Some common use examples for Ace, illustrated with Jacy:

For the full list of command line options see AceOptions.

Parsing

Prepare for parsing, outputting only the top best MRS, nicely formatting the output (one EP per line)

$ ace -g grammar.dat -Tf1
$ 犬 が 吠える

SENT: 犬 が 吠える
[ LTOP: h0
INDEX: e1 [ e TENSE: pres MOOD: indicative PROG: - PERF: - ASPECT: default_aspect PASS: - SF: prop ]
RELS: < [ udef_q_rel<-1:-1> LBL: h3 ARG0: x2 [ x PERS: 3 ] RSTR: h4 BODY: h5 ]
 [ "_inu_n_rel"<-1:-1> LBL: h6 ARG0: x2 ]
 [ "_hoeru_v_1_rel"<-1:-1> LBL: h7 ARG0: e1 [ e TENSE: pres MOOD: indicative PROG: - PERF: - ASPECT: default_aspect PASS: - SF: prop ] ARG1: x2 ] >
HCONS: < h4 qeq h6 > ]

NOTE: 1 readings, added 183 / 63 edges to chart (22 fully instantiated, 26 actives used, 11 passives used)      RAM: 1327k

Batch Parsing

With sensible limits and using tnt for POS tagging (which enables unknown word processing).

cat FILE | ace --max-chart-megabytes=1920 --max-unpack-megabytes=2048 --tnt-model wsj.tnt -g eng.dat -Tf -n 10 

Paraphrasing

Pipe parse output into generation:

$ ace -g grammar.dat -Tf1 | ace -g grammar.dat -e
$ 犬 が 猫 を 追う 。
NOTE: 1 readings, added 323 / 101 edges to chart (31 fully instantiated, 37 actives used, 17 passives used)     RAM: 1996k
犬 が 猫 を 追う
猫 を 犬 が 追う
NOTE: 85 passive, 431 active edges in final generation chart; built 88 passives total. [2 results]

Compiling the Grammar

This produces the compiled grammar.

$ ace -G grammar.dat -g path-to/config.tdl

When you compile a grammar, ace gives you a brief summary:

2057 types (601 glb), 16751 lexemes, 63 rules, 12 orules, 52 instances, 22622 strings, 168 features

These refer to (FCB thinks):

types

number of non-leaf types

of which X are created by ace as glbs

lexemes

lexical items

rules

grammar rules

orules

inflectional rules

?both with and without orthographic effect

instances

???

everything else: roots, chart-mapping rules, ???

string

distinct strings

features

distinct features

AceUse (last edited 2018-01-24 08:59:33 by FrancisBond)

(The DELPH-IN infrastructure is hosted at the University of Oslo)