Culture-Related Metaphor Affective Computing

The metaphor is a frequent phenomenon.It is not merely a kind of rhetoric but also a mechanism of human cognition.In recent years,the metaphor has been widely studied in multiple disciplines,such as the cognitive linguistic,the brain science,and the computer science.In the area of natural language processing(NLP),the metaphor computing plays a critical role,influencing the performance of related NLP tasks,such as machine translation,dialogue system,and opinion mining.Different from the previous metaphor computing studies,which focus on metaphor detecting and metaphor understanding,this article takes metaphor affective computing as the task.One of the most important elements in metaphor affective computing is the language environment.It mainly includes two factors,the context and the cultural background.Related works in metaphor computing have realized the usefulness of the context,however,the influence of cultural factors has not been fully detected.This paper believes that the cultural factors can remarkably drive the results of metaphor affective computing.For instance,the readers that are not familiar with Chinese culture may not able to judge the sentiment of the metaphor "the boss is an iron rooster".So this paper brings out the idea of culture-related metaphor affective computing.It has built a culture-related metaphor sentiment analysis system and a culture-related metaphor affective category analysis system.The article proposes the view of "cultural attribute mapping" based on the Conceptual Metaphor Theory.The view holds that the "cultural attributes" of a concept are defined by the related conceptual metaphors in that culture.The work have built a "cultural attribute database" based on the idea of cultural attribute mapping.The cultural attribute database then serves as the source of cultural background information in metaphor sentiment analysis.The work builds a culture-related and attention-based Long-Short Term Memory neural network to perform binary(positive/negative)sentiment classification on metaphors.In the metaphor affective category analysis study,the article notices that the traditional Chinese literature critics always discuss affects along with the mood(tone,atmosphere,artistic conception).The mood is regarded as the consequence of the harmonious interaction of the emotion and the environment.So the article takes the mood as affective feature and proposes a Chinese culture-related taxonomy of affects.The taxonomy divides the mood of text into 12 categories,including Tranquil,Chill,Magnificent,etc.Based on the holistic and context-related characteristics of the mood,this article comes up with the idea of "cultural affect mapping".The mapping is used to describe the mood feature of words.Then the work performs metaphor affective category analysis based on "cultural affect mapping" and a Long-Short Term Memory neural network.In a word,the work proposes the idea of culture-related metaphor sentiment analysis and culture-related metaphor affective category analysis.It has built two culture-related deep learning-based systems to achieve culture-related metaphor affective computing.