This paper develops a cognitive map of the Yan’an spirit that is based on the performance characteristics of the Yan’an spirit. An entity boundary perception model based on multi-task learning is proposed, and the traditional named entity recognition task is decomposed into entity boundary perception and entity classification, which completes the entity recognition based on MTBA-NER. The entity-oriented relationship extraction is based on the Bi-LSTM network to classify the expression of features, and the softmax function is used to extract the classification results. The political value has always been the value of Yan’an spirit, which has a value of 0.274 today, while the economic value, cultural value and educational value have a value of 0.322, 0.367 and 0.245, respectively. The analysis of the connotation of the Yan’an spirit based on cognitive mapping can help reflect and promote its value in contemporary times.