Bulskov

Uploaded from authorPOINTLite
Views:
 
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Integration of a formal ontological framework with a linguistic ontology : 

Integration of a formal ontological framework with a linguistic ontology Henrik Bulskov & Hanne Erdman Thomsen Dept. Of Computer Science, Roskilde University & Dept. of Computational Linguistics, CBS

Introduction: OntoQuery: 

Introduction: OntoQuery

Introduction: The SIMPLE lexicon: 

Introduction: The SIMPLE lexicon <SemU id="USEM_N_rugbrxoed_ARF_1" naming="rugbrød (rye bread)" ... weightvalsemfeaturel=”WVSFTemplateArtifactFoodPROT WVSFUnificationPathConcreteentity-ArtifactAgentive-FoodTelicPROT TSVP_FOOD_TS_domaine_D TSVP_ARTIFACT_TS_classificateur_de_nom_C"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_N_bread_ARF_1" semr="SRIsa"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_V_bake_1" semr="SRCreatedby"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_V_eat_1" semr="SRObjectoftheactivity"> <RWeightValSemU weight="ESSENTIAL" target="USEM_N_flour_of_rye_SUF_1" semr="SRMadeof"> </SemU>

Introduction: The prototype ontology on ’bread’: 

Introduction: The prototype ontology on ’bread’

Introduction: OntoLog: 

lack [WRT: vitaminD] (lack of vitamin D / vitamin D deficiency) Introduction: OntoLog Rules of specialization A [Rel: B] is a specialization of A lack [WRT: vitaminD] IS-A lack A [Rel: B1] is a specialization of A [Rel: B2] if B1 is a specialization of B2 lack [WRT: vitaminD] IS-A lack [WRT: vitamin]

A main challenge for the project: 

A main challenge for the project The integration of a comprehensive static semantic lexicon structured in a database with the generative ontology framework (Andreasen et al. DKE 2003) OntoLog: lack [WRT: vitamin] Prototype (SIMPLE): {vitaminmangel} ≠ {mangel,vitamin} vitaminmangel = mangel på vitamin (vitamin deficiency = lack of vitamin)

OntoLog descriptors in SIMPLE?: 

OntoLog descriptors in SIMPLE? IS-A relation supplies superordinate concept Other SIMPLE relations give supplementary relations OntoLog descriptor for a lemma: nearest superordinate [REL1: concept1, REL2: concept2] An example: brød (bread)

SIMPLE: rugbrød (brown rye bread): 

SIMPLE: rugbrød (brown rye bread) <SemU id="USEM_N_rugbrxoed_ARF_1" naming="rugbrød (brown rye bread) … weightvalsemfeaturel=”WVSFTemplateArtifactFoodPROT WVSFUnificationPathConcreteentity-ArtifactAgentive-FoodTelicPROT TSVP_FOOD_TS_domaine_D TSVP_ARTIFACT_TS_classificateur_de_nom_C"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_N_bread_ARF_1" semr="SRIsa"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_V_bake_1" semr="SRCreatedby"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_V_eat_1" semr="SRObjectoftheactivity"> <RWeightValSemU weight="ESSENTIAL" target="USEM_N_flour_of_rye_SUF_1" semr="SRMadeof"> </SemU> naming=”brown rye bread" descr= ”bread [CREATED_BY: bake_1 OBJ_ACTIVITY: eat_1 MADE_OF: flour_of_rye]”

SIMPLE: hvedebrød (wheat bread): 

SIMPLE: hvedebrød (wheat bread) <SemU id="USEM_N_hvedebrxoed_ARF_1" naming="hvedebrød (wheat bread)” … weightvalsemfeaturel=” WVSFTemplateArtifactFoodPROT WVSFUnificationPathConcreteentity-ArtifactAgentive-FoodTelicPROT TSVP_FOOD_TS_domaine_D TSVP_ARTIFACT_TS_classificateur_de_nom_C"> <RWeightValSemU weight="PROTOTYPICAL" comment="Type-defining semantic relation" target="USEM_N_brxoed_ARF_1" semr="SRIsa"> <RWeightValSemU weight="PROTOTYPICAL" comment="Type-defining semantic relation" target="USEM_V_bake_1" semr="SRCreatedby"> <RWeightValSemU weight="PROTOTYPICAL" comment="Type-defining semantic relation" target="USEM_V_eat_1" semr="SRObjectoftheactivity"> <RWeightValSemU weight="ESSENTIAL" comment="Type-defining semantic relation" target="USEM_N_flour_of_wheat_SUF_1" semr="SRMadeof"> </SemU> naming=”wheat bread" descr= ”bread [CREATED_BY: bake_1 OBJ_ACTIVITY: eat_1 MADE_OF: flour of wheat] ”

SIMPLE: brød (bread): 

SIMPLE: brød (bread) <SemU id="USEM_N_brxoed_ARF_1" naming="brød (bread)" … weightvalsemfeaturel=” WVSFTemplateArtifactFoodPROT WVSFUnificationPathConcreteentity-ArtifactAgentive-FoodTelicPROT TSVP_FOOD_TS_domaine_D TSVP_ARTIFACT_TS_classificateur_de_nom_C"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_N_food_FOO_1" semr="SRIsa"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_V_bake_1" semr="SRCreatedby"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_V_eat_1" semr="SRObjectoftheactivity"> <RWeightValSemU weight="ESSENTIAL" target="USEM_N_flour_SUF_1" semr="SRMadeof"> </SemU> naming=”bread" descr= ”food [CREATED_BY: bake_1 OBJ_ACTIVITY: eat_1 MADE_OF: flour] ”

SIMPLE: substans and entity: 

SIMPLE: substans and entity <SemU id="USEM_N_substans_SUB_1" naming="substans" … weightvalsemfeaturel=" WVSFTemplateSubstancePROT WVSFTemplateSuperTypeConcreteentityPROT TSVP_SUBSTANCE_TS_classificateur_de_nom_C"> <RWeightValSemU weight="PROTOTYPICAL" target="USEM_N_entity_ENT_1" semr="SRIsa"> </SemU> SIMPLE: entitet <SemU id="USEM_N_entity_ENT_1" naming="entitet (entity)" comment="dummy"> </SemU> naming=”substans" descr=”entity [] ”

Ontology of bread concepts: 

Ontology of bread concepts naming=”substans" descr=”entity [] ” naming=”stof (stuff)" descr= ”substans [] ” naming=”næring (nourishment)" descr= ”stuff [OBJ_ACTIVITY: consumption_1] ” naming=”mad (food)" descr= ”nourishment[OBJ_ACTIVITY: eat_1] ” naming=”brød (bread)" descr= ”food [CREATED_BY: bake_1 OBJ_ACTIVITY: eat_1 MADE_OF: flour] ” naming=”hvedebrød (wheat bread)" descr= ”bread [CREATED_BY: bake_1 OBJ_ACTIVITY: eat_1 MADE_OF: flour of wheat] ” naming=”rugbrød (brown rye bread)" descr= ”bread [CREATED_BY: bake_1 OBJ_ACTIVITY: eat_1 MADE_OF: flour of rye] ”

Many descriptors: 

Many descriptors naming=”substans" descr=”entity [] ” naming=”stof (stuff)" descr= ”substans [] ” naming=”næring (nourishment)" descr= ”stuff [OBJ_ACTIVITY: consumption_1] ” naming=”mad (food)" descr= ” nourishment[OBJ_ACTIVITY: eat_1] ” naming=”brød (bread)" descr= ”food [CREATED_BY: bake_1 MADE_OF: flour] ” naming=”hvedebrød (wheat bread)" descr= ”bread [MADE_OF: flour of wheat] ” naming=”rugbrød (brown rye bread)" descr= ”bread [MADE_OF: flour of rye] ” food [CREATED_BY: bake_1 MADE_OF: flour of rye] nourishment [OBJ_ACTIVITY: eat_1 CREATED_BY: bake_1 MADE_OF: flour of rye] stuff [OBJ_ACTIVITY: eat_1 CREATED_BY: bake_1 MADE_OF: flour of rye]

Choice of descriptors: 

Choice of descriptors The immediately superordinate concept rugbrød (‘brown rye bread’) = bread [MADE_OF: flour [MADE_OF: rye]]) mad bagt af rugmel (‘food baked with rye flour’) = food [CREATED_BY: bake, MADE_OF: flour [MADE_OF: rye]]) The first superordinate where no new features are added The first concept encountered in the top-ontology rugbrød = entity [OBJ_ACTIVITY: eat, CREATED_BY: bake, MADE_OF: flour [MADE_OF: rye]]

Problems in SIMPLE: 

Problems in SIMPLE Repetition of inherited features (brød) Lack of differentiating features (stof: descr= ”substans [] ”) næring ‘nourishment’: {nourishment, stuff [OBJ_ACTIVITY: consumption], stuff [OBJ_ACTIVITY: relational_act] } Synonyms Different superordinate concepts in the IS-A hierarchy Different ’other’ realtions Dummies are not always correctly located in the ontology

Descriptors: 

Descriptors OntoLog descriptor set for each lemma in SIMPLE: A simple descriptor orthographically identical to the lemma. Descriptors created by combining the lemma of the immediately subsuming concept with the supplementary relations coded in the linguistic ontology. In the case of synonyms, the union of the descriptor sets cf. 1 & 2 of the synonyms is used.

A query example: 

A query example Minimal text data base: A: Brød lavet af rug. (’Bread made of rye.’) B: Hævet rugbrød fremstilles med surdej. (‘Raised brown rye bread is made with leaven.’) C: Brød lavet af hvede (‘Bread made of wheat.’)

Query expansion: 

Query expansion {rugbrød (‘brown rye bread’)} Expansion in the current prototype: {1.00/rugbrød (‘brown rye bread’) + 0.50/brød (‘bread’) + 0.25/mad (‘food’)}.

Ranking in the current prototype: 

Ranking in the current prototype 1.00 B: Hævet rugbrød fremstilles med surdej. (‘Raised brown rye bread is made with leaven.’) 0.50 A: Brød lavet af rug. (’Bread made of rye.’) 0.50 C: Brød lavet af hvede. (‘Bread made of wheat.’)

New OntoLog expansion: 

New OntoLog expansion {rugbrød (‘brown rye bread’)} Expansion { 1.00/“rugbrød” (‘brown rye bread’) + 1.00/brød [MADE_OF: rug] (‘bread’[MADE_OF: ‘rye’]) + 0.50/brød (‘bread’) + 0.25/mad (‘food’) }

’New’ ranking: 

’New’ ranking 1.00 B: Hævet rugbrød fremstilles med surdej. (‘Raised brown rye bread is made with leaven.’) 1.00 A: Brød lavet af rug. (’Bread made of rye.’) 0.50 C: Brød lavet af hvede. (‘Bread made of wheat.’)

Concluding remarks: 

Concluding remarks OntoLog descriptor set for each lemma in SIMPLE A simple descriptor orthographically identical to the lemma. Descriptors created by combining the lemma of the immediately subsuming concept with the supplementary relations coded in the linguistic ontology. In the case of synonyms, the union of the descriptor sets cf. 1 & 2 of the synonyms is used. Future: Experiment: various ontology levels for 2 above Problems with dummies