Intelligent Agents for Information Presentation: Dynamic Description - PDF document

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/2781252 Intelligent Agents for Information Presentation: Dynamic Description of Knowledge Base Objects Article · September 1998 Source: CiteSeer CITATIONS READS 0 23 6 authors , including: Karin Maria Verspoor Robert Dale University of Melbourne Language Technology Group 283 PUBLICATIONS 4,225 CITATIONS 240 PUBLICATIONS 5,836 CITATIONS SEE PROFILE SEE PROFILE Maria Milosavljevic Cécile Paris The University of Edinburgh The Commonwealth Scientific and Industrial Research Organisation 40 PUBLICATIONS 388 CITATIONS 340 PUBLICATIONS 4,751 CITATIONS SEE PROFILE SEE PROFILE Some of the authors of this publication are also working on these related projects: We Feel View project EBMSummariser View project All content following this page was uploaded by Karin Maria Verspoor on 09 November 2014. The user has requested enhancement of the downloaded file.

In telligen t Agen ts for Information Presen tation� Dynamic Description of Kno wledge Base Ob jects Cornelia M� V ersp o or � Rob ert Dale � Stephen J� Green � � � � Maria Milosa vljevic � Cecile P aris y � and Sandra Williams � � Microsoft Researc h Institute � Macquarie Univ ersit y Sydney NSW ���� Australia f kversp�rdale�sj m g gre en �ma ri am �sw il lia �m ri� mq �ed u� au T el� ��� ��� ���� ����� F ax� ��� ��� ���� ���� y CSIR O Mathematical and Information Sciences Lo c k ed Bag �� North Ryde NSW ���� Australia Cecile�Paris�cm is� cs iro �a u T el� ��� ��� ���� ����� F ax� ��� ��� ���� ���� Keyw ords� natural language generation� user tailoring� m ultilingual information presen tation Abstract Users of the W orld Wide W eb ha v e needs and in terests whic h can help to determine what of the v ast quan tities of information a v ailable migh t b e relev an t to them� In telligen t agen ts migh t b e used to select con ten t for a particular user� Ho w ev er� it is also imp ortan t to consider how that con ten t is pro vided to a user� W e suggest that this information pr esentation m ust also tak e in to consideration the needs of a user� and discuss a set of agen ts whic h utilizes natural language generation tec hniques to presen t information in an appropriate w a y � In this pap er w e describ e t w o systems w e ha v e built whic h dynamically generate descrip� tions of kno wledge base en tities� and consider the extension of the tec hniques used there for m ultilingual information presen tation� W e describ e the notion of a phr asal lexic on as a basis for dynamic ob ject description� and prop ose a mo del for dynamic m ultilingual description whic h builds on that notion�

� In tro duction In the con text of the W orld Wide W eb� in telligen t agen ts are often discussed as a means of �nding information that is relev an t to a particular user�s needs at a particular p oin t in time� Those needs� ho w ev er� also in�uence the most e�ectiv e presen tation of the relev an t information� as migh t the needs of the information pro vider� It is therefore imp ortan t to consider w a ys of in tro ducing user�tailoring in to information presen tation� In this pap er w e explore the use of a set of in telligen t agen ts for information presen tation on the W orld Wide W eb� These agen ts use natural language pro cessing �NLP� tec hniques to generate appropriate descriptions� W e argue that a k ey b ene�t of this tec hnology is the p oten tial for data reuse under v arying presen tation constrain ts� and examine the particular case of m ultilingual presen tation to suggest one w a y forw ard for this tec hnology � This w ork builds on existing w ork in tailored kno wledge base description � the PEBA�I I system �Milosa vljevic� T ullo c h and Dale ����� � the ILEX system �Knott et al ������ and join t w ork �Dale in press� � and describ es ho w kno wledge base en tities can b e dynamically et al describ ed� W e also discuss the ease with whic h w e w ere able to p ort the PEBA�I I system to a new domain in the PO WER system� due to the use of a and a re�usable agen t� phr asal lexic on based system arc hitecture� and explore ho w these features of our approac h migh t also b e applied in a m ultilingual generation system� � T ailoring metho dologies Information presen tation should b e tailored to the needs of a user� and the needs of the informa� tion pro vider� F or instance� new comers to a domain will w an t to see information at a di�eren t lev el of detail than exp erts� In the con text of a virtual m useum� a m useum curator ma y ha v e a particular message he hop es to con v ey to visitors� and the visitors ma y ha v e an agenda in visiting a particular exhibition� Di�eren t p eople ma y ha v e di�eren t p ersp ectiv es on individual ob jects� John migh t w an t to kno w the history of an ob ject while Jane is more in terested in its comp osition� Suc h tailoring can b e ac hiev ed in v arious di�eren t w a ys� W e consider a few curren t metho d� ologies b elo w� ��� Adaptiv e Hyp ertext Brusilo vsky ������ reviews adaptiv e h yp ermedia systems� These systems construct user mo dels whic h driv e the adaptation of a h yp ermedia page� This adaptation migh t in v olv e� for example� limiting the bro wsing space a user has access to or prioritizing and�or annotating h yp erlinks to guide a user through the a v ailable information� These systems w ork with static textual do cumen ts� but v ary the access whic h a user has to those do cumen ts� supp ort v ariations in whic h links are A daptive navigation supp ort systems a v ailable to a user at a giv en p oin t� or in the order in whic h the links are presen ted� and adaptive include or suppress certain bits of text dep ending on the needs of a user� pr esentation systems Although the use of suc h approac hes is an imp ortan t step to w ards the tailoring of information to a user� they do not go as far as they migh t� F or example� they do not consider issues of textual coherence � what is the e�ect of limiting access to texts or c hanging the order or their presen tation� F urthermore� they require all the texts p oten tially a v ailable to a user to b e pre� written� whic h demands a large amoun t of w ork when man y lev els of v ariabilit y are in tro duced� and complete rewriting of texts when c hanges or additions are made in the information to b e con v ey ed� This is an issue for m ultilingual systems in particular� since the same information m ust b e expressed in sev eral di�eren t w a ys �languages��