Surpassing The Challenges of Inferencing for Natural Language [ READ ]
Aleshinloye Abass Yusuf, Nwojo Nnanna Agwu
The main objective of this research is to improve on the standard tree- matching approach of textual entailment by introducing inference engine for every standard sentence which is more efficient in doing inferencing than the textual entailment approach and the logical approach. The previous approaches have some limitation ranging from lack of background knowledge and inadequate representation of natural language. To overcome the present challenges a compromise is made on both logical and textual entailment approaches. The compromise is referred to as the Normalisation approach, where the inference is achieved by transforming the dependency tree generated from the parser in order to generate a simple version which can be easily handled by the inference engine. The approach involves goal inference; that is, inference from a goal to a set of known facts, which help control the search space by enforcing conditions that will help prove the current goal. The approach can be used in solving ontological problems and challenges that come with inferencing in natural language.
Aleshinloye Abass Yusuf, Nwojo Nnanna Agwu
The main objective of this research is to improve on the standard tree- matching approach of textual entailment by introducing inference engine for every standard sentence which is more efficient in doing inferencing than the textual entailment approach and the logical approach. The previous approaches have some limitation ranging from lack of background knowledge and inadequate representation of natural language. To overcome the present challenges a compromise is made on both logical and textual entailment approaches. The compromise is referred to as the Normalisation approach, where the inference is achieved by transforming the dependency tree generated from the parser in order to generate a simple version which can be easily handled by the inference engine. The approach involves goal inference; that is, inference from a goal to a set of known facts, which help control the search space by enforcing conditions that will help prove the current goal. The approach can be used in solving ontological problems and challenges that come with inferencing in natural language.
