WP6 meeting at Beijing, November 15-16, 2010

Contents

Sessions

November 15, 2010 20:00
Participants:
- Cefriel: Emanuele, Daniele
- Siemens: Yi (Huang)
- WICI: Yi (Zeng), Joshua

November 16, 2010 9:30
Participants:
- Saltlux: Tony, Stanley, Kevin
- VUA: Zhisheng, Gaston
- WICI: ...
- CEFRIEL: Emanuele, Daniele

Emanuele explained the sota of C-SPARQL and its integration into LarKC
Yi Z. presented the WICI's work done on Twitter to profile a user and his related interests WP2-WP6-WICI.ppt
- NEXT: it will be ported in order to use RDF
Joshua explained his work done on stream/RDF and Twitter
We all agree that all the contributions (WP2/WICI, WP3, WP4/CEFRIEL) provide a well-coverage of LarKC

Summary of the available components:
- SUNS (Siemens)
  - Input: RDF & RDF streams
  - Dataset: Glue & others
  - QL: SPARQL with probability
  - Output: <RDF, probability>, RDF streams (?)
  - Tech: ML for RDF
- C-SPARQL engine (CEFRIEL)
  - Input: RDF streams
  - Dataset: Glue
  - QL: C-SPARQL
  - Output: SPARQL outputs, RDF streams
  - Tech: DSMS, OWL-RL and SPARQL (1.1 - aggregates and subqueries)
- Twitlogic (Joshua)
  - Input: RDF, XML streams
  - Dataset: DBpedia, Twitter, Semantic Web dog food
  - QL: SPARQL
  - Output: SPARQL outputs
  - Tech: Semantic annotations, SPARQL and ontology mapping
- I-?ReaSearch (WICI)
  - Input: RDF
  - Dataset: User description, relationship network of the user (Twitter)
  - QL: ?
  - Output: Interests of the user
  - Tech: Cognitive inspired
We can combine all this component in order to build a recommender system of items (movies, music, etc.) that the user will look the next time.
Proposals:

Plan:
- Activate two parallel streams of work
  - Yi Z. and Joshua will start to integrate I-?ReaSearch and Twitlogic
  - Emanuele and Yi H. will continue the integration of C-SPARQL with SUNS, adding a Twitter stream (leveraging Joshua's Twitlogic)
- At this point we will start to integrate all the parts together (with an incremental approach)

	Dec (beginning)	Dec - Jan	Feb	Mar	Apr	May	Jun - Sept
WP6	scenarios (data sample and queries); data gathering	Wrappers of real data	GUI design			GUI development
WP3	design multiple indepentent WFs	1st experiments; final design (WFs and list of plug-ins)	2nd experiment and 1st implementation in LarKC
DoW	C-SPARQL; WP2; WP4	C-SPARQL; SUNS; WP3	D6.x

Wiki page: LBSMA

WP6 Traffic roadsign session and reasoning on noisy data
Tony says that he liked RND but he did not like RND4SSRM, because:
- too small
- too dirty
He proposed to extend it with other ?GeoData
Zhisheng agrees with this proposal
Tony proposes to use:
- microPOI, a dataset developed by Saltlux on which it'd like to work (dataset)
- korean gov data (dataset)
- ordnance survey (dataset)
- sensor network: real time traffic (stream)
- mobile phone data (stream)
Emanuele likes the idea of integrate RND with stream computing
Gaston warns about the fact that it could become too hard
Zhisheng wants a more "ontological" data
VUA requirements:
- what data would we like to use?
- we want Semantic data
Discussion of the new scenario
- Zhisheng proposed to consider the Traffic Larkc and to continue the development
- Emanuele says that it will be too hard to move Milano Traffic Application in Seoul and viceversa
- People flow was discussed as a alternative scenario
- Gaston proposes to work on an application to suggest how to fix OSM
- Tony says that we needed scale and reasoning intensive applications for larkc
- Emanuele says that in RSM there was enough reasoning, we used sparql because there was not a working rule engine to use instead of it
- Emanuele suggest to perform a careful risk management and start from fixing open street map as suggested by Gaston and move on to more complex scenarios if OSM problems are too easy to fix

Discussion about the OSMF
- Zhisheng proposes to interleave OSMF and RSM checking in one reasoner
- Tony recommends for engineering reasons to cascade OSMF in RSM
- Gaston says that the OSM fixer should receive inputs from the road sign stuff
- Tony says that also the road signs had errors, so the reuse of RSM data in OSMF could be dangerous
- Final decision is to start building the OSMF without the inputs of the RSM, then we could introduce them
- Principle (draft of): to fix a noisy data set one can use an oracle which is assumed to be less noisy than the dataset to fix
Timeplan:
- Feb 2011 - collection of noise examples (Saltlux)/techniques to solve the problem (VUA): set of data, definition of the solution and algorithm to solve them
- May 2010 - design & experiments: software that implements the algorithms (as plugin/workflow)
- July 2010 - implementation of demostrator and integration into the RSM + decision about how to combine components (OSMF + RSM) @ server side
TODO: Saltlux will share with Gaston the road sign ontology of RSM
There will be a joint WP4-WP6 phone call every 2 weeks with Skype + document sharing (Google Doc, Microsft Share)
- TODO: Zhisheng and Gaston will try Google Doc and will provide some feedbacks
- Next phone call: November 30, 10:00 am

Wiki page: RNS4RSM

Emanuele proposes to write an extension of the poster for ISWC In-use track - description of the problem, evaluation & lesson learned (the web of data is really noisy and should be fixed), anticipation of the future works (anticipation of the RND4RSM)
TODO (Saltlux): provide a first draft by November 26
TODO (Emanuele): work on the draft with Saltlux and close it by Dec, 1 and send it to Saltlux for revision
TODO (Saltlux): finalize the paper and send it to Alexey by Dec, 3
TODO (Alexey): to provide feedbacks for Dec, 6 (8:00PM)
TODO (Emanuele): submit to ESWC by Dec, 10

LarkcProject/WP6/Communication/2010-11-15_Beijing-meeting (last edited 2010-11-25 14:13:53 by ?DanieleDellAglio)