Performance of ERG-RMRS (erg-nomsg, 2006) vs RASP3-RMRS 1 It rained. RASP can't recognised `it' as expletive 2 Abrams barked. OK 3 The window opened. OK 4 Abrams chased Browne. OK 5 Abrams handed Browne the cigarette. OK 6 Abrams handed the cigarette to Browne. The rule here is |V1/v_np_pp| - this is also involved in 71 (`Abrams put Browne in the garden.') and 85 (`Abrams strikes Browne as old.') plus (potentially) `Abrams ran to the store.' These all behave differently semantically. The least bad approach seems to be to give the PP an adverbial modification analysis (which is OK for the last case). So for this example, to find what should be the ARG3, one has to go via the P. 7 Abrams bet Browne a cigarette that it rained. OK 8 Abrams knew that it rained. OK 9 Abrams intended to bark. Missing ARG1 of bark. 10 Abrams intended Browne to bark. Semantics here isn't same as ERG because Browne is taken as a separate argument. ARG1 of bark missing. 11 Every cat barked. OK 12 Every cat chased some dog. OK 13 My cat barked. OK 14 It barked. OK 15 The cat chased it. OK 16 The cat chased itself. OK 17 The cat chased one. This would be OK if we recognised non_temp_nom_rel a supertype of one_n_1_rel (which we currently don't but probably should). 18 Mine barked. OK 19 That opened. The RASP3-RMRS has a that_q without an associated nominal. To get this right would require having a subcase of S/np_vp for the case where the np is a bare DD1 - possible, but probably too much hassle to be worth it. 20 Cats bark. OK 21 Tobacco arrived. OK (though character alignment mismatch) 22 Some bark. No suitable RASP parse (but cf 19) 23 Some of the cats bark. OK, modulo spurious of_p_rel. However, this example only works if the index of the pp is set to be the noun index, rather than the p's event. Formally this is wrong, and it's possible it will cause problems with modification of Ps (although I don't think these are all that frequent). 24 No cat barked. OK 25 Did the dog bark? Nothing resembling a sensible parse for this from RASP 26 Which dog barked? OK 27 Whose dog barked? OK 28 Chase Browne! OK 29 Abrams wondered which dog barked. OK (Note use of underspecified ARG2-4 for the V1/v_s rule: which may not be necessary) 30 Abrams wondered whether Browne barked. I have chosen to give this the subordinating conjunction semantics (see examples 72 and 109). This isn't right, but we can't get exactly the right semantics, since that involves a null semantics conjunction. Giving the V1/v_pp rule (subord) the same semantics as the V1/vp_pp subord conj case will make things more robust since the two rules apply in the same places. 31 The dog that Browne chased barked. OK (Note use of underspecified ARG2-4) The N1/n_s rule in RASP is used both for relative clauses starting with `that' (in the ERG, the relative clause is coordinated with the noun) and for Ns like belief which subcat for a sentence (in the ERG, an ARG is created between the noun and the ltop of the sentence) There's no way currently of underspecifying the attachment. As things stand, we get the relative clause case right. 32 The dog to chase is barking. OK (Note use of underspecified ARG2-4) 33 The dog was chased by Browne. No connected arguments to chase and a spurious `by'. Getting this right (or at least better) would require a special form of S/np_vp for passives, maybe plus a special form |V1/v_pp|. A halfway position would be to leave in the `by' and assume that's used to find the agent. This will need to be fixed for Chemistry! 34 The dog chased by Browne barked. 35 The dog is barking. OK (check tense) 36 The dog has barked. OK (check tense) 37 The dog has been barking. OK (check tense) 38 The dog had been barking. OK (check tense) 39 The dog will bark. Modal semantics: getting this right would require a special entry for `will' 40 The dog is going to bark. 41 The dog could bark. OK 42 The dog couldn't bark. 43 The old dog barked. OK 44 The dog barked softly. OK apart from soft_a vs softly. The dog probably barked. The dog barked in the garden. The dog barks now. The garden dog barked. The tobacco garden dog barked. The cat is old. The cat is in the garden. The barking dog chased Browne. Chased dogs bark. Chasing the cat is old. That the cat chases Browne is old. Dogs chase whatever barks. The only RASP analysis of this with `bark' as a verb involves the V1/v_s rule The dog barked every day. When did the dog bark? Three of the dogs bark. Three bark. Browne's dog barks. Browne's barks. Twenty three dogs bark. Two hundred twenty dogs bark. Abrams arrived by car. Abrams kept barking. Browne squeezed the cat in. Browne squeezed in the cat. The picture of Abrams arrived. Abrams wiped the table clean. Abrams put Browne in the garden. The dog will bark if Browne arrives. Abrams and Browne arrived. Abrams, Browne and the dog arrived. The dog arrived and barked. The dog arrived and Browne barked. The dog barked didn't it? It is obvious that the dog barked. Abrams promised Browne to bark. Abrams seems to bark. Abrams believes Browne to be barking. It bothered Abrams that Browne barked. It took Abrams ten minutes to arrive. Abrams left it to Browne to bark. 85 Abrams strikes Browne as old. The semantics the ERG obtains is strike_v_as(e,B,old(A)), which is completely out of reach! Browne considers Abrams old. Abrams liked the idea that Browne could bark. Abrams barked from ten to three. Abrams was very old. Nearly every dog barked. Abrams barked very softly. Browne's chasing of cats bothered Abrams. It bothered Browne that Abrams chased cats. June third arrived. Abrams arrived at three twenty. Browne arrived on Tuesday morning. The cats found a way to bark. The happier dog chased Browne. There were cats in the garden. That dog chased Browne. Somebody chased Abrams. How happy was Abrams? The number five bothers Browne. Abrams could. Browne tried to. Don't bark! The dog arrived barking. The dog liked barking.