ºÝºÝߣ

ºÝºÝߣShare a Scribd company logo

Temporal Knowledge Acquisition and Modeling

?
?
Cyril Faucher
Cyril Faucher

This document describes a workflow for acquiring and modeling temporal knowledge from natural language expressions about access periods. The workflow includes: 1. Extracting temporal properties from natural language expressions using natural language processing techniques. 2. Representing the temporal properties as a controlled text constrained by a grammar that can be interpreted by machines. 3. Modeling the temporal knowledge intensionally as expressions in a linguistic model and extensionally as concrete dates in a temporal model using model-driven engineering techniques. 4. Exporting the temporal data for visualization and control tasks such as automatically creating iCalendar entries.

1 of 11
Download to read offline
1 Temporal Knowledge Acquisition and Modeling This work is granted by the French Research Agency (ANR- Contint, RelaxMultiMedias 2 project) Cyril Faucher1, Charles Teiss¨¨dre2,3, Jean-Yves Lafaye1, Fr¨¦d¨¦ric Bertrand1 1 L3i, Universit¨¦ de La Rochelle, France cyril.faucher@univ-lr.fr 2 MoDyCo - Universit¨¦ de Paris Ouest Nanterre La D¨¦fense ¨C CNRS 3 Mondeca, France EKAW 2010 - Knowledge Engineering and Knowledge Management by the Masses 11th October-15th October 2010 - Lisbon, Portugal
Introduction ? Context: Leisure and cultural event programming (festival, theater, etc.) & opening dates and hours (restaurant, exhibition, museums) ? Contrib: Temporal Knowledge Acquisition and Modeling Chain ? From natural language expressions to controlled text 2 ? Facilitating the work of human operators for acquisition of temporal properties/knowledge about ¡°Access Periods¡± ? NLP & MDE techniques ? Natural Language Processing ? Model Driven Engineering Natural Language Expressions Temporal properties Controlled text As model Extraction / Generation NLP MDE Faucher et al 2
From Natural Language Expressions to Controlled text ? Natural language expressions ? (1) The museum is open every day except Tuesday and the following French holidays: December 25, January 1, May 1, and August 15. Opening hours: Monday, Thursday, Saturday, Sunday: from 9 a.m. to 6 p.m. Wednesday, Friday: from 9 a.m. to 10 p.m. ? (2) Opening times: Monday to Friday: Lunch (12 noon to 2 p.m.) Dinner (6.30 p.m. to 11 p.m.). Saturday: Dinner (6.30 p.m. to 11 p.m.) ? (3) Opened every day from 10:00 to 18:00, except Tuesdays. ? Controlled text (close to natural language) ? is constrained by a grammar ? can be interpreted by a machine Faucher et al 3
Temporal knowledge ? Temporal knowledge is expressed as a set of intensional expression ? Opening times: Monday to Friday: Lunch (12 noon to 2 p.m.) Dinner (6.30 p.m. to 11 p.m.). Saturday: Dinner (6.30 p.m. to 11 p.m.) ? Translation into extension for visualization/control tasks ? extension / concrete dates: closed on Sunday = closed Sunday 2010-09-19, closed Sunday 2010-09-26, closed Sunday 2010-10-03 Faucher et al 4
Detailed workflow Text input frame Faucher et al 5
Detailed workflow Text input frame Annotations Faucher et al 6 Automatic annotation (Unitex patterns) + post-treatment to lift ambiguities
Detailed workflow Text input frame Annotations Serialization in XMI/XML Faucher et al 7
Detailed workflow 8 Text input frame Annotations Serialization in XMI/XML Calendar editing Text / Calendar Synchronization
Detailed workflow Serialization in XMI/XML Linguistic Model Access Period domain populates Faucher et al 9
Detailed workflow Linguistic Model (specific to the domain of Access Period) Temporal Model (as a pivot) that extends ISO 19108 (generic to define temporal expressions) model transformation Textual representation conforms to the grammar defined in Xtext iCalendar export Temporal data in extension over a time span Faucher et al 10
Conclusions ? Workflow synthesis ? Future work ? Query engine (upon intensional expressions) Natural Language expressions Extraction/ Generation NLP Linguistic Model Controlled text Temporal Model MDE MDE iCalendar MDE Calendar entry auto creation, Visualization and editing Faucher et al 11

More Related Content

Temporal Knowledge Acquisition and Modeling

  • 1. 1 Temporal Knowledge Acquisition and Modeling This work is granted by the French Research Agency (ANR- Contint, RelaxMultiMedias 2 project) Cyril Faucher1, Charles Teiss¨¨dre2,3, Jean-Yves Lafaye1, Fr¨¦d¨¦ric Bertrand1 1 L3i, Universit¨¦ de La Rochelle, France cyril.faucher@univ-lr.fr 2 MoDyCo - Universit¨¦ de Paris Ouest Nanterre La D¨¦fense ¨C CNRS 3 Mondeca, France EKAW 2010 - Knowledge Engineering and Knowledge Management by the Masses 11th October-15th October 2010 - Lisbon, Portugal
  • 2. Introduction ? Context: Leisure and cultural event programming (festival, theater, etc.) & opening dates and hours (restaurant, exhibition, museums) ? Contrib: Temporal Knowledge Acquisition and Modeling Chain ? From natural language expressions to controlled text 2 ? Facilitating the work of human operators for acquisition of temporal properties/knowledge about ¡°Access Periods¡± ? NLP & MDE techniques ? Natural Language Processing ? Model Driven Engineering Natural Language Expressions Temporal properties Controlled text As model Extraction / Generation NLP MDE Faucher et al 2
  • 3. From Natural Language Expressions to Controlled text ? Natural language expressions ? (1) The museum is open every day except Tuesday and the following French holidays: December 25, January 1, May 1, and August 15. Opening hours: Monday, Thursday, Saturday, Sunday: from 9 a.m. to 6 p.m. Wednesday, Friday: from 9 a.m. to 10 p.m. ? (2) Opening times: Monday to Friday: Lunch (12 noon to 2 p.m.) Dinner (6.30 p.m. to 11 p.m.). Saturday: Dinner (6.30 p.m. to 11 p.m.) ? (3) Opened every day from 10:00 to 18:00, except Tuesdays. ? Controlled text (close to natural language) ? is constrained by a grammar ? can be interpreted by a machine Faucher et al 3
  • 4. Temporal knowledge ? Temporal knowledge is expressed as a set of intensional expression ? Opening times: Monday to Friday: Lunch (12 noon to 2 p.m.) Dinner (6.30 p.m. to 11 p.m.). Saturday: Dinner (6.30 p.m. to 11 p.m.) ? Translation into extension for visualization/control tasks ? extension / concrete dates: closed on Sunday = closed Sunday 2010-09-19, closed Sunday 2010-09-26, closed Sunday 2010-10-03 Faucher et al 4
  • 5. Detailed workflow Text input frame Faucher et al 5
  • 6. Detailed workflow Text input frame Annotations Faucher et al 6 Automatic annotation (Unitex patterns) + post-treatment to lift ambiguities
  • 7. Detailed workflow Text input frame Annotations Serialization in XMI/XML Faucher et al 7
  • 8. Detailed workflow 8 Text input frame Annotations Serialization in XMI/XML Calendar editing Text / Calendar Synchronization
  • 9. Detailed workflow Serialization in XMI/XML Linguistic Model Access Period domain populates Faucher et al 9
  • 10. Detailed workflow Linguistic Model (specific to the domain of Access Period) Temporal Model (as a pivot) that extends ISO 19108 (generic to define temporal expressions) model transformation Textual representation conforms to the grammar defined in Xtext iCalendar export Temporal data in extension over a time span Faucher et al 10
  • 11. Conclusions ? Workflow synthesis ? Future work ? Query engine (upon intensional expressions) Natural Language expressions Extraction/ Generation NLP Linguistic Model Controlled text Temporal Model MDE MDE iCalendar MDE Calendar entry auto creation, Visualization and editing Faucher et al 11