SECTION 4 PRESENT EVALUATION RESULTS. IN SECTION 5 WE DISCUSS OUR CONC...

1977). CHAT-80 (Warren & Pereira, 1982), for in-

left, QS1, is our basic QA system, in which the

stance, was a DCG-based NL-query system about

Q

UESTION

P

ROCESSING

(QP), S

EARCH

(S) and

world geography, entirely in Prolog. In these

A

NSWER

S

ELECTION

(AS) subcomponents are indi-

systems, the NL question is transformed into a se-

cated. The outer block on the right, QS2, is another

mantic form, which is then processed further. Their

QA-System that is used to answer the inverted ques-

overall architecture and system operation is very

tions. In principle QS2 could be QS1 but parameter-

different from today’s systems, however, primarily

ized differently, or even an entirely different system,

in that there was no text corpus to process.

but we use another instance of QS1, as-is. The

block in the middle is our Constraints Module CM,

Inferencing is a core requirement of systems that

which is the subject of this paper.

participate in the current PASCAL Recognizing

Textual Entailment (RTE) challenge (see

https://traloihay.net and

.../RTE2). It is also used in at least two of the more

QS2

QS1

QA system

Question

CM

QP

QP

constraints

question proc.

module

S

search

AS

answer selection

Answers

Figure 1. Constraints Architecture. QS1 and QS2 are (possibly identical) QA systems.

The Question Processing component of QS2 is not

(6) “<C

AND

A

NS

> was the 33

^rd

what of the U.S.?”

used in this context since CM simulates its output by

modifying the output of QP in QS1, as described in

Having more than one possible inversion is in theory