AND SEE ALSO LEHNERT, 1978). IN ORDER TO GENER-THE CORRECT ANSWE...

1975, and see also Lehnert, 1978). In order to gener-

The correct answer is “1503”, which is in 4

th

ate meaningful auxiliary questions and constraints,

place, with a low confidence score. Using QA-by-

we need a model (“script”) of the situation the ques-

Dossier, we ask two related questions “ When was

tion is about. Among others, we have identified one

Leonardo da Vinci born?” and “ When did Leonardo

such script modeling the human life cycle that seems

da Vinci die?” The answers to these auxiliary ques-

common to different question types regarding peo-

tions are shown in Table 2.

ple.

Given common knowledge about a person’s life

expectancy and that a painting must be produced

3 Introducing QDC

while its author is alive, we observe that the best

dates proposed in Table 2 consistent with one an-

QA-by-Dossier-with-Constraints is an extension

other are that Leonardo da Vinci was born in 1452,

of on-going work of ours called QA-by-Dossier

died in 1519, and painted the Mona Lisa in 1503.

(QbD) (Prager et al., 2004). In the latter, defini-

[The painting date of 1490 also satisfies the con-

tional questions of the form “Who/What is X” are

straints, but with a lower confidence.] We will ex-

answered by asking a set of specific factoid ques-

amine the exact constraints used a little later. This

tions about properties of X. So if X is a person, for

example illustrates how the use of auxiliary ques-

example, these auxiliary questions may be about

tions helps constrain answers to the original ques-

important dates and events in the person’s life-cycle,

tion, and promotes correct answers with initial low

as well as his/her achievement. Likewise, question

confidence scores. As a side-effect, a short dossier

proach, and, for each question type, develop a set of

is produced.

auxiliary questions and appropriate constraints

among the answers. Third, for each question type,

we must determine how the results of applying con-

Score Born Score Died

straints should be utilized.