Finally I can write a RIP (Repose In Peace) on the Virtual Information Retrieval experiment. In the last strategy I was trying to detect possible differences between the two algorithms (LSI and CNG) to support the user in exploring the results of the query. Finally I found an effect when using the relative distance index between jumps.

Basically, when the user explores the results in the map, s/he select items one at a time. Our initial hypothesis was that the distance between each ‘jump’ was a function of the pertinence of the document (either detected by the user or as an absolute value).

We found that this hypothesis was not verified when looking at the distances from an absolute point of view, as the documents are not distributed equally in the map. On the contrary, if we look at these distances from a relative point of view (number of closer documents compared to the next document chosen), then we have a different overview of what happened.

It happens in fact that CNG is more efficient in splitting good results from the rest of the documents returned from the query. We have seen this through the heatmap, as LSI spreads the results all around. Subsequently, the user is more able to identify the good results as they are well isolated in the bottom part of the map. As the user moves from that selection the absolute distance increases a lot but the relative distance stays low. And certainly lower that what the user does using LSI (see test below).

> wilcox.test(distance_rel ~ as.factor(method))

Wilcoxon rank sum test with continuity correction

data:  distance_rel by as.factor(method)

W = 423460.5, p-value = 2.271e-12

alternative hypothesis: true mu is not equal to 0

In sum, we could not see an improvement of the performances as the task was maybe too difficult but we could register an interesting effect in the user strategy that reflects on the ability of each method to cluster the results in a way that facilitated the user.

Tags: clustering, information retrieval, Latent Semantic Analysis, map algorithms, spreading activation, user experience

Julian Beever is an English artist who is famous for his art on the pavements of England, France, Germany, USA, Australia and Belgium. Its peculiarity?  Beever gives his drawings an anamorphosis view, his images are drawn in such a way which gives them three dimensionality when viewing from the correct angle.

Copyright notice: the present content was taken from the following URL, the copyrights are reserved by the respective author/s.

Tags: anamorphosis maps, art, urban, urban exploration

As a final attempt to understand the exploration strategies used by the participants of the Visual Information Retrieval experiment we decided to try to map their jumps on the map in relation to the available possibilities calling this a “relative distance”. For instance, if two consecutive documents in the jump sequence were the absolute closest this would have produced a relative distance of 1.

On the contrary all the other documents which would have been at a shorter distance to the starting one, would count as more units on this relative distance.

This is a trick to understand to what extent the user tried to explore the nodes which were closer to the selected one, or, in other words, we wanted to understand whether the jumps distance was a function of the document pertinence.

The picture below represent a small excerpt of the results of this analysis. The last column tell use whether the user was reading a correct answer to the task.

Swisscom Auto-id is a spin-off of Swisscom telecommunication that aims at hitting the radio frequency tags business market. The main applications developed will target companies logistics:

Swisscom Auto-ID Services can help speed up your company’s supply management thanks to Radio Frequency Identification (RFID) technology. RFID, which allows high-speed automatic data recording using radio waves, is able to identify all manner of objects without direct contact, and can be used in applications such as logistics, inventory management and industrial automation.

One of the crucial point I am facing is that of being able to monitor the SMS activity of the participant of the field trial of STAMPS. This will provide useful information to compare the spatialized messaging with canonical strategies of communication. There are a couple of technical limitations which prevent a direct monitoring on the outgoing box so I have been thinking on possible workarounds.

1st solution: monitor the inbox for incoming messages from one of the group participant. This will require to know the numbers of all the participants which is possible. One of the limitation is that it will also parse other messages with possible infringement of the privacy.

2nd solution: ask the user to use a second phone knowing that this second device, the SMS exchanges will be monitored. A bit more fair but not practical to hang around with two phones in the pocket.

3rd solution: implement an internal private and direct one-to-one communication with other participants. This will be mostly transparent to the user but it will face the risk of not being used because people are accustomed to current SMS mechanisms.

Interesting meeting this afternoon with Martin Rajman on information retrieval techniques in relation to the STAMPS project. One of the first reference I picked up was that of the link mining, which is the data mining activity of constructing links between documents.

Then Dr. Rajman, gave me two references to the weighting scheme which are currently used to make rankings on a certain data structures: Prosit (also known as DFR) and Okapi (a variant of TF*IDF).

He mentioned briefly a technique they are working on at LIA to combine a selection technique with a Random Walk data structure which should be based on probabilistic formulas.

Basically, the building of the structural part on the graph is the result of how people would have walked our graph through many interactions.

An interesting comment Dr. Rajman made is that the different layers / properties we are trying to combine (social, semantic, geometric) are mostly precious if kept separated. These are maybe interesting to use at different time during the user interaction: for instance the user may start looking for a specific area on the map and then s/he might want to zoom on a more specific subset of the result, maybe this time focussing more on the social layer and using this to trim the results. We should enquiry how to use pertinence propagation across the different layers.

Another interesting thought from Dr. Rajman is that words are interesting to look at because they do not follow gaussian distributions in phenomena.

E. F. Churchill, J. Trevor, S. Bly, L. Nelson, and D. Cubranic. Anchored conversations: chatting in the context of a document. In Proceedings of the SIGCHI conference on Human factors in computing systems, pages 454 – 461, The Hague, The Netherlands, 2000. Association for Computing Machinery. [pdf]

———————-

This paper describes an application-independent tool called Anchored Conversations that brings together text-based conversations and documents. The design of Anchored Conversations is based on our observations of the use of documents and text chats in collaborative settings. We observed that chat spaces support work conversations, but they do not allow the close integration of conversations with work documents that can be seen when people are working together face-to-face. Anchored Conversations directly addresses this problem by allowing text chats to be anchored into documents. Anchored Conversations also facilitates document sharing; accepting an invitation to an anchored conversation results in the document being automatically uploaded. In addition, Anchored Conversations provides support for review, catch-up and asynchronous communications through a database. In this paper we describe motivating fieldwork, the design of Anchored Conversations, a scenario of use, and some preliminary results from a user study.

Pierre told me about this work today. I was a bit surprised that he slipped over this important reference till this moment. I think this paper is extremely important for my research as focusing on making links to between specific points of the document where the discussion concentrates. In this sense the document can be seen as a map on top of which the chat has been anchored. Can we find an asynchronous alternative to this?

Tags: collaboration tools, context, graphical user interface, interaction design, map algorithms

Fortunately there are Italians like Antonio Scarponi around. They represent the genius and the creativity of some of us against the barbarian disgrace that Italy is living in this period.

His latest project is called ‘Italia in fumo’ (Italy burned and transformed in smoke). The goal of the project is to raise the awareness around the italian constitution, a document that is largely unknown by most of the italians 🙁 .

Antonio proposes to use some common form of public display to spread the knowledge of the constitution articles. With the same graphic language of the european directive concerning the cigaret consumption, the project web site proposes a series of widgets that is possible to download and use to spread the constitution around. On the picture below, for instance, there is the model for building a cigaret box cover. I find it really neat and smart! Thanks Antonio!

Tags: education, information visualization, Italy, learning technology, pedagogy, politics

I finally join the wave started on the smart mob world and catch promptly by Nicolas and Fabien on the great advantages of having a web server running on a mobile phone. Why is this step so important for an ecology of the internet of things:

Providing access to a mobile phone from the Internet is not straightforward, as operators typically employ firewalls that prevent access from the Internet to phones inside that firewall. By implementing a custom gateway we could circumvent that limitation and we are now able to provide a webserver on a mobile phone with a global URL than can be accessed from any browser. In a sense, the mobile phone has now finally become a full member of the Internet.

It is clear that with this ‘little’ step a mobile phone becomes an ubiquitous ‘transductor’ from the bits world  to the atom world, a connection that was missing so far. We can already imagine lots of different geek application (aka ‘find my shoes’). Now, let’s start thinking for real on what to do with this technology which will not make my grandmother say “and then what?”.

Besides, any example out there yet on how to make an ‘hello world’?

Tags: new technology, internet of things, reverse engineering, smart mobs

Trying to reshape the experimental setup of my thesis, I brainstormed this afternoon with my respected colleagues Mirweis (aka ‘Sinalco’) and Gaelle (aka ‘Molinette’). The aim was to find an experimental setting that could support the hypotheses of the thesis in a closer look that the current experiment I am doing with Lorenzo.

So the starting hypotheses are that space provide a context for the receiver and a new structure to communication. Therefore we might expect shorter messages and a smaller number of iterations when dealing with a task. Other specific linguistic indicators might include: it might be easier to reach grounding, or there might be fewer repair acts.

The initial idea I had was to implement two different interfaces of the kind developed for the VIR experiment. A map and a search box. Using the search box, the users can browse a dataset of geolocalised messages. The users have to solve collaboratively a certain task and they have at their disposal a chat utility that allows them to coordinate. This utility is different in the two interfaces. In one case the users can anchor their conversation to specific places on the map. In the other case it resemble a traditional chat, standing on a side of the map.

As I was not sure of the task I thought that a bit of brainstorming could be optimal to reach a possible solution. Both Gaelle and Mirweis were very imaginative to sketch scenarios and to highlight concerns. Two mayor concerns emerged: on the case of the georeferenced chat it will be difficult to represent the temporal aspect of the discourse, which is embedded in the vertical sorting of the traditional chat; additionally, the dimensions of the chat tool should be maximally comparable in both situations.

Three elements were found fundamental to a good and balanced task: (1) the presence of countable aspects in the requirements of the task: i.e., temporal or numerical as a certain amount of money necessary to solve the task. This should be mixed with spatial elements necessary to accomplish the task: as locating a certain number of similar items, find the shortest path according to the criteria, or locate the best spot to position an activity using different criteria; (2) the difficulty of the task should be balanced so not to have scores too high or too low, so there should be different solutions to the task; (3) the task should not impose any preferential control to the users so ‘to force’ them to discuss possible solutions and act together to reach the goal.

We worked on three different scenarios, which I have to refine to reach a more specific task:

1. Find the closest books on a certain topics among the libraries of EPFL. This might integrate the need for different queries and discussions on where points needs to fall to compose the shortest path;

2. Organise a touristic tour in the center of Geneva for a group of visitors given a certain amount of money and time.

3. Discuss where to position the offices of a startup in Lausanne, given the activities of the company, the transport resources around, the condition of the soil and other factors.