Aspect-Based Sentiment Analysis (ABSA) is a task to detect the sentiment of specific aspects in text.For example, in the sentence “This phone has a great screen, but the battery is too small”, there are two aspects, “screen” and “battery”, and their sentiment polarities are positive and negative respectively.

ABSA is widely used. With it, research teams can analyze customers' feedback on products or services in many aspects and extract valuable insights from them. However, ABSA requires word-level recognition of various aspects involved in the training samples when labeling samples, so labeling training data for ABSA becomes a tedious task.