Semiautomatic Generation of Resilient Data Extraction Ontologies

1
Semiautomatic Generation of Resilient Data
Extraction Ontologies

• Yihong Ding
• Data Extraction Group
• Brigham Young University
• Sponsored by NSF

2
Data Extraction Ontology

• Goal extract data from web pages
• Components
• concepts
• relations between the concepts
• participation constraints
• Resilient
• Difficulty manual ontology generation is costly

3
Generation Procedure
Train
Test
Knowledge Selection Processing
Extraction Processing
Database
Knowledge Sources
4
Knowledge Collection

• Assumptions about knowledge base
• general
• contains meaningful relationships
• pre-existing
• XML or easy to transfer to XML
• Current input
• Mikrokosmos ontology Mik
• auxiliary data frame library

5
Selection of Concepts

• PROCEDURE ConceptSelection(Tdoc, Kbase)
• SourceDoc Parse(Tdoc)
• PrimarySelectedConceptsList
  MikroSelection(M-Ontology)
• SecondarySelectedConceptsList
  DataFrameSelection(DF-Library)
• ConflictHandling()
• SelectedSubgraphGeneration()

• MANY ISSUES
• selection strategies, conflict resolution,

6
Basic Selection Strategy

• Select from Mikrokosmos Ontology

• Afghanistan
• smaller than Texas.
• Area 648,000 sq. km.
• Capital--Kabul,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• Population17.7 million.
• Agriculture Wheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

7
Basic Selection Strategy

• Select from Mikrokosmos Ontology
• concept names and their synonyms

• Afghanistan
• smaller than Texas.
• ArealtGeographicalAreagt 648,000 sq. km.
• CapitalltCapitalCitygtltFinancialCapitalgt--Kabul,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• PopulationltPopulationgt17.7 million.
• AgricultureWheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

8
Basic Selection Strategy

• Select from Mikrokosmos Ontology
• concept names and their synonyms
• concept values and their synonyms

• AfghanistanltNationgt
• smaller than TexasltUSStategt.
• ArealtGeographicalAreagt 648,000 sq. km.
• CapitalltCapitalCitygtltFinancialCapitalgt--KabulltCapi
  talCitygt,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• PopulationltPopulationgt17.7 million.
• AgricultureWheatltFoodStuffgtltAgriculturalProductgt,
  corn, barley,rice, cotton, fruit, nuts, karakul
  pelts, wool, mutton.

9
Basic Selection Strategy

• Select from Mikrokosmos Ontology
• concept names and their synonyms
• concept values and their synonyms
• Select from Data Frame Libraries

• Afghanistan
• smaller than Texas.
• Area 648,000 sq. km.
• Capital--Kabul,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• Population17.7 million.
• Agriculture Wheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

10
Basic Selection Strategy

• Select from Mikrokosmos Ontology
• concept names and their synonyms
• concept values and their synonyms
• Select from Data Frame Libraries
• extract result based on the data frames

• Afghanistan
• smaller than Texas.
• Area 648,000ltAreagtltMileagegt sq. km.
• Capital--Kabul,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• Population17.7ltTimegt millionltPopulationgtltPricegt.
  
• Agriculture Wheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

11
Document-Level Conflict

• Afghanistan
• smaller than Texas.
• Area 648,000ltAreagtltMileagegt sq. km.
• CapitalltCapitalCitygtltFinancialCapitalgt--KabulltCapi
  talCitygt,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• Population17.7ltTimegt millionltPopulationgtltPricegt.
  
• Agriculture Wheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

12
Concept-Level Conflict

• Afghanistan
• smaller than Texas.
• ArealtGeographicalAreagt 648,000ltAreagt sq. km.
• Capital--Kabul,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• PopulationltPopulationgt 17.7 millionltPopulationgt.
  
• Agriculture WheatltFoodStuffgtltAgriculturalProductgt
  , corn, barley,rice, cotton, fruit, nuts, karakul
  pelts, wool, mutton.

13
Relation Retrieval

• Theoretical solution
• all paths in the subgraph
• too expensive NP-Complete
• Heuristic solution
• find the shortest path between any two nodes
• set a threshold distance

14
Participation Constraints

• AfghanistanltNationgt
• smaller than Texas.
• Area 648,000 sq. km.
• CapitalKabulltCapitalCitygt,
• Other cities--Kandahar Mazar-e-Sharif Konduz
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• Population 17.7 million.
• Agriculture Wheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

CapitalCity 11 IsA.CITY.PartOf Nation 11
15
Participation Constraints (cont.)

• AfghanistanltNationgt
• smaller than Texas.
• Area 648,000 sq. km.
• Capital--KabulltCitygt,
• Other citiesltCitygt--KandaharltCitygt
  Mazar-e-SharifltCitygt KonduzltCitygt
• Terrain Landlocked mostly mountains and desert.
• Climate Dry, with cold winters and hot summers.
• Population 17.7 million.
• Agriculture Wheat, corn, barley,rice, cotton,
  fruit, nuts, karakul pelts, wool, mutton.

City 11 PartOf Nation 1
16
Performance Evaluation

• Speed of generation
• Precision and recall of the generation process
• Precision and recall of the generated ontology

17
Generation Time with Distance Threshold
18
PR of Generation Process
19
Conclusion

• Data Extraction Ontology generated
• Knowledge sources exploited
• Many issues applied
• Many more to explore