Individual Linguistic Granular Computing: A Granulation-Degranulation-Based Approach.

Area of Science:

• Artificial Intelligence
• Computational Linguistics
• Uncertainty Quantification

Background:

• Current linguistic granular computing often uses interval-based granulation with random sampling for degranulation.
• This approach faces limitations in handling semantic ambiguity and contextual variability effectively.
• Existing methods struggle to align computational interpretations with human perception.

Purpose of the Study:

• To propose a novel granulation-degranulation approach for linguistic granular computing.
• To address uncertainty in semantic representation and contextual choices using probability sampling.
• To enhance the alignment of computational linguistic models with human perception.

Main Methods:

• Developed a probability-sampling-based degranulation method utilizing truncated and mirrored power-law distributions.
• Established a relationship between the power-law index and linguistic term interpretation uncertainty, based on the central limit theorem.
• Designed an optimization framework integrating interval-based granulation with probability-sampling degranulation.

Main Results:

• The proposed approach demonstrated superior performance in semantic analysis of online reviews.
• Experimental validation on aircraft engine risk assessment confirmed the method's effectiveness and practicality.
• Comparative analysis showed advantages over traditional computing-with-words models and large language models.

Conclusions:

• The novel probability-sampling-based degranulation method effectively handles linguistic uncertainty.
• The integrated approach improves the accuracy and human-like interpretation of semantic data.
• This research offers a significant advancement in linguistic granular computing and its applications.