[Attempto] Source Code of RACE reasoner

[Norbert E. Fuchs]

On 21 Sep 2012, at 15:24 , Huan Nguyen wrote:

> I am just so curious that whether you have any plan to extend ACE with temporal operators (which supported by temporal logics)? currently I think this is the only considerable limitation of this cool language.

Dear Huan

I take the liberty to send this message also to the Attempto mailing list.

In earlier implementations of ACE we had times, events and states. Citing for example from our earliest publication

N. E. Fuchs, R. Schwitter. Specifying Logic Programs in Controlled Natural Language, Proceedings CLNLP 95, ELSNET/COMPULOG-NET/EAGLES Workshop on Computational Logic for Natural Language Processing, University of Edinburgh, April 1995.

> DRT with Eventualities (DRT-E)
> 
> DRT-E investigates further details of the semantics of verbs, taking into account the theory of underlying eventualities (events or states) and handling temporal and aspectual information [Kamp & Reyle 93, Reichenbach 47]. A prototypical implementation of DRT-E is described in [Brown 94].
> 
> Investigating sentences like
> 
>      The customer enters the card.
>      SimpleMat checks the card.
> 
> and
> 
> Every customer has a personal code.
> 
> we recognise that the first two sentences are naturally understood as a report of temporally ordered events, while the second sentence describes something like a condition or state.
> 
> Verbs like enter or check introduce the existence of an event in much the same way as a noun phrase introduces the existence of an object. Events involve some kind of change in the universe of discourse, they persist through a certain interval of time and come eventually to a culmination point. They imply that some non-temporal condition, which is true when the event starts, is terminated by the event, and is replaced by further events.
> 
> States differ from events. A state verb such as have expresses a quality that is true indefinitely – it involves the continuation of a condition.
> 
> Sometimes the distinction between state-sentences and event-sentences is recognisable from the syntactic form of the verb, but it is well known that it is not the verb alone which decides about the eventuality introduced by a new sentence. The different thematic roles may exert a major influence.
> 
> In our approach, we represent the statement that E1 is the event of X1 entering X2 as enter(E1,X1,X2) and use the special predicate cul(E1,T1) to express that the event E1 culminates at time T1. The relation between the reference time T1 of E1 and the speech time N is established with the help of the additional predicate at(T1,N). With these notational changes, the DRS looks like
> 
> N E1 T1 X1 X2 E2 T2 X3 X4
> customer(X1) 
> card(X2) 
> enter(E1,X1,X2) 
> cul(E1,T1) 
> at(T1,N) 
> named(X3,simplemat) 
> card(X4) 
> check(E2,X3,X4)
> X4 = X2
> cul(E2,T2) at(T2,N)
> 
> As mentioned above, a state is temporally extended and homogenous. It describes a static situation S that holds or does not hold at a given time T. The function of the verb have is twofold: it introduces a discourse referent S which represents a state of affair and it provides a descriptive characterisation of this state represented by S. We will retain this information as a predicate of the form have(S,X1,X2). The additional predicate hold(S,T) asserts that S holds at T and the condition at(T,N) indicates that the eventuality described is located at the same time as the utterance time of the discourse that the DRS is taken to represent.
> 
> X1
> customer(X1)
> =>
> N S T X2
> personal_code(X2) 
> have(S,X1,X2) 
> hold(S,T) 
> at(T,N) 

For some time we even considered introducing temporal operators like "before", "after" and "while", but we never did for two reasons: first, classifying verbs as event or state is not easy – see above citation – and second, the development of ACE and its tools was often driven by the projects we were involved in – in recent years the semantic web where there is no need for temporality. Thus at some time we decided to remove times, events and states from the language, concretely from the DRS representation derived from ACE texts, and we simplified the ACE parser accordingly. We have no plans to reintroduce these features.

Now you have two possibilities: First, reintroduce yourself temporality into ACE and the DRS language and modify the ACE parsing engine APE accordingly. Second, use prepositional phrases to express temporality which was done years ago in an Israelian project for signal processing. Here is a modification of the above example

A customer enters a card at a time T1.
SimpleMat checks the card at a time T2 and T2>T1.

Best regards.

   --- nef