CROSER

148 views
117 views

Published on

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
148
On SlideShare
0
From Embeds
0
Number of Embeds
33
Actions
Shares
0
Downloads
2
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • The matching method adopted by CroSeRto provide meaningful link recommendations to the user is based on three main steps:Service descriptions in languages different from English are translated in English using machine translation tools (Bing APIs)The descriptions translated in English are analyzed by CroSeR using Explicit Semantic Analysis and are indexed using the Vector Space Model, thus each service is represented as a vector of weights in a multi-dimensional spaceGiven an input service, a set of top-k best matching services in the target catalog are retrieved; the matching score for a couple of source and target service is defined by the cosine similarity between the descriptions of the services (both described in the Vector Space Model)Explicit Semantic Analysis describes documents in a Vector Space Model defined using Wikipedia as reference knowledge.
  • “Arbitrati e conciliazioni” (translated as “Arbitrations and conciliations”) –> “Legal -litigation support”3 - CroSeR is able to find matchings that cannot be discovered by simple Machine Translation + keyword comparison; 2 examples are shown in the figure [see the description in the upper most figure; literal translation to English are shown in round brackets]4 – We also notice that we can find several CroSeR recommendations that are meaningful even when not correct wrt the Gold standard. As an example, “Briefwahl”, literally translated as “Absentee Ballot” has been mapped to “Postal Voting” by domain experts. However (according to Wikipedia), Postal Voting and Proxy Voting are two established practices of Absentee Ballot. In this case, it seems that establishing two broaderMatch links between Briefwahl and Proxy Voting, and between Briefwahl and Postal Voting, would be a much more appropriate choice. Although no service described as “Absentee Ballot” occurs in the target service catalog, CroSeR recommends both Proxy Voting and Postal Voting as possible matches for Briefwahl. Using CroSeR, the PA user could modify their links and make more appropriate choices.
  • CROSER

    1. 1. Cross-language Linking of eGov Services to the LOD Cloud Fedelucio Narducci*, Matteo Palmonari*, Giovanni Semeraro° *DISCO, University of Milan-Bicocca, Italy °Department of Computer Science, University of Bari Aldo Moro, Italy Semantic Web Access & Personalization research group “Antonio Bello” http://www.di.uniba.it/~swap AI for Smart Cities Workshop AI*IA 2013 - 25th Year Anniversary XIII Conference Turin (Italy), December 5, 2013
    2. 2. EC 6 axes to evaluate ‘city smartness’ http://www.smart-cities.eu/model.html 2
    3. 3. Cross-language Linking of Open Government Data • A large amount of Open Government Data in many languages*: o 1,000,000+ datasets published online (February 2013) o 40 different countries o 24 different languages *http://logd.tw.rpi.edu/iogds_data_analytics 3
    4. 4. Cross-language Linking of eGov Service Descriptions • Government service catalogs are part of the LOD cloud o Effective Service Delivery (ESD)-toolkit o European Local Government Service List (LGSL) • 2000+ interlinked public services in 6 languages 4
    5. 5. Cross-language Linking of eGov Services Why is it useful? • Advantages for PAs o Compare local service offerings with best practices in other countries o Support interoperability among PAs of different countries and other service providers o Enrich service descriptions with additional information via links to LGSL (e.g., link to life event ontologies) • Advantages for citizens o Find eGov services when in a foreign country o Towards cross-language service access Costly and Error Prone Activity Catalogs of several hundreds of services 5
    6. 6. Cross-language Linking of eGov Services Why is it challenging? ≈ sameAs links • • Semantic heterogeneity o o Challenging cross-language matching problem Most of the approaches: • not a mere “translation” problem cultural bias • Ultra-short descriptions use structural information [Spohr et al. 2011, Fu et al. 2011, Wang et al. 2009] or long textual descriptions [Knoth et al. 2011] or report problems when automatic translation returns descriptions with heterogeneous vocabulary [Hertling & Paulheim 2012] 6
    7. 7. CroSeR Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English o Based on automatic translation and Explicit Semantic Analysis Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) TRY IT @ http://siti-rack.siti.disco.unimib.it:8080/croser/ 7
    8. 8. CroSeR Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English o Based on automatic translation and Explicit Semantic Analysis Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) 8
    9. 9. CroSeR Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in • Load a catalog English o Based on automatic translation and Explicit Semantic Analysis Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) 9
    10. 10. CroSeR Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in • Load a catalog English o Based on automatic translation and Explicit Semantic Analysis • Select a source service Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) 10
    11. 11. CroSeR Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in • Load a catalog English o Based on automatic translation and Explicit Semantic Analysis • Select a source service • Look at the retrieved services (link recommendations) Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) 11
    12. 12. CroSeR Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in • Load a catalog English o Based on automatic translation and Explicit Semantic Analysis • Select a source service • Look at the retrieved services (link recommendations) • Link SKOS broader / exact / narrower match Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) 12
    13. 13. Cross-language Service Retriever • CroSeR o A tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English o Based on automatic translation and Explicit Semantic Analysis Web tool to support the linkage of a source eGov service catalog represented in any language to a target catalog represented in English Based on Machine Translation and Explicit Semantic Analysis (ESA) 13
    14. 14. CroSeR: Matching Approach Machine Translation of Service Descriptions Extraction of ESA-based representations and indexing (Vector Space Model) Top-k Service Retrieval by Cosine Similarity 14
    15. 15. Experiments: Discussion • CroSeR finds matchings that cannot be discovered by machine translation + keyword comparison • CroSeR’s recommendations can support the users to refine the links GT:“Absentee Ballot” 16
    16. 16. References (for details) • Model and Experimental Evaluation @ISWC 2013 o F. Narducci, M. Palmonari, G. Semeraro. Cross-Language Semantic Retrieval and Linking of E-Gov Services. The Semantic Web - ISWC 2013 12th International Semantic Web Conference, Sydney, NSW, Australia, October 21-25, 2013, LNCS 8219, 130-145, Springer, 2013 • System Demo @ECIR 2014 o F. Narducci, M. Palmonari, G. Semeraro. CroSeR: Cross-language Semantic Retrieval of Open Government. 36th European Conference on Information Retrieval, Amsterdam, the Netherlands, April 13-16, 2014. To Appear 17

    ×