Uyghur Metaphor Detection Via Considering Emotional Consistency

Uyghur Metaphor Detection Via Considering Emotional Consistency Qimeng Yang Xinjiang University yangqm@xju.edu.cn Long Yu Xinjiang University yul@xju.edu.cn Shengwei Tian Xinjiang University tsw@xju.edu.cn Jinmiao Song Xinjiang University sjm@dlnu.edu.cn Abstract Metaphor detection plays an important role in tasks such as machine translation and human- machine dialogue. As more users express their opinions on products or other topics on social media through metaphorical expressions, this task is particularly especially topical. Most of the research in this field focuses on English, and there are few studies on minority languages that lack language resources and tools. Moreover, metaphorical expressions have different meanings in different language environments. We therefore established a deep neural network (DNN) framework for Uyghur metaphor detection tasks. The proposed method can focus on the multi- level semantic information of the text from word embedding, part of speech and location, which makes the feature representation more complete. We also use the emotional information of words to learn the emotional consistency features of metaphorical words and their context. A qualitative analysis further confirms the need for broader emotional information in metaphor detection. Our results indicate the performance of Uyghur metaphor detection can be effectively improved with the help of multi-attention and emotional information. 1 Introduction With the rapid development of digital technology and the Internet, social media has become a powerful platform where people can express their opinions on various topics such as politics, finance, education, and other social issues. It is worth noting that more users use a lot of metaphors in online texts to express their thoughts and emotions. According to statistical research, metaphors appear in every three sentences in natural language (Cameron, 2003; Steen et al., 2010; Shutova and Teufel, 2010). Metaphor involves not only language expression, but also the cognitive process of conceptual knowledge (Lakoff, 2003). According to Lakoff and Johnson, metaphor is a conceptual mapping. More specifically, metaphor is a concept used to describe another concept. They are widely used in oral and written language to convey rich linguistic and emotional information. For instance, in the metaphorical utterance: knowledge is treasure., we use treasure to describe knowledge to emphasize that knowledge can be valuable. To take another metaphorical instance as an example: this is an ocean of flowers. Ocean has broad characteristics, which means that the flower area is large. Metaphor detection is an important subtask of natural language processing, which provides a more complete representation for semantic analysis (Lakoff, 2003). More- over, interpreting metaphors helps to improve the performance of tasks such as machine translation and human-machine dialogue analysis (Rentoumi et al., 2012). The existing computational models of metaphor detection are mainly based on lexicon- s (Mohler et al., 2013; Dodge et al., 2015) and supervision methods (Gao et al., 2018; Stowe et al., 2019; Le et al., 2020). Lexicon-based methods do not require data annotation, but they cannot detect new metaphoric usage and capture contextual information. The supervised method can obtain text context in- formation from the sentence level, thereby obtaining broader semantic information. However, it requires …

Content truncated to 500 words