuller grammar non-congruent would be used for generation of a nominalized situation, but in the mini-grammar the non-congruent choice is inhibited by setting its probability to zero.

Figure 5.1: The initial entry to the mini-grammar
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\setlength{\un...
...0mm}{5mm}INFORMATION FOCUS}}
\end{array} \right. $ }
\end{picture}\end{figure}

The other entry point for the mini-grammar, which in a fuller grammar would be the starting point for generation of noun phrases, has the entry condition thing. In the mini-grammar, all choices in the subnetwork for thing are inhibited except those which generate a proper name. The subnetwork for thing is only entered by a re-entry realization rule from the main network for situation, in order to generate a proper name noun phrase to fill a participant role in the clause.

The TRANSITIVITY region of the network is shown in Figure 5.2. Only material and relational process types are covered in the mini-grammar, so the choice of mental process type is inhibited by a zero probability value. Material processes are restricted to a choice of 5 verbs in the CULTURAL CLASSIFICATION system.

Figure 5.2: The TRANSITIVITY network
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\setlength{\un...
...{\line(1,0){4}}
\put(15,0){0 (others)}
\end{picture}}
\end{picture}\end{figure}

The MOOD network is shown in Figure 5.3. The output feature information is the entry condition for 3 simultaneous systems, REFERENCE POINT LOCATION, RELEVANT PASTNESS and MOOD(2). This part of the network handles tense and aspect. The Nigel grammar, following Halliday's mainstream analysis, divides primary tense into past, present and future, as described in detail in [Matthiessen & Bateman 1991] Chapter 7. However, Halliday also recognises an alternative analysis in which future tense is regarded as part of modality (i.e. a disposition to do something), and this is the view adopted in Genesys and the mini-grammar.

Figure 5.3: The MOOD network
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\begin{center}...
...4}}
\put(15,0){0 (others)}
\end{picture}}
\end{picture}\end{center}\end{figure}

REFERENCE POINT LOCATION therefore leads to only two choices in the located system: performer's present or other time reference (i.e. past). If the reference point is located at performer's present, the MODALITY network, shown in Figure 5.4, is entered, and if unconditional disposition is selected, a future tense form is generated. RELEVANT PASTNESS, despite its name, handles aspect not tense: the choice of relevant past generates a perfective aspect form. The realization of these tense, aspect and modality combinations is described in Section 5.2.

Figure 5.4: The MODALITY network
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\begin{center}...
...){\line(1,0){4}}
\put(25,0){6 unmodalized}
\end{picture}\end{center}\end{figure}

The CONTINUING PERIOD network, together with RELEVANT PASTNESS in the MOOD network, also handles aspect. CONTINUING PERIOD distinguishes progressive and non-progressive aspect. It consists of the single system shown in Figure 5.5.

Figure 5.5: The CONTINUING PERIOD network
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\setlength{\un...
...1){\line(1,0){4}}
\put(30,0){3 continuing}
\end{picture}\end{center}\end{figure}

The POLARITY network consists of a single system, shown in Figure 5.6.

Figure 5.6: The POLARITY network
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\setlength{\un...
...,21){\line(1,0){4}}
\put(30,0){2 negative}
\end{picture}\end{center}\end{figure}

The INFORMATION FOCUS network, shown in Figure 5.7, has a system to select between marked and unmarked information focus, i.e. whether special emphasis should be placed on a particular item. In the demonstration implementation, the emphasis is shown by displaying the marked item in upper case. In the mini-grammar, information focus can only be marked on polarity, other candidates for marked focus (such as Agent or Process) being inhibited by a zero probability value.

Figure 5.7: The INFORMATION FOCUS network
\begin{figure}\renewedcommand{baselinestretch}{1}\small\normalsize\setlength{\un...
...5,1){\line(1,0){4}}
\put(30,0){8 unmarked}
\end{picture}\end{center}\end{figure}


next up previous contents
Next: The realization rules Up: The Mini-grammar Previous: The Mini-grammar   Contents
Graham Wilcock 2001-11-15