Daejeon, July 27–31
This study investigates the interpretation and distribution of Korean aspectual forms in English–Korean literary translation, focusing on the contrast between -ko iss- and -e iss-. Korean aspectual forms encode different construals of situations: -ko iss- typically presents an event as ongoing, whereas -e iss- represents a resultant state.
In translation from English, where aspectual distinctions are often less explicit, translators must select a specific aspectual construal, making interpretive choices explicit in the target text (Kim & Ahn 2018). Previous studies have mainly described these forms in grammatical terms, but less attention has been paid to their discourse-level functions. This study therefore examines how Korean aspectual forms contribute to discourse organization and meaning construction, focusing on -e iss- and its contrast with -ko iss-.
This study adopts a discourse-semantic approach to aspect, integrating event structure and information structure. Korean aspectual marker, -e iss- is analyzed as selecting the result-state subevent from a complex event consisting of a process and its resulting state (Muk 2015). The preceding event is backgrounded, while the resultant state becomes the main descriptive content.
The study further adopts a discourse-semantic perspective (Venhuizen et al. 2018), in which linguistic expressions organize meaning into foregrounded and backgrounded information. Within this framework, -e iss- foregrounds the resultant state while suppressing the causal chain leading to it. This is consistent with studies showing that aspectual forms foreground particular phases of an event structure (Kinsui 2000).
Accordingly, the contrast between -ko iss- and -e iss- reflects a difference in viewpoint placement: the former foregrounds the ongoing process, while the latter foregrounds the resultant state.
The corpus consists of Min Jin Lee’s Pachinko and its Korean translation. The texts were segmented into sentences and processed with MeCab-ko for Korean and spaCy for English. Two aspectual patterns were extracted from the Korean data: -ko iss- and -e iss-. The extraction produced 95 examples of -e iss- and 345 examples of -ko iss-.
For qualitative analysis, each Korean example was manually aligned with its English source sentence. Cases were examined in relation to the local sentence, surrounding discourse, and differences between the English and Korean expressions. GPT-5.4 was used as an auxiliary tool to externalize interpretive contrasts, not to determine the analysis independently. Each prompt included the English sentence, the Korean translation, and four preceding and one following sentence as discourse context. The model was asked to generate alternative Korean translations, compare event/state construals, and explain their discourse effects. By minimizing lexical and stylistic changes, the alternatives served as a controlled basis for examining the contribution of the published aspectual form. Outputs were manually reviewed, and the final analysis was based on researcher validation of recurrent patterns.
1. Convergence and disambiguation: -e iss- resolves aspectual underspecification
The first finding is that structurally diverse English expressions systematically converge on -e iss- in Korean translation:
(1) KO: 양진은 여자들이 자는 곁방의 온돌바닥 따뜻한 아랫목에 책상 다리를 하고 앉아 있었다. (p.38, 39)
EN: Yangjin was seated cross-legged nearby one of the hot spots on the heated floor of the alcove room where the women slept. (p.20)
(2) KO: 양진은 입을 살짝 벌린 채 가만히 앉아 있었다. (p.152)
EN: Yangjin sat still with her mouth opened slightly. (p.103)
(3) KO: 경희는 평소에 수레 옆에 걸어놓는 뭉툭한 시장 의자에 앉아 있었다. (p.268)
EN: She was sitting on the stubby market stool that normally hung on the side of the cart. (p.186)
English forms such as the simple past, progressive, and passive often permit both eventive and stative interpretations. Korean resolves this ambiguity through -e iss-, which enforces a stative construal. Eventive alternatives such as 앉았다would foreground an action not intended in context. Thus, -e iss- eliminates eventive readings that remain implicit in English.
2. Stabilizing Scenes vs. Advancing Events: The Role of -e iss- and -ko iss- in Discourse
The second finding concerns the discourse-level contrast between -e iss- and -ko iss-. -E iss- foregrounds the result-state subevent while backgrounding the preceding process, yielding a stable post-event configuration. By contrast, -ko iss- foregrounds the ongoing phase of an event and situates the viewpoint within its unfolding, thereby contributing to narrative progression.
Taken together, this contrast shows that the two forms produce systematically different discourse effects. While -e iss- organizes discourse around a stabilized configuration anchored in a post-event perspective, -ko iss- organizes discourse around the unfolding structure of an event by situating the viewpoint internally. As a result, the distinction between the two forms is not merely aspectual, but reflects a deeper opposition between spatially structured representation and temporally structured progression.
3. Externalizing interpretive space through LLMs and its limitations
The third finding concerns the methodological contribution and limitations of the LLM-based approach. By generating alternative translations and explanations, the LLM externalizes interpretive possibilities that would otherwise remain implicit. This enables systematic comparison of available interpretations and clarifies why a particular aspectual form is preferred in context.
Aspectual interpretation can therefore be examined within a quasi-experimental framework, demonstrating the potential of LLMs as exploratory tools in Digital Humanities research.
At the same time, outputs are sensitive to prompting conditions and may lack consistency. Because the model relies on distributional patterns rather than explicit linguistic theory, its explanations may not always be theoretically reliable.
These observations suggest that LLMs should be treated as tools for hypothesis generation rather than autonomous analytical agents. Final interpretation must remain the responsibility of the researcher.
This study has shown that the Korean aspectual constructions -e iss- and -ko iss- function not simply as aspectual markers, but as a system for organizing discourse-level viewpoint. By requiring a choice between viewing a situation as a completed state or as an ongoing event, these forms allow translators to resolve ambiguity in the English source.
Methodologically, the study demonstrates that LLMs can externalize interpretive variation and support systematic analysis. Rather than replacing human analysis, LLMs function as tools for exploring interpretive structure and bridging qualitative linguistic inquiry with computational approaches to meaning.
This study was supported by NIHU Digital Humanities Projects and JST, the Support for Pioneering Research Initiated by the Next Generation, Grant Number JPMJSP2185.