Cherubini PhDAnnualReport06

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Doctoral Program in Computer, Communication and Information Sciences

PhD Annual Report


Candidate: Mauro Cherubini Thesis Director: Prof. Pierre Dillenbourg
Lausanne, March 28, 2006

Name of the candidate

Mauro Cherubini

Probable title of the thesis

Collaborative Annotations of Maps: a Computational Model that Integrates Geometrical and Semantical Dimensions of Communication.

Keywords

i. Computer Supported Collaborative Learning / Work (CSCL / W); ii. Collaborative Annotations of Maps; iii. Locative Media; iv. Location Based Services; v. Informal Learning; vi. Spatial Cognition; vii. Visual Information Retrieval; viii. Spreading Activation; ix. Spatial Information Retrieval; x. Mobile Information Retrieval.

A precise abstract of the subject being studied

This work targets Collaborative Annotations of a Map in a mobile setting. This is a form of communication that makes explicit usage of the geographical/physical context as referent to the message content. The goal of this study is to develop a computational support for such communication, dening a model that enable to integrate spatial information with the textual information produced through computer-mediated communication. As rst step, I will analyse datasets of these particular messages with the aim of understanding their peculiarities in comparison with comparable canonical forms. After, I will use this

information to build a specic information retrieval engine that will support the users exploration of the information space.

Update of the general plan of the thesis

The research plan presented in the thesis proposal maintained the stated research objectives. However, some technical constraints and advancements, which happened during the last year, moved the focus of research to the direction detailed afterward. The advancement of the second tool presented on the research plan, the mobile application for annotation named STAMPS1 , brought me to concentrate on the city scenario sketched on the thesis project. Secondarily, I operated some choices on the algorithm to be used on the content-side of the search engine. Some pre-experiments brought me to explore the possibility to incorporate and extend a Spreading Activation model for information retrieval [Crestani, 1997]. I plan to run specic experiments to conrm our initial ndings (Experiment 2 on gure 4). I expect the rst experiments to take place in this semester. The current plan is to recruit 40 subjects and ask them to use the application during a 6 months period. During this timeframe dierent information retrieval technologies will be activated allowing us to register any emerging dierence of usage of the tool (Experiment 1 on gure 4). In parallel, I signed a collaboration with an English rm, named Proboscis, that developed and tested a similar application called Urban Tapestries2 . I obtained access to the datased collected by this company during the eld trial which took place in London, last year. This will allow me to compare our results with similar messages taken in a dierent context, increasing the scientic value of any produced result.

Abstract of the results obtained during the past year with a list of publications and technical reports

The thesis project was presented in several meetings: the CAIF workshop (Collaborative Artefact Supporting Collaboration) a three-days workshop that was organized in ChteaudOex in Switzerland [Cherubini, 2005b]. More recently a the project was presented at the Kailedoscope workshop organised in Oulu, Finland [Cherubini, 2005a]. From the submission of the research plan (the 13th of June 2005) dierent deliverables have been completed. Initially the ShoutSpace client was tested internally (see gure 1). During the fall 2005 I compiled a literature review on the subjects of visual information retrieval [Au et al., 2000] and information retrieval in a mobile context, which was presented internally on the 10th of November 2005. During the discussion, I presented a possible contribution for the thesis, namely the incorporation of a geometrical routing of energy in a spreading activation model for
1 See 2 See

the website: craftsrv1.ep.ch/research/stamps/ the website: urbantapestries.net/

Figure 1: ShoutSpace interface information retrieval [Crestani, 1997]. This led me to promote and develop an experimental framework that could be used to compare semantic retrieval engines. Figure 2 represents the experimental tool used in the tests. A publication on this theme is in preparation. During the last semester I completed the development of the mobile version of the application, STAMPS. On gure 3 is possible to see the interaction ow of an user that is posting a message at a certain place.

Abstract of the work done from the beginning of the project

I have been testing dierent approaches to investigate the thesis idea: how space can inuence communication. I started this project trying to nd a way for architecture students to change their approach of study towards urban planning. In this context, we developed a proposal aiming to provide a computational device to collaboratively build their notes on the city [Cherubini, 2003]. At the end of this period I elaborated a vision on how this technology could move to a broader audience to support citizenry while exploring their city space [Cherubini and Nova, 2004]. After, I decided to slightly change the focus of my work, concentrating on questions which are more of interests for the CSCL community: the inference and grounding processes. The approach chosen, focused on the small group and on the cognitive processes happening between the members while jointly solving or elaborating tasks [Cherubini, 2004]. Furthermore, my last work focused on the grounding process as a key element to understand collaborative knowledge

Figure 2: Visual Information Retrieval Experiment interface construction [Cherubini and van der Pol, 2005]. All the dierent elements described in this section and in the precedent are schematised in gure 4, which shows the thesis plan with an updated timeline of development.

Plan of the work to be achieved during the year ahead

Three phases will be required to achieve the thesis goals: (a) the observation of people communicating using the designed application; (b) the renement of the retrieval engine that could sustain this kind of communication; (c) the validation of this model through a second experimental phase. (a) Phase I: The observation of communication. This part of research should last 6 months: an experimental group of users will be gathered for providing experimental data to the framework. The results of this enquiry will be analysed to serve the development of the second phase. The data collected during this phase will be compared with those collected with alternative systems like Urban Tapestries (Q1 on gure 4). (c) Phase II: The computational algorithm. The observation of phaseI will provide some clues on the nature of this kind of spatial communication like the connection of semantic and geometry of the messages. Together with side experiments on the semantic retrieval engines (Experiment 2 on gure 4), this will suggest a possible mechanism to sustain this kind of communication 4

Figure 3: Interaction ow of the STAMPS interface that will be implemented in this phase. (c) Phase III: The verication. Finally, during the last phase the retrieval system implemented during the second phase will be evaluated with two experimental group of users. The phase three might interleave during the 6 months of phase I (Experiment 1 continued on gure 4).

8.1

How I plan to answer the research question

In this research, I will develop modelling schemas that enable to integrate spatial information, as embedded in maps, with the textual information produced through computer-mediated communication. This integration will be investigated through the development of a specic search engine that combines these two sources of information. Q1. What is the mapping between the geographical, the semantic and the social structure of a body of localised text-based asynchronous messages? To answer this question I propose to cluster the messages treating them as geometrical entities at rst and then as semantic entities. The mapping between 5

the two dierent grouping will dene a distortion between the two dierent aspect of the spatial communication. Additionally several other markers can be used to identify dierences of this kind of communication from a comparable form as newsgroups messaging (i.e., dispersion of the messages, structure form, number of links per node, specic pointer words frequency, etc.). Q2. Does the geographical structure of the message set facilitate the clustering of messages into meaningful patterns? The messages datasets contain information on the messages author, time, and threading. Using this information is possible to construct dierent maps using primarily one of these components at a time. Then dierences between the obtained maps can be measured with the specic aim of nding map invariants. Q3. Will a search engine based on space-interaction patterns support search processes that are less formal than keyword-based search process? To answer this question we will use the experimental setup described in the thesis proposal. Two group of users of the annotation tool will be exposed to dierent search engines: a control group using a plain keyword-matching engine; an experimental group using an advanced search engine which will relate the messages geometrical dimension with the semantical dimension. The observed dierences of usage of the tool by the two dierent groups will provide clues on the question.

Estimated achievement date

September / October 2007

10

Signatures
Prof. Pierre Dillenbourg Thesis director

Mauro Cherubini Candidate 20th of March 2006

References
[Au et al., 2000] Au, P., Carey, M., Sewraz, S., Guo, Y., and Rger, S. M. u (2000). New paradigms in information visualization. In Proceeding of SIGIR2000, pages 307309, Athens, Greece. ACM Press. Available from: http://km.doc.ic.ac.uk/www-pub/npiv-sigir2000.pdf. [Cherubini, 2003] Cherubini, M. (2003). Maptribe proposal. Tech Report 3, CRAFT, Ecole Polytechnique Fdrale de Lausanne, Lausanne, Switzere e land. Available from: http://www.i-cherubini.it/mauro/projects/ MapTribe/. 6

[Cherubini, 2004] Cherubini, M. (2004). A collaborative ontology for spatialised communication. In Position paper for the workshop Potential of Cognitive Semantics for Ontologies, part of FOIS2004, Torino, Italy. Available from: http://www.i-cherubini.it/mauro/publications/FOIS2004 WorkshopPositionPaper CherubiniExtAbs.pdf. [Cherubini, 2005a] Cherubini, M. (2005a). Stamps, a system for georeferrenced messaging. In Presentation at the workshop Multiple Technologies and Tools for Supporting CSCL: A step further. 21-23 November, Oulu, Finland. Kaleidoscope Network of Excellence. [Cherubini, 2005b] Cherubini, M. (2005b). A system for tagging messages, post-inferential semantics. In Presentation at the workshop CAIF (Collaborative Artefacts Supporting Collaboration). June, Chateau dOex, Switzerland. Ecole Polytechnique Fdrale de Lausanne. e e [Cherubini and Nova, 2004] Cherubini, M. and Nova, N. (2004). To live or to master the city: the citizen dilemma: Some reections on urban spaces fruition and on the possibility of change ones attitude. Imago Urbis, Universitas de Quilmes, Buenos Aires, Argentina, (2). Available from: http://www.i-cherubini.it/mauro/publications/ Cherubini Live or Master 21apr04.pdf. [Cherubini and van der Pol, 2005] Cherubini, M. and van der Pol, J. (2005). Grounding is not shared understanding: Distinguishing grounding at an utterance and knowledge level. In CONTEXT05, the Fifth International and Interdisciplinary Conference on Modeling and Using Context, Paris, France. Available from: http://www.i-cherubini.it/mauro/ publications/Cherubini vanderPol CONTEXT05 dc.pdf. [Crestani, 1997] Crestani, F. (1997). Application of spreading activation techniques in information retrieval. Articial Intelligence Review, 11(6):453482. Available from: http://www. springerlink.com/(o0ccgl550sriscezu3ugap45)/app/home/ contribution.asp?referrer=parent&backto=issue,2,2;journal, 49,50;linkingpublicationresults,1:100240,1.

Figure 4: Thesis plan with updated timeline

You might also like