Information Technology Reference
In-Depth Information
Fig. 17.4
Graph example to answer
What causes lung cancer?
To compose an answer, we just have to gather the nodes and relationships point-
ing at the lung cancer node, e.g., in this case,
smoking
,
radon gas
and
tobacco
use.
The algorithm with this information performs the answer displayed in figure 17.5.
A simple interface, using Php and HTML languages, translates the graph into text:
Fig. 17.5
Automated answer to the question
What causes lung cancer?
17.3
Answering How-Questions
To address
how-questions
, the first step it to locate the causal paths connecting the
concepts involved in the question. How-questions usually involve two concepts in
the same question, the concept 'cause' and the concept 'effect'. Thus, if we ask
How X causes Y?
, the objective is to find the causal path connecting
X
(as head of
this path) to
Y
. Once the 'cause' concept has been located in the database (it will
appear in the Relationship table as cause_concept), a recursive algorithm finds the
possible paths connecting the concept cause with the concept effect. The main steps
of it are the following: