Competitive Intelligence based on Social Networks for Decision Making

COMPETITIVE INTELLIGENCE BASED ON SOCIAL NETWORKS FOR DECISION MAKING Fco Fe nando de la Rosa T oyano1, Ma ía Te esa Gómez López1and Ra ael Ma ínez Gasca1 1 Depa amen o de Lenguajes y Sis emas In o má icos. Dp o. Lenguajes y sis emas in o ma icos. Uni e si y o Se ille, Spain { osa , may egomez, gasca}@us.es Abs ac . In p e ious wo ks a amewo k has been p esen ed o ex ac om in e ne he scien i ic communi y in e es ed in a speci ic opic. The p ocess uses sea ch engines que y esul s and e-mails add ess co-occu ences o ob ain he in isible colleges and sub opics o a communi y. This wo k p esen s he use o his echnique o implemen se e al compe i i e in elligence asks o help in decision making in a esea ch a ea. In o de o show an illus a i e pu pose, his echnique is applied o analyze he social ne wo k o pa icipan s in se e al Veille S a égique Scien i ique & Technologique edi ions. Keywo ds: Social ne wo ks ex ac ion, compe i i e in elligence, sea ch engine mining, emails add ess mining. 1 In oduc ion One o he consequences caused by in e ne expansion is he exponen ial g ow h o public in o ma ion. This in o ma ion is s o ed in a se o in e linked he e ogeneous sou ces. Sea ch engines play a key ole in in e ne sea ching in o ma ion, bu he gene al app oaches o analyze he in o ma ion sys ems canno in eg a e di e en sou ces. The analysis o he s o ed in o ma ion in a sys ema ic and au oma ic way can help o decision making in compe i i e in elligence. In his con ex , he co- occu ences analysis becomes impe a i e o implemen in elligence compe i i e asks. In p e ious wo ks a p ocess o ex ac au oma ically he scien i ic communi y in e es ed in a speci ic opic was p esen ed. The p ocess ca ies ou a sys ema ic sea ch engines que ies. These que ies a e specially designed o ex ac he goal ne wo k h ough a pos e io analysis o he sea ch engines eques s. Analysing his ex ac ed in o ma ion is also possible o de e mine he communi y sub opics o in e es . This pape discusses he use o his echnique o implemen se e al compe i i e in elligence asks. Fo example: sea ching o expe s, ga he ing o in o ma ion, o ganiza ions collabo a ions analysis, coun ies collabo a ions analysis o p oduc s impac s analysis. Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 93 The pape is di ided in o he ollowing sec ions. Sec ion 2 p esen s he ela ed wo ks and compa es hem wi h ou p oposal. Sec ion 3 p esen s he Social ne wo ks opic- d i en ex ac ion algo i hm. Sec ion 4 analyzes some Compe i i e In elligence Tasks. In o de o illus a e he pu pose, he de ined p ocess is applied o ex ac he social ne wo k o pa icipan s in he Veille S a égique Scien i ique & Technologique (VSST). Finally, conclusions and u u e wo k a e p esen ed. 2 Rela ed Wo ks P e ious wo ks ha e de ined social ne wo ks ex ac ion p ocesses o di e en in o ma ion web sou ces: sea ch engines [8][10][12], cha s [14], DBLP [5], FOAF a chi es [12][13], Sou ceFo ge [1], mailing lis s [2], e c. The e iew o his pape is es ic ed o app oaches using sea ch engines o ex ac social ne wo ks. One o he sys ems ha uses sea ch engines o ex ac social ne wo ks is REFERRAL WEB [8] which was designed o use Al a is a engine, he ob ained ne wo k is ocused on a speci ic pe son (egocen ic ne wo k). Fo his eason, i only needs o know he name o he ego. By mean o an en i y ecogni ion sys em ex ac s a lis o ela ed people. To measu e he signi icance o he ela ionship be ween X and Y ( he a iable Y con ains any o he pe sons ela ed o X) uses he Jacca d coe icien [7]. The abo e p ocess may be epea ed ecu si ely. Recen ly wo sys ems POLYPHONET [10][11] and FLINK [12][13] we e launched, hese sys ems ob ain he social ne wo k using a lis o he membe s o a de e mined communi y. The basic algo i hm o bo h sys ems mus do a que y o each pai X and Y o c ea e he a ini y ma ix (X and Y a e wo di e en names o he lis ). Bo h sys ems use a h eshold o de e mine when ela ionships a e meaning ul. These sys ems ha e se e al disad an ages, o example he calcula ed a ini ies a e di icul o unde s and. The e ms co-occu ence on he indexed pages may be due o se e al ac o s: co-au ho ship, pa icipa ion in he same e en (i.e., p og am commi ee), e e enced in he same pape , e c. By using pe sonal names in he que y could add ambigui y o he esul s, since he p obabili y ha he name e e s o mo e han one pe son is high. Some imes se e al names e e ence o he same pe son (i.e., ‘Ra ael Ma ínez Gasca’ o ‘R.M. Gasca’). Ano he disad an age is he high cos o ex ac ing he social ne wo k, since in he wo s case o each pai o membe s a que y mus be pe o med. This is a majo p oblem gi en ha licenses o use sea ch engines may ha e limi a ions. Fo example, Google does no allow mo e han 1000 que ies pe day ( o calcula e a ne wo k o 500 ac o s would need 125 days). The scalable algo i hm implemen a ions [11] ha e educed he numbe s o necessa y que ies o comple e a whole ne wo k (acco ding o he au ho , o 503 ac o s 19,852 que ies a e needed, 20 days). The e iewed wo ks use lis s o names o ex ac social ne wo ks. In con as he e is he possibili y o eplacing he lis s o names by e-mails add ess. In [9] is desc ibed Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 94 an algo i hm o ex ac ing social ela ionships o a speci ic in e ne domain using e- mails add ess. This algo i hm is used o social enginee ing a acks in secu i y o compu e sys ems a ea and has a high cos o ex ac ing he social ne wo k, o each pai o e-mail add ess a que y mus be pe o med. This pape p oposes a social ne wo k ex ac ion algo i hm based on e-mails o imp o e he complexi y o e iewed algo i hms. The que y model ha is used makes i mo e obus o ambigui y, and allows a clea in e p e a ion o he ela ionships ex ac ed. I does no use lea ning p ocess. Unlike p e ious wo k [5], he p oposed algo i hm does no need an ini ial lis o e-mails and i uses heu is ics o d i ing he cons uc ion o a social ne wo k by opic and by impo ance o he ne wo k membe s. The p ac ice cases p esen ed in his pape ha e been de eloped wi h TREDAR ool. This ool pe mi s o de ine business p ocesses in an in e ac i e way. The speci ic analyzed case is a compe i i e in elligence p ocess [19] ha pe mi s ecollec and ex ac da a om in e ne abou VSST communi y and pe o m di e en ypes o analyses o he decision making o sol e di e en que ies. This ool also allows us o sp ead and isualize he ob ained da a o decision-making suppo . In his way, all he necessa y s ages o p o ide a igilance se ice and compe i i e in elligence a e in eg a ed h ough web echnology. 3 Social ne wo ks opic-d i en ex ac ion algo i hm In his sec ion he algo i hm o social ne wo ks opic-d i en ex ac ions o malize in [6] is desc ibed. The algo i hm is di ided in o h ee s eps: • To seed e-mails add ess ex ac ion: To s a he ex ac ion p ocess is necessa y o ha e a se o e-mails. Di e en p ocesses o ex ac ing e-mail add esses om web can be de ined. The e-mails add ess selec ion p ocess used by de algo i hm is desc ibed below: o To pe o m que ies wi h he ‘< opic>‘ and isi he web documen s e u ned by he sea ch engine o To ex ac he e-mails om he <a> ags o he h ml pages and o Fo each e-mail add ess, o check h ough he que y ‘<email>’ ‘< opic>’ he associa ion deg ee o e-mail o he opic. The p ocess conside s a high associa ion deg ee i he numbe o documen s e u ned by he p e ious que y exceeds a h eshold. • To expand he e-mail add ess: The algo i hm expands he social ne wo k wi h new ela ionships ex ac ed om sea ch engine que ies esul . Fo each mail add ess he algo i hm do: o To c ea e a que y using he schema: ‘<use name>’ ‘<domain>’ ile ype:pd . Fo example, he use name o he e-mail add ess osa @ us.es is osa and he domain is us.es, hence he que y is: ‘ osa ’ ‘us.es’ ile ype:pd and ex ac s he esul con ex s. Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 95 o The con ex s we e analyzed using egula exp essions, o analyze only he con ex s in which e i y he e-mail appea ance. Each new e-mail ha appea s in he con ex is added o he social ne wo k as a new node and he ela ionships a e added. The nodes and he ela ionships a e associa ed wi h a coun e which will ep esen hei impo ance in he ne wo k. I any e-mail add ess appea s again a some con ex , he coun e will inc ease by one uni , also he ela ionships. • Social ne wo ks opic-d i en: This p incipal p ocess is i e a i e and s o es he e-mails add ess in a p io i y queue. o Ini ially, he queue is ini ialized wi h he e-mails add ess seeds. o In each i e a ion an e-mails add ess is ex ac ed om he queue op, and i is checked he associa ion deg ee o he e-mail add ess o he opic. I he associa ion deg ee exceeds a h eshold, hen i is expanded wi h he neighbo s o he e-mails add ess. o The i e a ion is epea ed un il he queue is emp y and o he s a egies can be implemen ed, o example limi ing he numbe o web pages isi ed o he numbe o emails-add ess o he social ne wo k. o A e comple ing he expansion, he new nodes a e added o he queue. Al hough i is possible ha a pe son can ha e mo e han one e-mail add ess ( o example a pe sonal and a ins i u ional e-mail add ess), i is also ue ha an e-mail add ess ypically iden i y a pe son. I could a oid he ambigui y p oblems o o he me hodologies, bu no he a ie y p oblems. Rea anging he queue se e al d i en s a egies can be implemen ed: maximizing he associa ion deg ee o e-mail o he opic o o he social ne wo ks analysis measu es such as he deg ee, page ank o he be weenness. 4 Compe i i e In elligence Tasks In o de o illus a e he social ne wo ks opic-d i en ex ac ion echnique, he ‘VSST’ and ‘In e lligence compe i i e’ opic has been used. The goal is he ex ac ion o he social ne wo k o he communi y o he se e al Veille S a égique Scien i ique & Technologique (VSST) edi ions. The 327 seeds we e ex ac ed om di e en VSST web pages edi ions. A e unning he ex ac ion algo i hm, he social ne wo k had 2.107 nodes and 7.102 edges. O hese nodes, 432 nodes exceeded he minimum h eshold, only hese nodes we e expanded. Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 96 Table 1. Numbe s o documen s and que ies associa ed o each opic The social ne wo k ex ac ed can be used o ind expe s in ce ain opics, [15] uses he numbe o pages o de elop expe s anking. In ou case we can use he que y <email> "< opic>" o es ima e de numbe o documen s associa ed and de elop a opic anking o expe s, as i is shown in Table 1 and Table 2. The e a e al e na i es o he impac measu e such as Mindsha e. The ad an age ha we ha e p oposed in he expe s inding ask is he own o he social ne wo k, his allows using he ARS measu es o classi y he expe s, o example: deg ee, au ho i y, cen ali y, be weens, e c [16]. Table 2. Numbe s o documen s associa ed o each opic An app op ia e selec ion o email add esses can be used o model an a ea o in e es and op imize he ga he e o in o ma ion. Fo example, using he que y <email> ile ype:pd o download pd documen s o expe s in he a ea. This ga he ing o in o ma ion can be used o e ine he expe anking. In o de o analyze he global opics impac o he social ne wo k, i is possible analyzing he impac o p oduc s. In [5] his app oach is used o analyze wo ARS ools, Pajek and Ucine . I concluded ha he global impac o Ucine on he social ne wo k was a 20% la ge han Pajek. P obably he di e ence o impac is due o i s usabili y. The mayo p oblem o his app oach is he wo ds ambigui y. By using Figu e 1 i is possible o analyze he coope a ion ela ions be ween he coun ies o he VSST pa icipan s. In his case unde lines he s a egic posi ion o F ance and his his o ical connec ing wi h he Magh eb coun ies. And using Figu e 2 is possible o analyze he coope a ion be ween di e en o ganiza ions h ough hei Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 97 Fig. 1. VSST coun ies collabo a ions domains. Fo example, analyze he domains we can disco e new compe i o s (IALE, CDE, ISCOPE, IMCSLINE, e c), new clien s (LAPOSTE, CEA, EADS, e c) o new in es iga ion cen e s (IRIT, INIST, e c). And analyzing he ela ionships we can app ecia e he cen al posi ion o di e se uni e si ies in he ne wo k. Also using he IP domains is also possible o alloca e he o ganiza ions [5]. Figu e 3 ep esen s in a lexical ne wo k he VSST communi y opics o in e es . In o de o build his map, wo asks ha e been execu ed: (1) he key wo ds o he pape published in VSST e en s in 2007 ha e been ex ac ed (2) he concu ency ne wo k has been calcula ed [18]. Using his map, i is possible o de e mine he in e es ing cen e s o he esea che ha pa icipa e in he e en s. The ollow opics o in e es ha e been ex ac ed in his example: ex -mining, na u al language p ocessing, c ea i i y and inno a ion, in o ma ion e ie al and il e ing, co-wo d analysis, business-in elligence. The esul s o he VSST case s udy a e online a ailable in h p://www.lsi.us.es/~ osa /index.php/F osa /MapaLis aVSST2007Es. 5 Conclusions and u u e wo k In his pape some echniques ha e been combined helping in he making-decision p ocess o an esea ch o ganiza ion. The opics ha ha e been used a e : ex ac ion o in o ma ion om he web, analysis o social ne wo ks and co-occu ences. The que ies ha can be sol ed wi h hese echniques a e : • Which a e he expe s in an speci ic a ea? • Which a e he mos in luence g oups ? • How he o ganiza ions a e s uc u ed? • Which a e he mos in luen ial coun ies ? • Which a e he goal ne wo k? Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 98 This wo k p o ides a amewo k o making-decision p ocesses o analyse he dis ibu ed da a in di e en and he e ogeneous sou ces, non cen e ed in any da abase. Figu e 4 shows in a isual way he main ideas de eloped in his wo k. As u u e wo k we p opose an au oma ic classi ica ion o he domains using he a ailable in o ma ion in he esea h cen e webs. Also i will be in e es ing o de elop echniques o ex ac in o ma ion abou he en i ies o be ep esen ed in he maps (names, depa men s, elephone numbe s…) Acknowledgmen s. This wo k has been pa ially unded by Jun a de Andalucía (P08-TIC-04095). Re e ences 1. K. C ows on and J. Howison. (2005) The Social S uc u e o F ee and Open Sou ce So wa e De elopmen , Vol. 10, No. 2, Fi s Monday. 2. A. Culo a, R. Bekke man, and A. McCallum. (2004) ‘Ex ac ing social ne wo ks and con ac in o ma ion om e-mail and he web’, P oceedings o he Con e ence on Email and Spam. 3. F. de la Rosa T., S. Pozo and R. M. Gasca. (2005) ‘Análisis y isualización de comunidades cien í icas con in o mación Ex aída de la web’, IEEE La in Ame ica T ansac ions, Vol. 3, No. 1, ISSN: 1548-0992. 4. F. de la Rosa T, R. M. Gasca, L. González y F. Velasco (2005). Análisis de Redes Sociales median e Diag amas Es a égicos y Diag amas Es uc u ales. Redes: Re is a Hispana pa a el Análisis de Redes Sociales. ISNN: 1579-0185. Vol. 8. 5. F. de la Rosa T., F. and Gasca, R.M. (2007) ‘Sis emas de in eligencia web basados en edes sociales’, Re is a Hispana pa a el Análisis de Redes Sociales, Vol. 12, ISSN: 1579- 0185. 6. F. de la Rosa T. and R. M. Gasca (2008) Au oma ic ex ac ion o social ne wo ks by opics o in e es . IJCAT , Vol. 33 , N . 4, p. 292-299. 7. P. Jacca d (1901) ‘É ude compa a i e de la dis ibu ion lo ale dans une po ion des Alpes e des Ju a’, Bull Soc Vaudoise Sci Na , Vol. 37, pp.547–579. 8. H. Kau z, B. Selman and M. Shah (1997) ‘The hidden web’, AI Magazine, Vol. 18, No. 2, pp.27–35. 9. J. Long (2005) The Google Hacke ’s Guide, Re ie ed, ISBN 1-59749-176-4. 10. Y. Ma suo, H. Tomobe, K. Hasida, and M. Ishizuka. (2003) ‘Mining social ne wo k o con e ence pa icipan s om he web’, P oceedings o he In e na ional Con e ence on Web In elligence, pp.190–194. 11. Y. Ma suo, J. Mo i, M. Hamasaki, H. Takeda, T. Nishimu a, K. Hashida and M. Ishizuka. (2006) ‘Polyphone : an ad anced social ne wo k ex ac ion sys em, P oceedings o WWW2006. Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 99 12. P. Mika. (2004) ‘Boo s apping he FOAF-Web: an expe imen in social ne wo k mining’, P oc. 1s Wo kshop F iend o a F iend, Social Ne wo king and he Seman ic Web. 13. P. Mika. (2005) ‘Flink: Seman ic web echnology o he ex ac ion and analysis o social ne wo ks’, Jou nal o Web Seman ics, Vol. 3, No. 2. 14. P. Mu on. (2004) In e ing and Visualising Social Ne wo ks on In e ne Relay Cha , In oVis, Aus in, TX, pp.35–43. 15. Da Sil a, A, Manniana, B., Quoniam, L. and Ros aing, H. ‘Sea ching o Expe s on he In e ne ’ Compe i i e In elligence Rewiew, USA, . 11, n. 4, 2000. 16. Sco , J. P. (2000) ‘Social Ne wo k Analysis: A Handbook. Second edi ion’ Sage Publica ions. 17. X. Canale a, P. Ros, A. Vallejo, D. Ve ne and A. Zaballos. "A sys em o ex ac social ne wo ks based on he p ocessing o in o ma ion ob ained om In e ne ", P oceedings o he 11 h In e na ional Con e ence o he Ca alan Associa ion o A i icial In elligence(CCIA 2008), IOS P ess, ISBN: 978-1-58603-925-7, 2008 18. N. Coul e , I. Mona ch and S. Konda. (1998). “So wa e enginee ing as seen h ough i s esea ch li e a u e: A s udy in co-wo d analysis”. Jou nal o he Ame ican Socie y o In o ma ion Science, 49(13), pp. 1206-1223. 19. P. Esco sa P.y R. Maspons (2001), De la Vigilancia Tecnológica a la In eligencia Compe i i a. Mad id. P en ice Hall. 20. V.A. Bucheli y F. González (2007) He amien a in o má ica pa a igilancia VIGTECH-. A ances en Sis emas e In o má ica, . 4 n.1, pp 117-126, Junio 2007, Facul ad de Minas, Uni e sidad Nacional de Colombia, Medellín Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 100 Fig. 2. VSST o ganiza ions collabo a ions Ac as de los Talle es de las Jo nadas de Ingenie ía del So wa e y Bases de Da os, Vol. 3, No. 1, 2009 ISSN 1988–3455 SISTEDES, 2009 101