You are on page 1of 6

A Novel Perspective of Common Sense

Reasoning
in e-Examination
M.Nandhini
#1
!.!h"vanes#ari
#$

#1
Assistant Professor,
Department of Computer Science, Pondicherry University
Puducherry, India-14.
#
Department of Computer Science, Pondicherry University
Puducherry, India-14.
mnandhini2005@yahoo.com, ammu.bhuvi.1989@gmail.com
Abstract Over a period of time, the researchers
have been studied the theoretical parts of how to
represent common sense knowledge. However, very
few methodologies only have been identified to
explore the fact of common sense knowledge in
artificial intelligence. The common sense reasoning
has been studied in different areas, such as
philosophy, psychology, and computer science. In
this paper, a new methodology for evaluating online
examination using common sense reasoning is
proposed. Thereby, different areas of common sense
reasoning are identified and able to achieve the
obective with the help of natural language
processing. Hence, the effect and efficiency of
common sense reasoning in academic is analy!ed
with set of facts.

Keywords - "ommon sense reasoning, #rtificial
intelligence, e$examination, natural language processing,
academic.
%. %N&R'()C&%'N
Artificial %ntelligence is the "se of programs to ena*le
machines to perform tas+s #hich h"mans perform "sing
their intelligence.
,ig 1. Artificial intelligence
&he "sage of artificial intelligence is to solve the
pro*lems "nder vario"s categories s"ch as expert
s-stems . +no#ledge *ased s-stems +no#ledge
representation . management +no#ledge ac/"isition
m"lti-agent s-stems and distri*"ted A% intelligent
organi0ations nat"ral lang"age processing ontologies
intelligent interfaces1 m"ltimedia virt"al realit-
comp"ter vision . image processing ne"ral net#or+s
"ncertaint- . pro*a*ilistic reasoning model-*ased
reasoning non-monotonic reasoning common sense
reasoning case *ased reasoning spatial and temporal
reasoning constraint programming logic programming
a"tomated theorem proving ro*otics planning and
sched"ling pattern recognition /"alitative reasoning
213.
,ig $. &as+s of Artificial %ntelligence.
%n this paper are handling the concept of nat"ral
lang"age processing "sing common sense in order to
ma+e the s-stem to "nderstand the common sense *ased
data and to reason it *- "sing its +no#ledge.
&he nat"ral lang"age processing gives machines the
a*ilit- to read and "nderstand the lang"ages that
h"mans spea+. Man- researchers hope that a s"fficientl-
Proceedings of National Conference on Advanced Computing and CommunicationNCACC11, April.1, !"11
po#erf"l nat"ral lang"age processing s-stem #o"ld *e
a*le to ac/"ire the +no#ledge on its o#n *- reading the
existing text availa*le over the internet 2$3. Some
straightfor#ard applications of nat"ral lang"age
processing incl"de information retrieval 4or text mining5
and machine translation.
Nat"ral lang"age "nderstanding re/"ires a large
amo"nt of *ac+gro"nd 6common sense7 +no#ledge
a*o"t the sit"ation "nder disc"ssion. %n man- respects
"sing this +no#ledge is at the core of reasoning and
acting in traditional Artificial intelligence. &he parent
disciplines the s"*8ect covered and the applications of
common sense reasoning are narrated thro"gh the ,ig9.
,ig 9. Common Sense Reasoning - Applications
,or example #hen a reader 4#ho generates /"estion5
read an article a*o"t a criminal conviction the #riter
4s-stem: one sho"ld give repl-5 #ho gives the res"lt
from the traits and facts perceived form the environment
thro"gh N;P. &he narrative chain pro8ect of N;P aims
to learn this +no#ledge *- processing large amo"nt of
text speech and learning in #hich events to occ"r
together. An example of learned narrative event chain is
sho#n in ,ig<.
=ere it is concerned #ith A% ena*ling program to *e
a*le to process learn and reason #ith the common
+no#ledge to *e a*le to excel in a #ide range of non-
specific information tas+s. &his goal "nfort"natel-
re/"ires good proced"res #hich are the +no#ledge
representation and common sense:#orld +no#ledge.
%n theor- if the comp"ter is endo#ed #ith good
+no#ledge representation data*ase incl"ding a
comprehensive common sense data*ase it is a*le to
process and respond in plain text English also have the
a*ilit- to process and reason #ith English texts. &he
tas+ of common sense reasoning is pro*a*l- the least
researched area tho"gh it is incl"ded in man- #a-s in
+no#ledge representation tas+.
%n this paper foc"sing a*o"t ho# the s-stem
generates set of /"estions *- sh"ffling from the /"estion
*an+ and ho# to eval"ate the ans#ers *- "sing the
ans#er +e- #hich is generated a"tomaticall- *- the
s-stem #ith its common sense reasoning in order to
avoid an- mischievo"s activities and "tili0e the
efficienc- of the s-stem.
,ig <. ;earned narrative event chain
%%. C'MM'N SENSE REAS'N%N>
A. !vervie"
Common sense reasoning is the *ranch of
Artificial intelligence concerned #ith replicating
h"man thin+ing. &here are several components to
this pro*lem incl"ding1
(eveloping ade/"atel- *road and deep
common sense +no#ledge *ases.
(eveloping reasoning methods that exhi*it
the feat"res of h"man thin+ing incl"ding
a*ilit- to 2931
o Reason #ith +no#ledge that is
tr"e *- defa"lt.
o Reason rapidl- across a *road
range of domains.
o &olerate "ncertaint- in -o"r
+no#ledge.
o &a+e decisions "nder incomplete
+no#ledge and perhaps revise that
*elief or decision #hen complete
+no#ledge *ecomes availa*le.
(eveloping ne# +inds of cognitive
architect"res that s"pport m"ltiple
reasoning and methods.
&he factors #hich "se the common sense reasoning
are explained in fig ?.
Common sense reasoning is re/"ired for solving the
pro*lems in the common sense #orld. &he explanation
is that the common sense #orld is characteri0ed *- a
different informatics sit"ation than that #ithin an-
formal theor- 293. %n order to ma+e comp"ters perform
common sense reasoning in the same #a- as #e do
need to a"tomate the common sense reasoning 2<3. As
defined in 2?3. 2@3 common sense reasoning is a
process that involves the comp"ters in ta+ing
information a*o"t certain aspect of a scenario of #orld
$
Ad#iparasa$t#i %ngineering College, &elmaruvat#ur, 'epartment of Computer (cience ) %ngineering
and ma+ing inferences a*o"t other aspects of the
scenario *ased on o"r common sense reasoning or
+no#ledge of ho# the #orld #or+s.
%n s"mmar- tho"gh man- of the proposals dealt the
iss"es in the common sense reasoning 2A3 2B3 some of
the open iss"es are not -et addressedC pro*a*l- no
proper mechanism for ma+ing the s-stem to possess
proper res"lt a*o"t the common sense *ased pro*lems
d"ring examination no s"ita*le model for the reasoning
o"t the pro*lems no proper g"idance for dealing the
common sense *ased pro*lems and so on.
=ence from these perspectives in this #or+ #e
proposed a different common sense reasoning areas that
can *e applied to examinations. %n this paper #e are
also proposing a*o"t ho# the comp"ters reason o"t the
ans#ers for the /"estions d"ring the examinations.
&h"s the common sense reasoning can *e applied to the
comp"ters *- "sing the nat"ral lang"age processing and
the +no#ledge representation techni/"es to ma+e the
comp"ters to reason o"t for an- /"er-.
=ence it can *e achieved *- feeding the semantic
+no#ledge to the s-stem and can ma+e the s-stem to
"nderstand the text #hich is given as the inp"t and to
ma+e them it to reason o"t the /"estion #ith a correct
o"tp"t.
Dhile to the average personEs the term Fcommon
sense F is regarded as s-non-mo"s #ith Fgood
8"dgementE to the A% comm"nit- it is "sed in a
technical sense 21G3 to refer to the millions of *asic facts
and "nderstandings possessed *- most people.
Common sense reasoning is th"s the h"ge amo"nt of
h"man experience encompassing +no#ledge a*o"t the
spatial ph-sical social temporal and ps-chological
aspects of t-pical ever-da- life. !eca"se it is ass"med
that ever- person possesses common sense s"ch
+no#ledge is t-picall- omitted from social *ehavio"r.
A f"ll "nderstanding of an- text re/"ires a s"rprising
amo"nt of common sense #hich c"rrentl- onl- possess.
%t is o"r p"rpose to find #a-s to provide s"ch common
sense to machines2H3. &o do so ne# concept has *een
proposed in #hich the comp"ter generates the /"estions
to the st"dents d"ring the online examination and the
s-stem itself "ses its common sense reasoning to find
the correct ans#er for the /"estion and to store the
correct ans#er in its data*ase #hich can *e later "sed
for eval"ating the ans#ers and prod"cing the res"lts.
%t can *e done *- feeding the *oth the s-ntactic and
semantic data to the s-stem memor- and to ma+e the
s-stem to "nderstand the common facts *- appl-ing the
facts #hich is alread- stored in memor-.
&his is possi*le *- "sing a large scale common sense
reasoning *ase #ith an integrated nat"ral lang"age
processing that s"pports man- practical reasoning tas+s
over real #orld iss"es. &o do so the initial stage is to
ma+e the text as a sensi*le one.
#. $a%in& sense of te't
'f sorts of different semantic +no#ledge that are
researched Since comp"ters do not possess common
sense reasoning it is "nderstanda*le #h- the- #o"ld *e
so *ad at ma+ing sense of text"al information.
A comp"ter can pla- chess /"ite #ell -et it cannot
even "nderstand a simple childrenEs stor-.
>iven a sentence F% ate some chips #ith m- l"nchE
is a common sense-deprived nat"ral lang"age
"nderstanding s-stem is not li+el- to +no# that FchipsE
pro*a*l- refer to Fpotato chipsE and pro*a*l- not
Fcomp"ter chipsE.
Dhile +e-#ord-spotting s-ntactic lang"age parsing
and statistical methods have all assisted in text"al
anal-sis there is a little s"*stit"te for the
comprehensiveness and ro*"stness of interpretation
afforded *- large-scale common sense.
Ditho"t common sense a comp"ter might *e a*le to
g"ess that the sentence F% had an a#f"l da-E is negative
*- spotting the mood +e-#ord Fa#f"lE *"t given the
sentence F% got fired toda-E the comp"ter reader #o"ld
not +no# #hat to thin+.
%n contrast a common sense +no#ledge *ase sho"ld
*e a*le to reason a*o"t the sit"ation of a person Fgetting
firedE. Perhaps it +no#s some things a*o"t Fgetting
firedEC people sometimes get fired *eca"se the- are
incompetent. A possi*le conse/"ence of getting fired is
not having mone-. People need mone- to pa- for food
and shelter.
Even if the +no#ledge *ase does not have direct
affective +no#ledge a*o"t Fgetting firedE tho"gh its
net#or+ of related +no#ledge it sho"ld a*le to sense the
reason and the sit"ation of Fgetting firedE "s"all- *ears
man- negative connections s"ch as fear anger and
sadness.
C. Imp(ementin& the optimi)ed reasonin&
system
&he scope of implementing the optimi0ed reasoning
to the s-stem is to ma+e the s-stem in s"ch a #a- that it
sho"ld *e a*le to compare #ith #hat are in o"r opinion
a*o"t a partic"lar statement. =o#ever the +e-
differences #ill *e spelled o"t in the follo#ing section.
%n order to optimi0e the lexical characteri0ation and
#ord similarit- determination the common sense
reasoning is optimi0ed for the formali0ed logical
reasoning for ma+ing the practical context-*ased
inferences over real-#orld texts.
&hat it reasons simpl- and gracef"ll- over text is
perhaps o#ed to the fact that its +no#ledge
representation is itself semi-str"ct"red English.
Common sense reasoning is more or less same li+e the
context"al reasoning. %n context"al common sense
reasoning it is highl- applica*le to text"al information
management *eca"se it allo#s a comp"ter to *roadl-
categori0e texts along interesting dimensions s"ch as
topic and affect. &he nat"ral lang"age allo#s a
comp"ter to "nderstand novel or "n+no#n concepts *-
emplo-ing str"ct"ral analogies to sit"ate them #ithin
#hat is alread- +no#n.
%%%. (ES%>N ', PR'P'SE( SIS&EM
A. !n(ine *'amination
%n the online examination proposed to organi0e a
competitive exam in #hich a*le to test the candidateEs
9
Proceedings of National Conference on Advanced Computing and CommunicationNCACC11, April.1, !"11
performance in the field of act according to sit"ation
logical se/"encing and reasoning classification
decision ma+ing assertion and reason ver*al reasoning
co"rses of action and theme detection voca*"lar- and
no"n phrases2113.
&he /"estions for the online exam #ill *e generated
*- the s-stem from its data*ase accordingl- to the
candidateEs mental a*ilit- reasoning po#er intelligent
approach and the time ta+en for ans#ering the previo"s
/"estion for #hich the s-stem has its o#n set of
ans#ers #hich is derived from the s-stem memor- *-
"sing its a*ove said common sense reasoning. &he
/"estions #ill *e *ased on the m"ltiple-choice
/"estions. Each /"estion #ill *e given #ith fo"r
ans#ers in #hich onl- one #ill *e the correct ans#er. %f
the "ser gives the #rong ans#er the s-stem #ill
prod"ce the correct ans#er *- retrieving it from its
data*ase.
%nitiall- the ans#ers #ill *e stored in the data*ase.
&he ans#ers for the /"estions are derived *- the s-stem
*- "sing its common sense reasoning and +no#ledge
representation techni/"es.&he common sense reasoning
ma- *e classified into different areas *- "sing #hich
the s-stem can a*le to find the relevant ans#er for ever-
/"estion. %t is *riefl- explained in the s"cceeding
section. ;ater the ans#ers #ill *e eval"ated *- referring
the ans#er +e- #hich is stored in the data*ase and
prod"ce the res"lts of the exam.
#.Areas of common sense reasonin&, app(ied
in e-e'amination
&he common sense reasoning ma- *e s"*divided into
different areas.
1+ ,er-a( reasonin&. 'ne of the ma8or areas is
that the primaril- no"ns ver*s and ad8ectives organi0ed
onto discrete FsensesE that are lin+ed to the semantic
relations s"ch as the s-non-m relation and F%sAE
=ierarchical relation. ,or example "sing a ver* 4drin+5
#ith a no"n phrase 4coffee5 and in a preposition phase
4in morning5 to represent a semi str"ct"red English
#hich comprises of common sense reasoning #hich
inc"lcates the concept of semantic net#or+ and is given
in ,ig.@. &he ver*al reasoning is often "sed
interchangea*l- #ith Fver*al a*ilit-E and Fver*al
aptit"deE. &here is no #idel- accepted definition of the
difference *et#een ver*al a*ilit- and ver*al aptit"de
and as far as ps-chometric tests are concerned the t#o
terms are interchangea*le. %t can *e classified into
ver*al speed tests 4Spelling and grammar5 and ver*al
po#er tests 4Critical reasoning5 as sho#n in ,ig A.
%ll"stration1 1
Pedro goes either h"nting or fishing ever- da-. %f it is
sno#ing . #ind- then Pedro goes h"nting. %f it is s"nn-
and not #ind- then Pedro goes fishing. Sometimes it
can *e sno#ing and s"nn-.
Dhich of the follo#ing statements m"st *e tr"eJ
%f it is not s"nn- and it is sno#ing then Pedro
goes h"nting.
%f it is #ind- and Pedro does not go h"nting
then it is not sno#ing.
%f it is #ind- and not s"nn- then Pedro goes
h"nting.
%f it is #ind- and s"nn- then Pedro goes
h"nting.
%f it is sno#ing and s"nn- then Pedro goes
h"nting.
,ig @. Example of ver*al reasoning.
,ig A. Classification of Ker*al Reasoning
,rom the a*ove example the s-stem gets the details
of Pedro #ho goes to fishing dail- and if it seems to *e
sno#ing and #ind- the person #ho named Pedro #ill
go for h"nting rather than fishing. ;i+e#ise if it is of
s"nn- and not #ind- the person goes to fishing.
So from the fetched details the s-stem recogni0e that
the #ord Pedro #hich is the name of a person *eca"se
the #ord Pedro is the no"n. !- "sing no"n #e +no# the
#or+ #hich #as done *- the person. !- chec+ing all the
logics it #ill prod"ce the correct ans#er as
If it is "indy and Pedro does not &o huntin& then it is
not sno"in&.
+ Act accordin& to situation. Consider the
small s"*set of the concepts #hich allo#s "s to act
according to the sit"ation. %t is f"rther classified into one
of the six affect categories +no#n as happ- sad angr-
fearf"l disg"sted and s"rprised. &he effect of an-
"nclassified concept can *e assessed *- finding all the
paths #hich lead "s to each of these categories and then
8"dging the strength and fre/"enc- of each set of paths
<
Ad#iparasa$t#i %ngineering College, &elmaruvat#ur, 'epartment of Computer (cience ) %ngineering
as in ,ig.B. %n sit"ation reaction test the certain
sit"ation is descri*ed and needs to choose the most
s"ita*le reaction to the given sit"ation amongst the
alternatives provided.
,ig B. Example for act according to the sit"ation.
I((ustration.
Dhile travelling in -o"r car certain persons stop -o"
on the #a- as+ing -o" to ta+e an in8"red child to the
hospital. Dhat #ill *e -o"r reaction for thisJ2193.
As+ them to leave -o"r #a- and then drive a#a-.
As+ them to first call the police
%mmediatel- ta+e the child to the hospital
>et o"t of the car and as+ some other person to help
,rom the s-stem point of vie# the s-stem #ill
recogni0e the car as an o*8ect and a person is travelling
in the car. Since a child gets in8"red some *od- is
as+ing the child to ta+e him to the hospital. So the
s-stem #ill diagnose the child as a h"man and reali0e
the sit"ation is a critical sit"ation. So according to the
sit"ation it #ill respond that the child to ta+en to the
hospital immediatel-. %t is all done *- comparing the
semantic data #ith the different o*8ects and the
sit"ation.
,rom the a*ove sit"ation descri*ed it demands that
the person sho"ld immediatel- render the help as+ed for
and ta+e the child to the hospital and prod"cing the
ans#er as
Immediate(y ta%e the chi(d to the hospita(.
/+ Assertion and reasonin&. %n order to
implement the assertion in common sense and
reasoning #e need to process the FRelaxation phaseE
over the net#or+ meant to smooth over the semantic
gaps and to improve the connectivit- of the net#or+.
%nitiall- d"plicate the assertions are merged 4since man-
common factors are "ttered m"ltiple times5 and an
additional Meta data field called Ffre/"enc-E is added to
each predicate-relation to trac+ ho# man- times
something is "ttered. Secondl- the F%sAE hierarchical
relation is "sed historicall- to FliftE the +no#ledge from
the child node to the parent node. An example of this is
given *elo#21?3.
24%sA FappleE Ffr"itE5C
4%sA F*ananaE Ffr"itE5C
4%sA FpeachE Ffr"itE53
AN(
24Propert-'f FappleE Fs#eetE5C
4Propert-'f F*ananaE Fs#eetE5C
4Propert-'f FpeachE Fs#eetE53
%MP;%ES
4Propert-'f Ffr"itE Fs#eetE5
%K.C'NC;)S%'N
%n this paper #e have presented the #a- for ma+ing
the s-stem to "nderstand the text in a meaningf"l
manner and to prod"ce the res"lts for it *- "sing its
common sense reasoning techni/"es. &his #or+
s"pports the practical text"al reasoning tas+s incl"ding
semantic disam*ig"ation and classification analog-
ma+ing and so on. %t has a simple str"ct"re and its
"nderl-ing representation is *ased on nat"ral lang"age
fragments ma+ing it partic"larl- #ell s"ited to text"al-
reasoning pro*lems. Motivated *- the range of concepts
availa*le *ased on common sense reasoning the
concept reflects far richer set of concepts and semantic
relations. Common sense reasoning is a research area
that is poised to redefine the possi*ilities for intelligent
information management. ,"rthermore the common
sense reasoning research #ill *e expanded to prod"ce
the consistent and sophisticated res"lts #hich #o"ld *e
carried o"t from the academic point of vie#.
RE,ERENCES
213 Charnia+ E. . Mc(ermott (. Introduction to Artificia(
Inte((i&ence Reading MA1 Addison Desle- 1HB?.
2$3 ;"ger >eorgeC St"**lefield Dilliam ?th Ed Artificial
%ntelligence1 Str"ct"res and Strategies for Complex Pro*lem
Solving. &he !en8amin:C"mmings P"*lishing Compan- $GG<.
293 R"ssell St"art L.C Norvig Peter $nd Ed Artificial %ntelligence1
A Modern Approach )pper Saddle River Ne# Lerse-1
Prentice =all $GG9.
2<3 Amir E. . Ma-nard-Reid P. 6;iSA1 A Ro*ot (riven *-
;ogical S"*s"mption7 ,ifth S-mposi"m on the ;ogical
,ormali0ation of Commonsense Reasoning $GG1.
2?3 !oden M. 6Creativit- and Comp"ters7 in Artificial
%ntelligence and Comp"ters (artnall &. ed. (ordrecht the
Netherlands1 Ml"#er 1HH< pp. 9-$@.
2@3 !rachman R. . ;eves/"e =. Mno#ledge Representation and
Reasoning San ,rancisco CA1 Morgan Ma"fmann:Elsevier
$GG<.
2A3 !rings8ord S. 6Animals Nom*animals and the &otal &"ring
&est1 &he Essence of Artificial %ntelligence7 Lo"rnal of ;ogic
;ang"age and %nformation $GGG H1 9HA-<1B.
2B3 Mlen+ M. ,or*"s M. &omai E. Mim=. and M-c+elhahn
!.6Solving Ever-da- Ph-sical Reasoning Pro*lems *-
Analog- "sing S+etches.7 Proceedings of $Gth National
Conference on Artificial %ntelligence 4AAA%-G?5 Pitts*"rgh
PA $GG?.
2H3 M"eller E. Commonsense Reasoning San ,rancisco CA1
Morgan Ma"fmann $GG@.
21G3 Rapaport D. 6S-ntactic Semantics1 ,o"ndations of
Comp"tational Nat"ral-;ang"age )nderstanding7 in Lames =.
,et0er Aspects of Artificial %ntelligence (ordrecht the
Netherlands1 Ml"#er Academic P"*lishers 1HBB.
2113 Pearl L. Pro*a*ilistic Reasoning in %ntelligent S-stems San
Mateo CA1 Morgan Ma"fmann 1HBB.
21$3 Polloc+ L. 6(efeasi*le Reasoning #ith Karia*le (egrees of
L"stification7 Artificia( Inte((i&ence, $GG1.
2193 Nilsson Nils Artificial %ntelligence1 A Ne# S-nthesis. Morgan
Ma"fmann P"*lishers 1HHB.
21<3 !roo+s R. A. 6%ntelligence #itho"t Representation7 Artificia(
Inte((i&ence, 1HH1.
?
Proceedings of National Conference on Advanced Computing and CommunicationNCACC11, April.1, !"11
21?3 ,agin R. and =alpern L. and Moses I. and M. Kardi M.
0easonin& A-out 1no"(ed&e Cam*ridge MA1 M%& Press
$GG<.
@

You might also like