Grammar Writing: Week 10: Difference between revisions
No edit summary |
No edit summary |
||
| Line 16: | Line 16: | ||
What our current grammars are missing to express the generalization covered by the five rules above is the concept of '''identity'''. By identity, we mean that one larger feature structure contains a smaller feature structure twice, in different places. Or more simply, that the information at two places in a feature structure must be the same. | What our current grammars are missing to express the generalization covered by the five rules above is the concept of '''identity'''. By identity, we mean that one larger feature structure contains a smaller feature structure twice, in different places. Or more simply, that the information at two places in a feature structure must be the same. | ||
In order to signal such identity, we will use so-called boxed numbers, i.e. 1,2,3, ... with a box around it. Unfortunately, in this WIKI, we cannot represent such boxes. Therefore, we will use | In order to signal such identity, we will use so-called boxed numbers, i.e. 1,2,3, ... with a box around it. Unfortunately, in this WIKI, we cannot represent such boxes. Therefore, we will use symbols like the following as identity markers: ©1, ©2, ©3, ... | ||
Revision as of 13:11, 18 December 2017
Identity
Last week, you were supposed to write a grammar that accepts different kinds of verb phrases. The rules you had to write all took the following form:
1. VP --> V, COMPS <NP> + NP
2. VP --> V, COMPS <PP> + PP
3. VP --> V, COMPS <AP> + AP
4. VP --> V, COMPS <sentence> + sentence
For NPs, APs, PPs you would also have to write several rules.
Clearly, the rules above are very similar, but they fail to express the similarity. This is shown by the fact that if there were a verb that wants to combine with a VP-complement, then we would would need to add another rule to the ones above, as follows:
5. VP --> V, COMPS <VP> + VP
What our current grammars are missing to express the generalization covered by the five rules above is the concept of identity. By identity, we mean that one larger feature structure contains a smaller feature structure twice, in different places. Or more simply, that the information at two places in a feature structure must be the same.
In order to signal such identity, we will use so-called boxed numbers, i.e. 1,2,3, ... with a box around it. Unfortunately, in this WIKI, we cannot represent such boxes. Therefore, we will use symbols like the following as identity markers: ©1, ©2, ©3, ...