A P ac ical Agen -Based Me hod o Ex ac
Seman ic In o ma ion om he Web
J. L. A jona, R. Co chuelo, A. Ruiz, and M. To o
Escuela T´ecnica Supe io de Ingenie ´
ia In o m´a ica de la Uni e sidad de Se illa
Depa amen o de Lenguajes y Sis emas In o m´a icos
A da. de la Reina Me cedes, s/n, Se illa, Spain
{a jona,co chu,a uiz,m o o}@lsi.us.es
Abs ac . The seman ic Web will b ing meaning o he In e ne , mak-
ing i possible o web agen s o unde s and he in o ma ion i con ains.
Howe e , cu en ends seem o sugges ha i is no likely o be adop ed
in he o hcoming yea s. In his sense, meaning ul in o ma ion ex ac-
ion om he web becomes a handicap o web agen s. In his a icle, we
p esen a amewo k o au oma ic ex ac ion o seman ically-meaning ul
in o ma ion om he cu en web. Sepa a ing he ex ac ion p ocess om
he business logic o an agen enhances modula i y, adap abili y, and
main ainabili y. Ou app oach is no el in ha i combines diffe en ech-
nologies o ex ac in o ma ion, su he web and au oma ically adap o
some changes.
1 In oduc ion
In ecen yea s, he web has consolida ed as one o he mos impo an knowl-
edge eposi o ies. Fu he mo e, he echnology has e ol ed o a poin in which
sophis ica ed new gene a ion web agen s p oli e a e. They enable efficien , p e-
cise, and comp ehensi e e ie al and ex ac ion o in o ma ion om he as
web in o ma ion eposi o y. A majo challenge o hese agen s has become si -
ing h ough an unwieldy amoun o da a o ex ac meaning ul in o ma ion. The
ollowing impo an ac o s con ibu e o hese difficul ies:
–Mos web pages a e a ailable in human- eadable o ms, and hey lack a
desc ip ion o he s uc u e o he seman ics associa ed wi h hei da a.
–The layou and he aspec o a web page may change unexpec edly. This
changes may in alida e he au oma ic ex ac ion me hods used so a .
–The access o he page ha con ains he in o ma ion in which we a e in e -
es ed may in ol e na iga ing h ough a se ies o in e media e pages, such as
login o index pages.
Se e al au ho s ha e wo ked on echniques o ex ac ing in o ma ion om
he web, and induc i e w appe s a e amongs he mos popula ones [2,6,7,9,10].
The wo k epo ed in his a icle was suppo ed by he Spanish In e -minis e ial
Commission on Science and Technology unde g an TIC2000-1106-C02-01
They a e componen s ha use a numbe o ex ac ion ules gene a ed by means
o au oma ed lea ning echniques such as induc i e logic p og amming, s a is ical
me hods, and induc i e g amma s. Al hough induc ion w appe s a e sui ed, hey
do no associa e seman ics wi h he da a hey ex ac .
The e a e also some ela ed p oposals in he field o da abases, e.g., TSIM-
MIS [5] and ARANEUS [8]. Thei goal is o in eg a e he e ogeneous in o ma ion
sou ces such as adi ional da abases and web pages so ha he use can wo k on
hem as i hey we e a homogeneous in o ma ion sou ce. Howe e , hese p opos-
als lack a sys ema ic way o ex ac in o ma ion om he web because ex ac ion
ules need o be implemen ed manually, which makes hem no scalable and un-
able o eco e om unexpec ed changes on he web.
In his a icle, we p esen a p oposal ha achie es a comple e sepa a ion
be ween he logic an agen encapsula es, he na iga ion pa h o he page ha
con ains he in o ma ion i needs, and way o ex ac i . Gi ing he necessa y
mechanisms o associa e seman ics wi h he ex ac ed in o ma ion and o adap
au oma ically o some changes on he web.
2 Ou P oposal
Ou p oposal aims a p o iding agen de elope s wi h a amewo k in which
hey can ha e access o seman ically-meaning ul da a ha esides on he e oge-
neous, use - iendly web pages. I elies on using a numbe o agen s ha we call
in o ma ion channels o IC o sho . They allow o sepa a e he ex ac ion o
in o ma ion om he logic o an agen , and offe agen de elope s a good deg ee
o flexibili y. In o de o allow o seman ic in e ope abili y, he in o ma ion hey
ex ac e e ences a numbe o concep s in a gi en applica ion domain ha a e
desc ibed by means o on ologies.
2.1 The A chi ec u e
Figu e 1ske ches he a chi ec u e o ou p oposal. As we men ioned abo e, in o -
ma ion channels a e a i s co e because hey specialise in ex ac ing in o ma ion
om diffe en sou ces, and a e able o eac o in o ma ion inqui ies ( eac i -
i y) om o he agen s (social abili y). They ac in he backg ound p oac i ely
acco ding o a p edefined schedule o ex ac in o ma ion and o main ain he
ex ac ion ules upda ed.
Each in o ma ion channel uses se e al induc i e w appe s o ex ac in o -
ma ion so ha hey can de ec inconsis encies amongs he da a hey ex ac .
I such inconsis encies a e ound, hey hen use a o ing algo i hm o decide
whe he o use he da a ex ac ed by mos w appe s o egene a e he se o
ex ac ion ules on he fly. This may happen because o an unexpec ed change
o he s uc u e o he web page ha in alida es he ex ac ion ules used so a .
The e is also an agen b oke o in o ma ion ex ac ion ha ac s as a ade
be ween he agen s ha need in o ma ion om he web and he se o a ailable
in o ma ion channels. When an agen needs some in o ma ion, i con ac s he
Agen
socie y
Agen pla o m (FIPA)
B oke
In e ne
In o ma ion
channel 1
In o ma ion
channel 2
In o ma ion
channel n
Fig. 1. P oposed a chi ec u e
b oke , which edi ec s he eques o he app op ia e in o ma ion channel, i
possible. This way, agen s need no be awa e o he exis ence o diffe en ICs,
which can hus be adap ed, c ea ed o emo ed om he sys em anspa en ly
(yellow pages).
2.2 In o ma ion Channels
. An in o ma ion channel is cha ac e ised by he ollowing ea u es: a se o
on ologies, a se o ex ac ion ules, and a na iga ional documen .
–The On ologies. The on ologies [3] associa ed wi h an IC desc ibe he con-
cep s ha define he seman ics associa ed wi h he in o ma ion we a e going
o ex ac om a high-le el, abs ac poin o iew. An on ology allows o
define a common ocabula y by means o which diffe en agen s may in e -
ope a e seman ically.
–The Ex ac ion Rules.The ex ac ion ules allow o define how o ha e access
o he in o ma ion in which we a e in e es ed. To gene a e hem, we need
a se o sample pages con aining es da a on which we use induc i e ech-
niques. To endow he sample pages wi h seman ics, we also need o anno a e
hem wi h DAML [1] ags ha associa e he concep s ha appea in hem
wi h hei on ologies. Once he sample pages ha e been anno a ed, we can
gene a e he ex ac ion ules using a ool called w appe gene a o .
–The Na iga ional Documen . A na iga ional documen defines he pa h o
in e media e pages, such as login o index pages, we need o isi in o de
o ha e access o he page ha con ains he in o ma ion in which we a e
in e es ed. Roughly speaking, a na iga ional documen may be iewed as a
s a e machine in which each s a e is a web page, and he ansi ions be ween
s a es may be iewed as he na iga ion om he page associa ed wi h he
ini ial s a e o he page associa ed wi h he des ina ion s a e.
Once we ha e se up an abs ac channel, we can send i messages o e ie e
in o ma ion abou a gi en book by means o he b oke . The con en o he
messages in DAML is based on an on ology ha defines communica ion [4].
3 Conclusions and Fu u e Wo k
In his a icle, we ha e ske ched some ideas ha show ha he p ocess o ex-
ac ing in o ma ion om he web can be sepa a ed om he business logic o a
web agen by means o abs ac channels. They ely on induc i e w appe s and
can analyse web pages o ge in o ma ion wi h i s associa ed seman ics au o-
ma ically. Thus, we a e con ibu ing o b ing oge he he communi y o agen s
p og amme s and he seman ic web.
In he u u e, we a e going o wo k on an implemen a ion o ou amewo k
in which da a sou ces can be mo e he e ogeneous (da abases, news se e s, mail
se e s, and so on). Ex ac ion o in o ma ion om mul imedia sou ces such as
ideos, images, o sound files will be also paid much a en ion.
Re e ences
1. DARPA (De ense Ad anced Resea ch P ojec s Agency). The da pa agen ma k
up language (daml). h p://www.daml.o g, 2000. 699
2. W. W. Cohen and L. S. Jensen. A s uc u ed w appe induc ion sys em o ex-
ac ing in o ma ion om semi-s uc u ed documen s. In Wo kshop on Adap i e
Tex Ex ac ion and Mining (IJCAI-2001), 2001. 697
3. O. Co cho and A. G´omez-P´e ez. A oad map on on ology specifica ion languages.
In Wo kshop on Applica ions o On ologies and P oblem sol ing me hods. 14 h
Eu opean Con e ence on A ificial In elligence (ECAI’00), 2000. 699
4. S. C anefield and M. Pu is. Gene a ing on ology-specific con en languages. In
P oceedings o On ologies in Agen Sys ems Wo kshop (Agen s 2001), pages 29–35,
2000. 700
5. H. Ga c´ıa-Molina, J. Hamme , K. I eland, Y. Papakons an inou, J. Ullman, and
J. Widom. In eg a ing and accessing he e ogeneous in o ma ion sou ces in TSIM-
MIS. In The AAAI Symposium on In o ma ion Ga he ing, pages 61–64, Ma ch
1995. 698
6. C. A. Knoblock. Accu a ely and eliably ex ac ing da a om he web: A machine
lea ning app oach. Bulle in o he IEEE Compu e Socie y Technical Com-mi ee
on Da a Enginee ing, 2000. 697
7. N. Kushme ick. W appe induc ion: Efficiency and exp essi eness. A ificial In el-
ligence, 118(2000):15–68, 1999. 697
8. G. Mecca, P. Me ialdo, and P. A zeni. ARANEUS in he e a o XML. Da a
Enginee ing Bulle in, Special Issue on XML, Sep embe 1999. 698
9. I. Muslea, S. Min on, and C. Knoblock. W appe induc ion o semis uc u ed,
web-based in o ma ion sou ces. In P oceedings o he Con e ence on Au oma ed
Lea ning and Disco e y (CONALD), 1998. 697
10. S. Sode land. Lea ning in o ma ion ex ac ion ules o semi-s uc u ed and ee
ex . Machine Lea ning, pages 1–44, 1999. 697