Can ace show generator input?

Is there a flag I can pass to ace so that I can see the generator input, preferably post-vpm?

I believe you can get what you want by increasing verbosity (-v). See “internal MRS” below:

$ ace -g erg-2018-x86-64-0.9.34.dat -e -v the-cat-slept.mrs
NOTE: loading frozen grammar ERG (2018)
NOTE: semantic index hash contains 29743 entries in 65536 slots
NOTE: max-ent model hash contains 464219 entries in 1048576 slots
NOTE: 10439 types, 40320 lexemes, 362 rules, 67 orules, 108 instances, 49510 strings, 233 features
permanent RAM: 3k


external MRS:
[ LTOP: h0 INDEX: e1 [ e SF: prop TENSE: past MOOD: indicative PROG: - PERF: - ] RELS: < [ _the_q<0:3> LBL: h2 ARG0: x3 [ x PERS: 3 NUM: sg IND: + ] RSTR: h4 BODY: h5 ]  [ _cat_n_1<4:7> LBL: h6 ARG0: x3 ]  [ _sleep_v_1<8:14> LBL: h7 ARG0: e1 ARG1: x3 ] > HCONS: < h0 qeq h7 h4 qeq h6 > ICONS: < > ]

internal MRS:
[ LTOP: handle0 INDEX: event1 [ event SF: prop E.TENSE: nonpresent E.MOOD: indicative E.ASPECT.PROGR: - E.ASPECT.PRF: - ] RELS: < [ _the_q<0:3> LBL: handle2 ARG0: ref-ind3 [ ref-ind PNG.PN: 3s IND: + ] RSTR: handle4 BODY: handle5 ]  [ _cat_n_1<4:7> LBL: handle6 ARG0: ref-ind3 ]  [ _sleep_v_1<8:14> LBL: handle7 ARG0: event1 ARG1: ref-ind3 ] > HCONS: < handle0 qeq handle7 handle4 qeq handle6 > ICONS: < > ]

mrs-to-dg 0.0002
fixup result:
[ LTOP: handle0 INDEX: event1 [ event SF: prop E.TENSE: nonpresent E.MOOD: indicative E.ASPECT.PROGR: - E.ASPECT.PRF: - ] RELS: < [ _the_q<0:3> LBL: handle2 ARG0: ref-ind3 [ ref-ind PNG.PN: 3s IND: + ] RSTR: handle4 BODY: handle5 ]  [ _cat_n_1<4:7> LBL: handle6 ARG0: ref-ind3 ]  [ _sleep_v_1<8:14> LBL: handle7 ARG0: event1 ARG1: ref-ind3 ] > HCONS: < handle0 qeq handle7 handle4 qeq handle6 > ICONS: < > ]
added 131 grammar entities
semlook 0.0039
packed 0.0015
probability = 1.000000:
The cat slept.
1 hyps / 1 reconstructed / 1 readings
5 decompositions produced for 5 passive edges
unpacked 0.0014
NOTE: 61 passive, 159 active edges in final generation chart; built 61 passives total. [1 results]

NOTE: generated 1 / 1 sentences, avg 628k, time 0.00706s
5 total hypotheses generated
5 total nodes reconstructed
NOTE: glb hash: 0 direct hits, 0 collisions, 4605 misses
NOTE: 541 subsumption tests; qc filters 92.8% leaving 39, of which ss passes 0.0% = 0 ; 7.7% = 3 generalizable
NOTE: unify filters: 6370 total, 1766 rf (27.7%), 141 qc (2.2% / 8.0%), 89 success (1.4% / 63.1%), 0 bad orth (0.0% / 0.0%)
NOTE: 5 / 61 (8.2%) passive edges were connected to roots
TIMERS (7 calls = ~ 1.0µs overhead):
mrs printing                    	16.8µs	for 3 events = 5.6µs per event
extract QC                      	0.0µs
main unify                      	0.0µs
build edge                      	0.0µs
hypothesize                     	0.9µs	for 2 events = 0.4µs per event
instantiate                     	444.9µs	for 1 events = 444.9µs per event
mrs                             	115.9µs	for 1 events = 115.9µs per event
NOTE: transfer did 194 successful unifies and 172 failed ones
1 Like

Thank you, @goodmami – that was exactly what I was looking for!

I’m pretty sure passing enough -v flags will get you that — did you try that? My memory could be wrong. If that doesn’t work then I doubt there is another flag to do it.

Thanks – it did!