Natural Language Processing for Automated Inference

1
Natural Language Processing for Automated
Inference
2
Tokenizer
The pipeline
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
3
Tokenizer
Adapted from PennBioTagger
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
4
Tokenizer
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
5
Tokenizer
Customized tags transduction activation"
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
6
Tokenizer
Sleator Temperley LinkParser Relationship
Extractor
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
7
Tokenizer
Abstracts Relex output from syntactical origins
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
8
Tokenizer
PLN Novamente AI Engine
Gene, Protein, and Malignancy Tagger
Nominalization Tagger
Sentence Extractor
Semantic Mapper
Probabilistic Inference
9
What it does

• Any of the sentences
• Kim kissed Pat.
• Pat was kissed by Kim.
• Is mapped into the set of relationships
• subj(kiss_0, Kim)
• obj(kiss_0, Pat)
• inheritance(kiss_0, kiss)

10
How the semantic mapping rules look like

• The rule
• by(x, y) inheritance(x, transitive_event) ?
  subj(x, y)
• Maps the relex-produced relationship
• by(prevention, inhibition)
• Into the abstract conceptual relationship
• subj(prevention, inhibition)
• Which is suitable for inference by PLN.

11
Example (Bioliterate)
Premise 1 Importantly, bone loss was almost completely prevented by p38 MAPK inhibition. (PID 16447221)
Premise 2 Thus, our results identify DLC as a novel inhibitor of the p38 pathway and provide a molecular mechanism by which cAMP suppresses p38 activation and promotes apoptosis. (PID 16449637)
(Uncertain) Conclusions DLC prevents bone loss. cAMP prevents bone loss.
12
Importantly, bone loss was almost completely prevented by p38 MAPK inhibition. Importantly, bone loss was almost completely prevented by p38 MAPK inhibition.
_subj-n(bone, loss) _obj(prevention, loss) _subj-r(almost, completely) _subj-r(completely, prevention) by(prevention, inhibition) _subj-n(p38 MAPK, inhibition) subj (prevention, inhibition) obj (prevention, loss) obj (inhibition, p38_MAPK) obj (loss, bone)
13
Thus, our results identify DLC as a novel inhibitor of the p38 pathway and provide a molecular mechanism by which cAMP suppresses p38 activation and promotes apoptosis. Thus, our results identify DLC as a novel inhibitor of the p38 pathway and provide a molecular mechanism by which cAMP suppresses p38 activation and promotes apoptosis.
_subj(identify, results) as(identify, inhibitor) _obj(identify, DLC) _subj-a(novel, inhibitor) of(inhibitor, pathway) _subj-n(p38, pathway) subj (inhibition, DLC) obj (inhibition, pathway) inh(pathway, p38)
14
Background knowledge utilized

• Implication
• AND
• inh x causal_event
• inh y causal_event
• subj(y, x)
• subj(x, z)
• subj(y,z)

15
Probabilistic Inference

• Abduction
• Inh inhib1, inhib
• Inh inhib2, inhib
• -
• Inh inhib1, inhib2
• Similarity Substitution
• Eval subj (prev1, inhib1)
• Inh inhib1, inhib2
• -
• Eval subj (prev1, inhib2)

• Deduction
• Inh inhib2, inhib
• Inh inhib,causal_event
• -
• Inh inhib2, causal_event

16
Probabilistic Inference

• And
• Inh inhib2, causal_event
• Inh prev1, causal_event
• Eval subj (prev1, inhib2)
• Eval subj (inhib2, DLC)
• -
• AND
• Inh inhib2, causal_event
• Inh prev1, causal_event
• Eval subj (prev1, inhib2)
• Eval subj (inhib2, DLC)

• Unification
• ForAll (x, y, z)
• Imp
• AND
• Inh x, causal_event
• Inh y, causal_event
• Eval subj (y, x)
• Eval subj (x, z)
• Eval subj (y, z)
• AND
• Inh inhib2, causal_event
• Inh prev1, causal_event
• Eval subj (prev1, inhib2)
• Eval subj (inhib2, DLC)
• -
• Eval subj (prev1, inhib2)

17
Probabilistic Inference

• Implication Breakdown (Modus Ponens)
• Imp
• AND
• Inh inhib2, causal_event
• Inh prev1, causal_event
• Eval subj (prev1, inhib2)
• Eval subj (inhib2, DLC)
• Eval subj (prev1, DLC)
• -
• Eval subj (prev1, DLC)

18
(No Transcript)