Method and device for extracting constructive opinions in online course evaluation content

The invention discloses a method for extracting constructive opinions in online course evaluation content and a device for implementing the method. According to the method, firstly, the prediction probability of a target word is obtained through a BERT coding layer and a full connection layer, then a screening algorithm is improved based on a heuristic multi-span decoding algorithm, the defects of an original algorithm in the aspect of redundancy removal are overcome, and a target description fineness coefficient is introduced to obtain an entity target for course comments. And inputting the entity target into a Bi-LSTM model with an attention mechanism to extract word features, and adjusting weights corresponding to words according to the entity target to obtain a target opinion pair. And finally, extracting constructive opinions from the course evaluation data by using a K-means-based clustering algorithm. According to the method, the most frequently appearing opinions can be effectively extracted from a la