2 DDQ MODULEI.E., “WHAT KIND OF *

3.2 DDQ module

i.e., “What kind of

?”, “What year was

held?”

The word generality score

_G

was computed using

and “Where is

?”.

the same Mainichi newspaper text described above,

The DDQ module selects the best DQ based on its

while the SDCFG for the dependency ambiguity

linguistic appropriateness and the ambiguity of the

score

_D

for each phrase was the same as that used

phrase. The linguistic appropriateness of DQs can

in (C. Hori et. al., 2003). Eighty-two types of inter-

be measured by using a language model, N-gram.

rogative sentences were created as disambiguating

Let

mn

be a DQ generated by inserting the

-th

queries for each noun and noun-phrase in each ques-

phrase into the

-th template. The DDQ module

tion and evaluated by the DDQ module. The linguis-

selects the DQ that maximizes the DQ score:

tic score

indicating the appropriateness of inter-

_mn

_L

_mn

_D

_D

_n

_G

_G

_n

rogative sentences was calculated using 1000 ques-

tions and newspaper text extracted for three years.

where

is a linguistic score such as the loga-

The structural ambiguity score

_D

was calculated

rithm for trigram probability, and

_L

,

_D

, and

_G

based on the SDCFG, which was used for the screen-

are weighting factors to balance the scores.

ing filter.

Hence, the module can generate a sentence that

is linguistically appropriate and asks the user to dis-