Information Technology Reference
In-Depth Information
Once the graph is performed, it is stored in a database attending to the following
structure:
Fig. 17.3
Causal database structure
Causal information stored in terms of nodes and relationships will be useful to
find answers to what-questions. Next, we briefly summarize the overall process:
•
Locate the sought concept within the database and point to the records where it
is contained.
•
If the user is asking for causes, locate the records of the 'relationship' table with
the sought concept as effect, and retrieve the cause_concept information from
the 'Concept' table (besides location, specification and intensity attributes if ex-
ists).
•
If the user is asking for effects, locate the records of the 'relationship' table with
the sought concept as cause_concept, and retrieve the effect_concept informa-
tion (with all the attributes) from the 'Concept' table.
•
Compose each part of the answer 'translating' the information into these ob-
tained records, linking the records obtained by 'ands' and processing the infor-
mation contained; i.e., modifiers and quantifiers.
•
When composing the answer, evaluate the type of causal connector to make a
correct reading of the causal relationship and place the relationship modifier (if
it exists) in the right place, taking into account the type of causal connector. For
example, if the causal link is 'cause' in figure 17.4 you can say that
smoking
'cause'
lung cancer
; but if the causal connector is 'due to', the reading is upside
down: so, you can not say that
tobacco use is due to lung cancer but lung cancer
is due to smoking
.
Graphically, what this algorithm achieves is to mimic a causal graph without draw-
ing it, tracking the links contained in the database. For example, if the user asks
What causes lung cancer?, the following graph is deployed:
