Performance Evaluation with knowledge base has 10,000 instances
System performance during processing queries in 800,000 instances knowledge base
System performance during processing queries in 1,000,000 instances knowledge base
Friday, April 29, 2011
Monday, April 25, 2011
Semantic-PlaceBrowser: Understanding Place for Place-Scale Context-Aware Computing
I presented this paper at the Pervasive 2010 Workshop: Ubicomp in the Large: Collaborative Sensing and Collective Phenomena (http://www.socionical.eu/index.php?option=com_content&view=article&id=85&Itemid=91)
However, the proceeding is not published anywhere and many of my colleagues asking me about the paper. I would use my blog to post and introduce a little bit about it.
The aim of the paper is to present an idea that pervasive mobile devices can browser the semantic meaning of sensed items in the physical world. Actually, browsing the physical world have interested many researchers:
Castelli et al [11] with "Browsing the world", Nakamura et al [12] with "Ambient Browser" and Bainbridge [13] "A map-based place-browser" have similar approach. In [11], authors use RFID reader to sense the environment for RFID tags. In [12], is an interesting approach, which not actually sense the world but because the similar title that persuades me to review. The authors implement a browser located in the kitchen, users with RFID tagged in their hand (gloves, bracelet) when user moving their hands toward the computer, the RFID reader in the computer detects tags and automatically moves to the new web links.
Our Semantic PlaceBrowser has similar to the PlaceSense approach in the sensing part, but different from the understanding and browsing for semantic meaning. For illustration, we introduce the sensing for Bluetooth phones. Because Bluetooth phones are pervasive and easy to approach. We have discussed the bluetooth sensing in [10].
Actually, the Semantic PlaceBrowser utilizes the PlaceComm framework architecture (See figure below). The browser focuses on the discovering meaning in the KB from the mobile device.
The different point of the Semantic PlaceBrowser is every tag (bluetooth device) is stored in a knowledge base. It has semantic links to other objects, entities or people such as hasOwner(device, person), ownDevice(person, device).. and so on. Therefore, whenever is detected, we not only know about its presence but also know more about its owner, for example, this device is belong to my friend, so if the Semantic PlaceBrowser found it, it means that my friend is around here.
The browsing is implemented by SPARQL queries send from an agent from user's mobile device. The knowledge base agent from the server side will recieve the SPARQL (in raw text format) then parsing it, then it actually queries the Knowledge base to get an answer.
That is it folks. Thanks for reading.
Link to download the full article:
https://sites.google.com/site/tuannguyenlatrobe/publications/SemanticPlaceBrowserCameraDue.pdf?attredirects=0&d=1
This is the demonstration on how it works: Part 1 Preparation
Part 2: Semantic PlaceBrowser
To get the source code, scripts and Netbeans project, to see how it works. Please visit this address:
https://sites.google.com/site/tuannguyenlatrobe/research/running-semantic-placebrowser
However, the proceeding is not published anywhere and many of my colleagues asking me about the paper. I would use my blog to post and introduce a little bit about it.
The aim of the paper is to present an idea that pervasive mobile devices can browser the semantic meaning of sensed items in the physical world. Actually, browsing the physical world have interested many researchers:
Castelli et al [11] with "Browsing the world", Nakamura et al [12] with "Ambient Browser" and Bainbridge [13] "A map-based place-browser" have similar approach. In [11], authors use RFID reader to sense the environment for RFID tags. In [12], is an interesting approach, which not actually sense the world but because the similar title that persuades me to review. The authors implement a browser located in the kitchen, users with RFID tagged in their hand (gloves, bracelet) when user moving their hands toward the computer, the RFID reader in the computer detects tags and automatically moves to the new web links.
Our Semantic PlaceBrowser has similar to the PlaceSense approach in the sensing part, but different from the understanding and browsing for semantic meaning. For illustration, we introduce the sensing for Bluetooth phones. Because Bluetooth phones are pervasive and easy to approach. We have discussed the bluetooth sensing in [10].
Actually, the Semantic PlaceBrowser utilizes the PlaceComm framework architecture (See figure below). The browser focuses on the discovering meaning in the KB from the mobile device.
The different point of the Semantic PlaceBrowser is every tag (bluetooth device) is stored in a knowledge base. It has semantic links to other objects, entities or people such as hasOwner(device, person), ownDevice(person, device).. and so on. Therefore, whenever is detected, we not only know about its presence but also know more about its owner, for example, this device is belong to my friend, so if the Semantic PlaceBrowser found it, it means that my friend is around here.
The browsing is implemented by SPARQL queries send from an agent from user's mobile device. The knowledge base agent from the server side will recieve the SPARQL (in raw text format) then parsing it, then it actually queries the Knowledge base to get an answer.
That is it folks. Thanks for reading.
Link to download the full article:
https://sites.google.com/site/tuannguyenlatrobe/publications/SemanticPlaceBrowserCameraDue.pdf?attredirects=0&d=1
This is the demonstration on how it works: Part 1 Preparation
Part 2: Semantic PlaceBrowser
To get the source code, scripts and Netbeans project, to see how it works. Please visit this address:
https://sites.google.com/site/tuannguyenlatrobe/research/running-semantic-placebrowser
PlaceSense: A Tool for Sensing the Community
This paper is motivated from the work of Kindberg et al. [1] and Paulos et al. [2] by their contributions:
- The pervasive of Bluetooth mobile phone are turned on and appear everywhere. The mobile phones are named by default or intentionally by their owner that reflex the owners’ personality.
- The re-examine sceenario “Familiar strangers” (Milgram 1970 )using modern technology (Bluetooth). Using bluetooth name discovery , we can found a person appear in range and can check for re-appear, or even the not re-appear.
By applying these ideas, I would like to investigate further steps to see whether:
- Are there any similar thing happens in Melbourne, Australia? Yes, it is. People turn on their mobile bluetooth name and leave it in discoverable mode.
- The community pattern: are there any of my friends went to Preston market on Friday afternoon? Some does, some don’t. Furthermore, if I found one, there might be another guy around there, even thought that guy does not have any devices attached to him. Because they are friend of a friend (FOAF), if you want to meet him, just go there on Friday afternoon, you will have a chance
- Size does matter: many devices appear at specific places at specific time of a day. In PlaceSense there are a result of room usage during weekdays. In addition, if a place suddenly have a huge number of devices appear, there must be something happen there. In the real time pilot study, I sense the Flemmington Race course in Melbourne, and discover more than 6000 different mobile phones. How can I do it ? For the room monitor is easy, because my office is next to the seminar room. I just put my ubuntu server with a bluetooth dongle to scan every 20 seconds during 3 months. You can imagine the rest. As a poor man research, I support my self doing research , therefore I cannot afford of $$$ for a ticket to the luxury place in Flemmington Racecourse. Furthermore, it is not cheap to get a ticket that you can go any reserved marques. Sometimes the research is costly. My solution is apply for a volunteer job on Melbourne Cup day to do the PlaceSense. I did it in 2007, 2008 and repeated in 2009
Event thought the paper has been accepted (2008) and published (late 2008), I still went back to collect more data. I also found something interesting:
- Are there any person (I mean the devices, because the devices cannot travel by themselves to the race course) went here last year still come this year ? Yes, some people still attend this year. Majority are not (based on the bluetooth data) but may be they change their phone) May be the owner has upgraded to a new luxury phone?
- There are many discussion about place and location. In some extends, location is not important. Let look at the Melbourne Cup as an example, a place that have thousands of people get together and enjoy the race is more important than a location (Lat X, Lon Y)? Suppose that next year the Melbourne Cup is moved to Caufield racecourse instead of Flemmington racecourse. Thing that happens at place is very important, it involved people, places.
- After this paper had published. Every time I go to a new place, I turn my PlaceSense on, try to discover are there any “invisible devices” around me? Luckily, I always found some ! Refer back to Weiser’s papers “computers have woven themselves into the fabric of everyday life”. The place now have many things in it: people, devices, social context. It is worth of looking at.
- The outcome of this paper is not only the research data, but also the experiences and relationships that I have. I have a chance to enjoy the Melbourne Cup event (free entrance ticket), learn new culture, meet different people, doing a good thing (raising fund for hospitals that I have never been to, touch wood).
Hardware to do this:
- Bluetooth dongle (from ebay): AU$10 ($8+$2shipping cost)
- Nokia N95 (won on Ebay $240+$20shipping cost), Nokia E65, E61 I borrow from my friends. Because doing sensing, the battery will drain out quickly. Furthermore, when there are many devices around you, one phone is not enough.
Software: J2ME Midlet, Jsr 82 (bluetooth), Netbeans, Sun Wireless Toolkit, but put the scanning procedure to thread, if not, it will stall your phone.
That’s it folks, with the minimal cost, big effort, you can do the costly research. One of the interesting aspect of pervasive computing is develop a future with current technology. Therefore we usually see imagined scenarios in many publications in this area.
Thanks for reading this. Any comments would be appreciated.
Selected publications:
[1] T. Kindberg and T. Jones. “merolyn the phone”: A study of bluetooth naming practices (nominated for the best paper award). In J. Krumm, G. D. Abowd, A. Seneviratne, and T. Strang, editors, Ubicomp, volume 4717 of Lecture Notes in Computer Science, pages 318-335. Springer, 2007.
[2] E. Paulos and E. Goodman. The familiar stranger:anxiety, comfort, and play in public places. In CHI ‘04: Proceedings of the SIGCHI conference on Human fac- tors in computing systems, pages 223-230, New York, NY, USA, 2004. ACM.
Published link: http://portal.acm.org/citation.cfm?id=1671714
PlaceAware: A Tool for Enhancing Social Interactions in Urban Places
This paper’s approaches are different from the PlaceSense paper. Even thought, it is also inspired from Kindberg et al. [1] and Paulos et al. [2]
Instead of scanning for every devices appear near the field. I only look for some pre-known devices. (See the scenario 1). Actually in the scenario 1, it does not require you do any programing, just acknowledge your buddies to name their Bluetooth phone and exchange name to each other. When never both of them are close enough (according to Bluetooth specifications, the Bluetooth field is 10meters), they can “see each other”. The program on devices are just do that job automatically. So the users are just simply turned on their phones. The scanner device need to install the PlaceAware application, the person who need pickup at air port
Scenario 1: Airport pickup. (Actually, this scenario happend after the paper is accepted). But in the prototype, I set up a scenario in my office with my buddies. The main idea is discover a pre-known Bluetooth name appear in the field.
Scenario 2: This scenario is more complicated. I setup a lab and make use of many technologies: OWL-DL (protege), SPARQL, jade agents. The main idea is using SPARQL query to query the Knowledge Base that store anything happened in the community (in this scenario is the room).
In short, the Bluetooth mobile phones are pervasive, we just think of another way of using it as a tag for person discovery 
Even though Bluetooth is slow, and short-range, but its dis-advantages become advantages where people walking speed is much slower in crowded area, where the visibility is limited and within 10m we can easy recognize each other.
Even though Bluetooth is slow, and short-range, but its dis-advantages become advantages where people walking speed is much slower in crowded area, where the visibility is limited and within 10m we can easy recognize each other.Is it useful that building a knowledge base (KB) for a certain place that can capture context as much as possible. Later on, the KB can be used for something else.
Selected references:
[1] T. Kindberg and T. Jones. “merolyn the phone”: A study of bluetooth naming practices (nominated for the best paper award). In J. Krumm, G. D. Abowd, A. Seneviratne, and T. Strang, editors, Ubicomp, volume 4717 of Lecture Notes in Computer Science, pages 318-335. Springer, 2007.
[2] E. Paulos and E. Goodman. The familiar stranger:anxiety, comfort, and play in public places. In CHI ‘04: Proceedings of the SIGCHI conference on Human fac- tors in computing systems, pages 223-230, New York, NY, USA, 2004. ACM.
Subscribe to:
Posts (Atom)