There will be three major components: the client, the SmartMap server and the mapping service.
Each Location points in the Map has Properties, (Say Latitude, Longitude, No. of intersections, Nature of Place Etc). Similarly most often In Smart-Map Error Locations , The Driver makes the error due to some Properties of that Particular Error location.Idea of this Classifier is to build a Model or No. of Models out of this Error Location Properties. And when ever a new location is given to us, We can compare the Property of the new location with our models. If the Property matches then we can give an alert to the Driver, That There is a Probability of making mistake in that particular location of his/her Map.
Currently, this is implemented as a web service that serves xml/rdf in response to uploads in xml/rdf and n3/rdf, using Jena and JSP. It currently acts as a cache for the Basic Mapping service, abstracting it away to allow things like using multiple map services transparently, and integrating information from multiple mapping services. The idea is to implement logic in JSHOP or TRIPLE, allowing more interesting queries. We shall see if we can make a smarter mapping service.
The Basic Mapping service is a semantic wrapper for some existing web service. The current idea is to give DAML-S descriptions to a mapping service and allow the SmartMap service to dynamically discover additional mapping services.
Ontologies & System design will surely be the work for all of us.
Laks Molaga: Smart Map Classifier
Each Location points in the Map has Properties, (Say Latitude, Longitude, No of intersections, Nature of Place Etc). Similarly most often In Smart-Map Error Locations , The Driver makes the error due to some Properties of that Particular location.
Idea of this Classifier is to build a Model or No of Models out of this Error Location Properties. And when ever a new location is given to us, We can compare the Property of the new location with our models. If the Property matches then we can give an alert to the Driver, That There is a Probability of making mistake in that particular location of his/her Map. Full description of the Smart Map Classifier
David Mihalcik: Smart Map Service
David is responsible for the intermediate service that directly processes requests for map information and updates about a user's experience. This will be a triple store, enhanced to deal with the concrete nature of the world. At its most simple, it is a cache and abstraction layer between Laks's user agent and Fan's map service. My work on the system will include the general infrastructure of the service, including the link between the triple store and the inference engine, as well developing as a set of rules to break down routes heirarchically.
Currently the system is very simple. It takes items and exports them to SHOP to do inferencing, then returns SHOP's selected solution, which may include multiple possible routes. It only uses about 10 lines of SHOP code in the planning domain now, but more could be added later to take into account Laks's classifier. For more information about the SmartMap service, and its use of JSHOP, see /MiddleTier.
Fan Lin: Semantic Map Service
I focused on designing a fundamental Semantic Map Service (actually, a driving direction service). The service will return a possible route given a semantic description of start point and end point. Results will be a Semantic description of the route using Daml+OIL ontology. The service was written in JWSDP (Java Web Service Developer Pack), so it is a standard web service. With WSDL, we can easily add Daml-S description to the service. With the help of intelligent wrapper, the service can wrapp up the different existing map services such as Yahoo Map Service, MSN Map Service, Mapquest etc.
I tried several wrapper induction algorithms to generate wrapper for the service. All of them are too fixed to deal with the modifications of the web page style. So I choose to write the intelligent wrapper by hand. The basic idea is to consider the map description as a specific language. Then the pattern recognition methods can be used to search for the location of the description and NLP technologies (such as Context-Sensitive Grammer) can be used to parse the structure of this language.
The map description language has a specific vocabulary such as "Turn", "Left", "Right", "Start" etc. Using data mining (Association Rules mining), I got a list of distinguish vocabulary. If we investigate the distribution of these vocabulary and then use dynamic clustering to segement the web page, we can easily get the location of the description. The one-dimension clustering is trivial, but in the experiments, it definitly produced a very high precision in seeking the location.
Part of Speech
The identification of POS in all English vocabulary is difficult. However, in map description language, it is also trivial for the limited vocabulary. With statistical method, we can easily tell that "Turn", "Merge", "Enter" are verbs; "on", "onto", "towards" are preposition, and so on.
The basic grammer of map description language is simple: Action + prep + Location + Distance description. Like natural language, it has a lot of variations. I constructed a automata to parse this context-sensitive grammer.
In the experiments, the wrapper can handle different web page styles in Yahoo Map Service, MSN Map Service and MapQuest. It demonstrated my intention: the wrapper will deal with the language itself rather than the web page style. These NLP methods are effective because the language is simple. And this algorithm is a robust way to wrap up existing information. ( Algorithm is robust doesn't mean that my implementation is robust :p )
 Nicholas Kushmerick, Daniel S. Weld, Robert Doorenbos, Wrapper Induction for Information Extraction, Intl. Joint Conference on Artificial Intelligence, 1997, pp. 729-737
 Chun-Nan Hsu, Ming-Tzung Dung, Generating Finite-State Transducers For Semi-Structured Data Extraction From The Web, Information Systems, 1998(8), pp. 521-538
 Craig A. Knoblock, Kristina Lerman, Steven Minton, Ion Muslea, Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach, IEEE Data Engineering Bulletin, 2000(4), pp. 33-41
 Ion Muslea, Steve Minton, Craig Knoblock, A Hierarchical Approach to Wrapper Induction, Proceedings of the Third International Conference on Autonomous Agents, 1999, pp. 190--197
 Arvind Arasu, Hector Garcia-Molina, Extracting Structured Data from Web Pages, Technical Report
 Dieter Fensel, Ontology-Based Knowledge Management, IEEE Computer, 2002(11), pp. 56-59
 Gabriel L., Somlo, and Adele E.H. (2001). Incremental Clustering for Profile Maintenance in Information Gathering Web Agents. In: Proc. of the fifth Int. Conf. on Autonomous agents, pages 262-269.
 Mitchell T. (1997). Machine Learning, New York: McGraw-Hill,
 Christopher D. Manning and Hinrich Schutze (1999), Foundations of Statistical Natural Language Processing, MIT Press
|Nov. 21||Project Imagination||Finish|
|Nov. 28||Architecture/Interface Design, Ontologies, Feasibility Study||Start|
|Dec. 5||Design detail function parts||Wait|
|Dec. 12||Collaborative testing, Bug fixing||Wait|
|Dec. 19||Continue testing or Function extention||Wait|
Semantic Map Service
1. Ontology 2. Instance 3. Another Travel Ontology
Smart Map Classifier
1. Location Property Ontology 2. Loction Property Instances 3. New Location Property Instances 4. DownLoad Smart Map Classifier
Lin: My poor English
Laks : Dont worry Semantic web will understand your English
Laks : I have made a slight change to the Description in the Scenario part
David : I've fleshed out the description of my stuff and cleaned the description a little. I think everyone is supposed to have a one-paragraph description of their section.
Laks : DAML MAP http://www.daml.org/2001/06/map/