DomainSenticNet: An Ontology and a Methodology Enabling Domain-Aware Sentic Computing.

Area of Science:

• Sentic computing
• Natural Language Processing
• Knowledge Representation

Background:

• SenticNet and OntoSenticNet are key resources in sentic computing.
• Existing sentic computing applications lack domain-specific awareness.
• There is a need for adaptable methodologies to build domain-aware resources.

Purpose of the Study:

• To propose DomainSenticNet, an extension of the OntoSenticNet ontology.
• To introduce an unsupervised methodology for developing domain-aware sentic applications.
• To improve the interpretability of sentiment analysis in specific domains.

Main Methods:

• Extended OntoSenticNet to create DomainSenticNet, incorporating semantically related concepts and domain distributional information.
• Developed an unsupervised methodology to mine concepts from WordNet and Probase.
• Computed domain distributional information from Kickstarter crowdfunding campaigns.

Main Results:

• DomainSenticNet integrates 100,000 concepts with domain-specific information.
• The unsupervised methodology is replicable and adaptable.
• Applied to a Kickstarter tool, DomainSenticNet improved sentiment intensity interpretability.

Conclusions:

• DomainSenticNet facilitates the development of domain-aware sentic applications.
• The methodology supports the creation of adaptable, domain-specific resources.
• This work advances sentic computing research in domain-aware contexts and hybrid sentiment analysis systems.