1 / 69

Information Access through Textual Entailment: The Experience of the QALL-ME project

Information Access through Textual Entailment: The Experience of the QALL-ME project. Bernardo Magnini FBK-irst, Trento, Italy. Outline. The Qallme scenario Semantic Interpretation of user queries Suggested direction: textual entailment engines Interacting with the user

kalb
Download Presentation

Information Access through Textual Entailment: The Experience of the QALL-ME project

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Information Access through Textual Entailment: The Experience of the QALL-ME project Bernardo Magnini FBK-irst, Trento, Italy

  2. Outline The Qallme scenario Semantic Interpretation of user queries Suggested direction: textual entailment engines Interacting with the user Suggested direction: provide answers with as much structure as possible (RDF) Porting the system Suggested direction: learn as much as possible from data (user questions) Conclusions

  3. QALLME Question Answering Learning Technologies in a Multilingual and Multimodal Environment • Reference: FP6 IST-033860 • Contract Type: STREP • Start date: October 1st, 2006 • Duration: 36 months • Project Funding: 2.82 M euros http://qallme.itc.it

  4. Query Driven vs Answer Driven Information Access • How many people live in Trento? No answer in the first ten documents using Google. • When did Hitler attack Soviet Union? We find documents containing the question itself, no matter whether or not the answer is actually provided. • Current information access is query driven. • Question Answering proposes an answer driven approach to information access. See how Google and Yahoo answer to “Who is Bill Clinton?”

  5. SMS SMS INPUT OUTPUT VOICE VOICE TEXT TEXT MMS VIDEO DIGITAL ASSISTANT QALL-ME Scenario • Mobile Devices: Mobile Phones & PDA • Question Input: Voice/SMS • Answer Output: Voice/SMS/MMS/Digital Assistant (Images/Audio/Video/Maps and geo-referenced interactive maps)

  6. halloI am in Trento and I would like to visit a church in the centre of the town I would like to know the name and the location of one of these churches thanks QALL-ME: Requests from the QALL-ME benchmark

  7. halloI am in Trento and I would like to visit a church in the centre of the town I would like to know the name and the location of one of these churches thanks QALL-ME Questions To greet from the QALL-ME benchmark

  8. halloI am in Trento and I would like to visit a church in the centre of the town I would like to know the name and the location of one of these churches thanks QALL-ME Questions To contextualise from the QALL-ME benchmark • This is explicit context • Time is implicit

  9. hallo I am in Trento and I would like to visit a church in the centre of the town I would like to know the name and the location of one of these churches thanks QALL-ME Questions from the QALL-ME benchmark To ask

  10. halloI am in Trento and I would like to visit a church in the centre of the town I would like to know the name and the location of one of these churches thanks QALL-ME Questions from the QALL-ME benchmark To thank

  11. QALL-ME Resources • Qallme benchmark Acquisition for four languages (about 12,000 requests in total). Semantic annotations: transcriptions, speech acts, EAT, translations • Qallme Ontology: version 4 Both the QALL-ME benchmark and QALL-ME ontology are being made incrementally available at the project website(http://qallme.fbk.eu) under a creative common license Two papers at LREC 2008

  12. TOWN • TRENTO • Address • - VIA VERDI 3 QALL-ME Mobile Infrastructure QALL-ME Webservices Server Side APPLICATION Resource Interface (German/ English) Webservices Front-End APPLICATION Voicedata ASR Engine Manager ASR Resource Interface Virtual Phone Engine CLIENT LIBRARY API IP IP Application data TTS Engine Manager TTS Resource Interface IP IP API Waycom srl, Demo Prototype

  13. Showcases Cinema and Accommodation domain Automatic procedures for daily updating (Trento) Distributed services Cross-language More complex questions Mobile showcase Infrastructure has been consolidated Run on Comdata server Nokia N95 with GPS Speech input (Italian only) Cross-language: SMS only Navigation Text to Speech

  14. Shared Semantic representation Local Information Sources Service Provider English Answer Extractor German Answer Extractor QALL-ME central QA planner Spanish Answer Extractor Italian Answer Extractor Question Type Ontology Answer Type Ontology Speech Recognizers Dialog Models QALL-ME architecture

  15. Structured and Unstructured Data

  16. QALL-ME in a nutshell Presentation output Question A Entailment Engine Answer Representation Training M Presentation Template Question Annotation Qallme Ontology M QALL-ME Question Collection SM M User Data

  17. QALL-ME Architecture

  18. Outline The Qallme scenario Semantic interpretation of user queries Suggested direction: Entailment Engine Presenting information How to build the system Conclusions

  19. Question Interpretation Domain ontology (entailment-based Relation Extraction) Given: A domain ontology

  20. Question Interpretation Domain ontology (entailment-based RE) Given: A domain ontology describing binary relations of interest

  21. Question Interpretation Domain ontology (entailment-based RE) Question Given: A domain ontology describing binary relations of interest A natural language question

  22. Question Interpretation Domain ontology (entailment-based RE) Question Given: A domain ontology describing binary relations of interest A natural language question Determine ALL the relations of interest expressed by the question

  23. Question Interpretation (entailment-based RE) Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10

  24. Question Interpretation (entailment-based RE) Q1 Q2 Q3 Q4 Q5 Out of domain questions Q6 Q7 Q8 Q9 Q10

  25. The task: example INPUT: “ What science fiction movie can I see today at cinema Astra in Trento?” OUTPUT:

  26. The task: example R2: HasGenre(Movie,Genre) INPUT: “ What science fiction movie can I see today at cinema Astra in Trento?” OUTPUT: R2

  27. The task: example R5: IsInDestination(Cinema, Destination) INPUT: “ What science fiction movie can I see today at cinema Astra in Trento?” OUTPUT: R2, R5

  28. The task: example R7: IsInSite(Movie, Site) INPUT: “ What science fiction movie can I see today at cinema Astrain Trento?” OUTPUT: R2, R5, R7

  29. The task: example R8: HasDate(Movie, Date) INPUT: “ What science fiction movie can I see todayat cinema Astra in Trento?” OUTPUT: R2, R5, R7,R8

  30. Textual Entailment t:The technological triumph known as GPS … was incubated in the mind of Ivan Getting. h: Ivan Getting invented the GPS. TE tutorial at ACL 2007, Dagan, Roth, Zanzotto

  31. Applied Textual Entailment A directional relation between two text fragments: Text (t) and Hypothesis (h): • Operational (applied) definition: • Human gold standard - as in NLP applications • Assuming common background knowledge – which is indeed expected from applications TE tutorial at ACL 2007, Dagan, Roth, Zanzotto

  32. Distance-Based TE Engine Determines the best (less costly) sequence of edit operations that allow to transform T into H: - Linear distance - Tree Edit Distance Determines the cost of the three edit operations (insertion, deletion, substitution) Each rule has a probability representing the degree of confidence of the rule. Rules can be at different levels (e.g. lexical, syntactic)

  33. Entailment-based QA over structured data Input question Pattern repository Q: “Where is cinema Astra located?” Entailment engine

  34. Entailment-based QA over structured data Input question Pattern repository Q: “Where is cinema Astra located?” Entailment engine Q  P4

  35. Entailment-based QA over structured data Input question Pattern repository Q: “Where is cinema Astra located?” Entailment engine CONSTRUCT ?address WHERE { ?cinema rdf:type tourism:Cinema ?cinema tourism:name “Astra”. ?cinema tourism:hasPostalAddress ?addr. ?addr tourism:street ?address } Q  P4

  36. Entailment-based QA over structured data Input question Pattern repository Q: “Where is cinema Astra located?” Entailment engine CONSTRUCT ?address WHERE { ?cinema rdf:type tourism:Cinema ?cinema tourism:name “Astra”. ?cinema tourism:hasPostalAddress ?addr. ?addr tourism:street ?address } Q  P4 A: Corso Buonarroti, 16 - Trento Answer

  37. Entailment-based QA over structured data Input question Pattern repository Q: “What’s the address of Astra?” Entailment engine CONSTRUCT ?address WHERE { ?cinema rdf:type tourism:Cinema ?cinema tourism:name “Astra”. ?cinema tourism:hasPostalAddress ?addr. ?addr tourism:street ?address } Q  P4 A: Corso Buonarroti, 16 - Trento Answer

  38. Entailment-based QA over structured data Input question Pattern repository Q: “Where can I find a cinema in the city centre?” Entailment engine CONSTRUCT ?address WHERE { ?cinema rdf:type tourism:Cinema ?cinema tourism:name “Astra”. ?cinema tourism:hasPostalAddress ?addr. ?addr tourism:street ?address } Q  P4 A: Corso Buonarroti, 16 - Trento Answer

  39. Entailment-based QA over structured data Input question Pattern repository Q: “I want to see a movie at Astra. Where is it?” Entailment engine CONSTRUCT ?address WHERE { ?cinema rdf:type tourism:Cinema ?cinema tourism:name “Astra”. ?cinema tourism:hasPostalAddress ?addr. ?addr tourism:street ?address } Q  P4 A: Corso Buonarroti, 16 - Trento Answer

  40. Entailment-Based QA Language variations are held at textual level. Alleviate the need of lexical mapping (as in traditional NLI systems) Any textual entailment approach/algorithm can be used Distance-based, Machine Learning based Entailment rules with lexical and syntactic information Linguistic phenomena are independent from the database organization Re-usable across different tasks (e.g. Relation Extraction) Does not change in case of open domain QA

  41. Outline The Qallme scenario Semantic Interpretation of user queries Presenting information Suggested direction: provide answers with as much structure as possible (RDF) How to build the system Conclusions

  42. QALLME: RDF-based output RDF is a standard for representing knowledge in the Semantic Web RDF is independent both from languages and from media, allowing specific presentation components to be designed on top of it. All reasoning capabilities allowed by RDF will be available in order to draw inferences from answers. In order to represent the informative content of an answer, it seems natural to re-use concepts and relations already defined for the QALL-ME Ontology, rather then define a new set of predicates. Howeverthe informative content is not adequate for generating interactive QA presentations

  43. A closer look to SPARQL queries CONSTRUCT{ … } WHERE{ … }

  44. A closer look to SPARQL queries CONSTRUCT{ … } WHERE{ … } “Construct” portion Selects fragments of the ontology, that represent the “answer” (core answer PLUS relevant additional information, for different answer presentation strategies)

  45. A closer look to SPARQL queries CONSTRUCT{ … } WHERE{ … } “Construct” portion Returns fragments of the ontology in the form of an RDF graph, that represent the “answer” (core answer PLUS relevant additional information, useful for answer presentation) “Where” portion Represents the constraints necessary for answer extraction

  46. CONSTRUCT portion IN: What’s on at Modena? CONSTRUCT {?event qmo:hasPeriod ?period . ?event qmo:isInSite ?cinema . ?event qmo:hasEventContent ?movie . ?movie rdf:type ?movietype . ?movie qmo:name ?moviename . ?cinema qmo:hasGPSCoordinate ?coordinate . ?cinema qmo:name ?cinemaname . ?cinema qmo:hasPostalAddress ?postaladdress . ?postaladdress qmo:isInDestination ?destination . … qma:AnswerInstance a qma:AnswersObject ; qma:hasAnswerValue ?movie } 

  47. CONSTRUCT portion IN: What’s on at Modena? hasPeriod period event CONSTRUCT {?event qmo:hasPeriod ?period . ?event qmo:isInSite ?cinema . ?event qmo:hasEventContent ?movie . ?movie rdf:type ?movietype . ?movie qmo:name ?moviename . ?cinema qmo:hasGPSCoordinate ?coordinate . ?cinema qmo:name ?cinemaname . ?cinema qmo:hasPostalAddress ?postaladdress . ?postaladdress qmo:isInDestination ?destination . … qma:AnswerInstance a qma:AnswersObject ; qma:hasAnswerValue ?movie }  isInDestination Destination postalAddress isInSite hasPostalAddr. cinema hasEventContent hasGPSCoord. movie coordinate type name name movietype cinemaName moviename

  48. WHERE portion IN: What’s on at Modena? CONSTRUCT { … }  WHERE{ ?event qmo:hasPeriod ?period . ?event qmo:isInSite ?cinema . … { ?cinema qmo:name ”Supercinema Modena" } UNION { ?cinema qmo:name "Multisala Modena" } } . … FILTER (xsd:dateTime("2008-12-05T14:19:55") <= xsd:dateTime(fn:string-join(fn:string-join(xsd:string(?date),"T"),xsd:string(?time)))) … } …the name of the cinema is “SUPERCINEMA MODENA” or “MULTISALA MODENA”

  49. WHERE portion IN: What’s on at Modena? CONSTRUCT { … }  WHERE{ ?event qmo:hasPeriod ?period . ?event qmo:isInSite ?cinema . …  { ?cinema qmo:name ”Supercinema Modena" } UNION { ?cinema qmo:name "Multisala Modena" } } . … FILTER (xsd:dateTime("2008-12-05T14:19:55") <= xsd:dateTime(fn:string-join(fn:string-join(xsd:string(?date),"T"),xsd:string(?time)))) … } …the movie should be TODAY, and AFTER THE TIME OF THE QUERY

  50. Resulting RDF graph IN: What’s on at Modena? hasPeriod period event isInDestination Trento postalAddress HasDatePeriod isInSite hasPostalAddr. HasTimePeriod Dateperiod 11°7′0′′E cinema Timeperiod Longitude hasEventContent StartDate hasGPSCoord. Latitude movie StartTime 12/11/2008 46°4′0′′N coordinate type 21.00 name name Crime Modena La Fuga

More Related