Randomized trial demonstrates automatic identification of key elements in unstructured information, suggesting improved data processing efficiency.
Key Points
The research aims to develop a method for automatically identifying key elements in unstructured chemical order information using data mining techniques.
Applied generalization processing using forward shortest editing distance and clustering for unstructured information.
Employed a pre-trained language model (BERT) for feature extraction and created a mixed feature vector.
Utilized a least squares support vector machine model to optimize parameters for effective identification.
Demonstrated a significant reduction in the domain adaptation gap post-generalization, below the threshold.
Achieved a Spearman rank correlation coefficient close to 1 for effective feature extraction.
Successfully identified key elements like service location and fault components with high accuracy.