this is for holding javascript data
AT deleted To_maintain_a_high.tex
almost 9 years ago
Commit id: 28b5e72d6a4099692b49026d4c3d6d77b298b685
deletions | additions
diff --git a/To_maintain_a_high.tex b/To_maintain_a_high.tex
deleted file mode 100644
index e51497e..0000000
--- a/To_maintain_a_high.tex
+++ /dev/null
...
To maintain a high precision level of answers, the Syphon system should be
conservative about the questions it allows. For the baseline system,
in light of the analysis of question classes above,
we want to propose a simple Syphon which scans the question
for named entities%
\footnote{\textit{Named entity} in NLP parlance is something that is
not a plain English word carrying meaning, but e.g.\ a proper
name or a numerical value (date, monetary amount, \dots).
There exist specialized scanners that extract named entities
of specific types; as a simple baseline, we may consider
all strings that are titles of Wikipedia articles (or redirects,
that is essentially unique aliases). For example
\textit{Hillary Clinton} and \textit{President of the United States}
both are titles of such articles.}
and verbs which describe clear relationships
and verifies that no words without an assigned role remain
in the question. This should keep a fair share of the sample questions
covered and with a clear interpretation.
Fig.~\ref{fig:syphon} shows a simple visual concept of the initial Syphon.
diff --git a/layout.md b/layout.md
index 1b65108..7cd9ef5 100644
--- a/layout.md
+++ b/layout.md
...
sectionStructure_of_.tex
beginitemize__item_C.tex
sectionHow_Can_We_Kn.tex
To_maintain_a_high.tex
sectionA_Concrete_Pr.tex
beginitemize__item_S.tex
sectionImplementatio.tex