20.07.2013 Views

Notes on computational linguistics.pdf - UCLA Department of ...

Notes on computational linguistics.pdf - UCLA Department of ...

Notes on computational linguistics.pdf - UCLA Department of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Stabler - Lx 185/209 2003<br />

1. Download the file lrp-cnt.pl and modify the definiti<strong>on</strong> <strong>of</strong> parseOnce so that in additi<strong>on</strong> to printing out<br />

the number <strong>of</strong> steps in the search, it also prints out the number <strong>of</strong> steps in the derivati<strong>on</strong> (= the number <strong>of</strong><br />

nodes in the tree).<br />

2. Which 6 word sentence requires the l<strong>on</strong>gest search with this ES grammar? Put your example in at the end<br />

<strong>of</strong> the file.<br />

3. Add 1 or 2 rules to the grammar (d<strong>on</strong>’t change the parser) in order to produce an even l<strong>on</strong>ger search <strong>on</strong> a<br />

6 word sentence – as l<strong>on</strong>g as you can make it (but not infinite = no empty categories). Put your example at<br />

the end <strong>of</strong> the file, and turn in the results <strong>of</strong> all these steps.<br />

6.6.6 Additi<strong>on</strong>al Exercise (for those who read the shaded blocks)<br />

(61) Gibs<strong>on</strong> (1998) proposes<br />

For initial purposes, a syntactic theory with a minimal number <strong>of</strong> syntactic categories, such as<br />

Head-driven Phrase Structure Grammar (Pollard and Sag, 1994) or Lexical Functi<strong>on</strong>al Grammar<br />

(Bresnan, 1982), will be assumed. [Note: The SPLT is also compatible with grammars assuming a<br />

range <strong>of</strong> functi<strong>on</strong>al categories such as Infl, Agr, Tense, etc. (e.g. Chomsky 1995) under the assumpti<strong>on</strong><br />

that memory cost indexes predicted chains rather than predicted categories, where a chain is a set<br />

<strong>of</strong> categories coindexed through movement (Chomsky 1981).] Under these theories, the minimal<br />

number <strong>of</strong> syntactic head categories in a sentence is two: a head noun for the subject and a<br />

head verb for the predicate. If words are encountered that necessitate other syntactic heads<br />

to form a grammatical sentence, then these categories are also predicted, and an additi<strong>on</strong>al<br />

memory load is incurred. For example, at the point <strong>of</strong> processing the sec<strong>on</strong>d occurrence <strong>of</strong><br />

the word “the” in the object-extracted RC example,<br />

1a. The reporter who the senator attacked admitted his error,<br />

Det<br />

the<br />

NP<br />

Noun<br />

reporter<br />

NP<br />

Comp<br />

thati<br />

Det<br />

the<br />

S’<br />

NP<br />

S<br />

Noun<br />

senator<br />

S<br />

V<br />

admitted<br />

Verb<br />

attacked<br />

VP<br />

VP<br />

NP<br />

ei<br />

Det<br />

the<br />

NP<br />

Noun<br />

there are four obligatory syntactic predicti<strong>on</strong>s: 1) a verb for the matrix clause, 2) a verb for<br />

the embedded clause, 3) a subject noun for the embedded clause, and an empty category NP<br />

for the wh-pr<strong>on</strong>oun “who.”<br />

Is the proposed model a glc parser? If not, is the proposal a cogent <strong>on</strong>e, <strong>on</strong>e that c<strong>on</strong>forms to the<br />

behavior <strong>of</strong> a parsing model that could possibly work?<br />

(62) References. Basic parsing methods are <strong>of</strong>ten introduced in texts about building compilers for programming<br />

languages, like Aho, Sethi, and Ullman (1985). More comprehensive treatments can be found in<br />

Aho and Ullman (1972), and in Sikkel (1997).<br />

102<br />

error

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!