Low Resource Domain Subjective Context Feature Extraction via Thematic Meta-learning

Abstract

The volume of the data is directly proportional to the model's accuracy in data analytics for any particular domain. Once a developing field or discipline becomes apparent, the scarcity of the data volume becomes a challenging proponent for the correctness of a model and prediction. In the proposed state-of-the-art, a transitive empirical method has been used within the same contextual domain to extract features from a low-resource part via a heterogeneous field with factual data. Even though an example of text processing has been used for brevity, it is not limited. The success rate of the proposed model is 78.37%, considering model performance. But when considering human subject matter experts, the accuracy is 81.2%.