CM-Gen: A Neural Framework for Chinese Metaphor Generation with Explicit Context Modelling
4668Authors
nan
Abstract
AbstractNominal metaphors are frequently used in human language and have been shown to be effective in persuading, expressing emotion, and stimulating interest. This paper tackles the problem of Chinese Nominal Metaphor (NM) generation. We introduce a novel multitask framework, which jointly optimizes three tasks: NM identification, NM component identification, and NM generation. The metaphor identification module is able to perform a self-training procedure, which discovers novel metaphors from a large-scale unlabeled corpus for NM generation. The NM component identification module emphasizes components during training and conditions the generation on these NM components for more coherent results. To train the NM identification and component identification modules, we construct an annotated corpus consisting of 6.3k sentences that contain diverse metaphorical patterns. Automatic metrics show that our method can produce diverse metaphors with good readability, where 92% of them are novel metaphorical comparisons. Human evaluation shows our model significantly outperforms baselines on consistency and creativity.
Keywords
Full Content
Proceedings of the 29th International Conference on Computational Linguistics, pages 6468–6479 October 12–17, 2022. 6468 CM-Gen: A Neural Framework for Chinese Metaphor Generation with Explicit Context Modelling Yucheng Li 1, Chenghua Lin 2∗, Frank Guerin 1 1 Department of Computer Science, University of Surrey, UK {yucheng.li,f.guerin}@surrey.ac.uk 2 Department of Computer Science, University of Sheffield, UK c.lin@sheffild.ac.uk Abstract Nominal metaphors are frequently used in hu- man language and have been shown to be ef- fective in persuading, expressing emotion, and stimulating interest. This paper tackles the problem of Chinese Nominal Metaphor (NM) generation. We introduce a novel neural frame- work, which jointly optimizes three tasks: NM identification, NM component identification, and NM generation. The metaphor identifica- tion module is able to perform a self-training procedure, which discovers novel metaphors from a large-scale unlabeled corpus for NM generation. The NM component identification module emphasizes components during train- ing and conditions the generation on these NM components for more coherent results. To train the NM identification and component identi- fication modules, we construct an annotated corpus consisting of 6.3k sentences that con- tain diverse metaphorical patterns. Automatic metrics show that our method can produce di- verse metaphors with good readability, where 92% of them are novel metaphorical compar- isons. Human evaluation shows our model sig- nificantly outperforms baselines on consistency and creativity. 1 Introduction Metaphors are commonly used in human language. Usually, metaphors compare two different kinds of objects or concepts with the intent to make the expression more vivid, or to make unfamiliar things easier to understand (Paul, 1970). Accord- ing to contrastive studies of English and Chinese, metaphors are especially crucial in Chinese as there are fewer abstract words in Chinese, so that peo- ple tend to express abstract meaning via metaphors (Lian, 1994). In this paper, we focus on the generation task of a special type of Chinese metaphor – Nomi- nal Metaphors (NMs). NMs (比喻 in Chinese) ∗Corresponding author 1. 这个[孩子]tenor 壮的像[牛]vehicle This [boy]tenor is as strong as a [bull]vehicle. Nominal 2. [生活]tenor好比[旅行]vehicle, 没有计划就难以前行 [Life]tenor is a [journey]vehicle, we cannot move on without a plan. Nominal 3. Meta股价[跳水]metaphorical META stock price [dives]metaphorical. Verbal 4. 他可以像大厨一样烹饪 He can cook like a pro. Literal Table 1: Examples of Chinese nominal metaphor, verbal metaphor, and NM components. Note that when the words “like” or “as” are used as COMPARATORS, we also call these special NMs 明喻 (Similes). are figures of speech associating a noun with an- other noun through a COMPARATOR such as like, be, become in English and 像,是,变成 in Chi- nese. Examples and NM components are shown in Table 1. In addition to the COMPARATOR (bold), there are three other components in a nom- inal metaphor: TENOR, VEHICLE, and CONTEXT (text with underline). TENOR is the subject of the metaphor, and VEHICLE is the source of the im- agery (i.e., the object of metaphor). CONTEXT is used to explain the comparison and is crucial for understanding the comparison (more details about NM and NM components are given in § 2.1). …
Content truncated to 500 words
Labeled Text
AbstractNominal metaphors are frequently used in human language and have been shown to be effective in persuading, expressing emotion, and stimulating interest. This paper tackles the problem of Chinese Nominal Metaphor (NM) generation. We introduce a novel multitask framework, which jointly optimizes three tasks: NM identification, NM component identification, and NM generation. The metaphor identification module is able to perform a self-training procedure, which discovers novel metaphors from a large-scale unlabeled corpus for NM generation. The NM component identification module emphasizes components during training and conditions the generation on these NM components for more coherent results. To train the NM identification and component identification modules, we construct an annotated corpus consisting of 6.3k sentences that contain diverse metaphorical patterns. Automatic metrics show that our method can produce diverse metaphors with good readability, where 92% of them are novel metaphorical comparisons. Human evaluation shows our model significantly outperforms baselines on consistency and creativity.
Publication Details
- Paper ID:
- 4668
- Year:
- 2022
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan