Made in China versus Made in Spain. A corpus-based study comparing AD in Chinese and Spanish

Yuchen Liu, Helena Casas-Tost, Sara Rovira-EstevaThe Journal of Specialised Translation 2022 · DOI

TL;DR

This study explores whether translating Audio Description (AD) scripts from Spanish into Chinese is a viable shortcut to address China’s shortage of AD professionals. By comparing a bilingual corpus of eight films, the researchers found that while core content selection is remarkably similar, Chinese AD is significantly more "interventionist"—offering higher information density, more subjective interpretation, and explicit narrative guidance compared to the more neutral Spanish approach.

What Problem Does This Paper Address?

Audio Description (AD) is the "voice-over" service that describes visual elements of media for people with sight loss. In Western countries like Spain, AD is a professionalized field governed by strict standards of neutrality. In China, however, AD is still in its infancy, largely dependent on volunteers and facing a severe scarcity of trained scriptwriters.

The authors investigate a potential solution: AD translation. Instead of writing scripts from scratch (ex novo), could China import and translate scripts from "pioneer" countries? To answer this, they needed to determine if the "what" and "how" of audio description are consistent enough across these two vastly different languages and cultures to make translation feasible.

Methods

The researchers employed a corpus-based contrastive analysis using a multimodal approach.

  1. Corpus Creation: They compiled a bilingual corpus of eight films that had existing AD in both Spanish and Chinese.
  2. Annotation Scheme: They developed a custom labelling scheme with 31 distinct labels across four blocks: verbal-visual elements (text on screen), characters, actions, and settings.
  3. Software & Analysis: Using ATLAS.ti, they manually tagged the scripts while synced to the video. They then performed quantitative statistical analysis (Mann-Whitney tests) to find significant differences and qualitative "deep dives" into specific scenes to understand the why behind the data.

Datasets

The study utilized a newly compiled bilingual multimodal corpus consisting of eight films:

  • Cinema Paradiso, Rain Man, Jurassic Park, Eat Drink Man Woman, The Chorus, The Curious Case of Benjamin Button, Frozen, and The Imitation Game.
  • Spanish Source: All Spanish scripts were produced by ONCE (the Spanish National Organization of the Blind) using the Audesc system.
  • Chinese Source: Scripts came from various volunteer groups like the Shanghai Film Critics Society and the Yangguang Volunteer Association.

Key Insights and Ideas

  • The "Paternalistic" vs. "Neutral" Divide: The most striking finding is the difference in authorial voice. Spanish AD follows the AENOR standards, which emphasize neutrality—describing only what is seen. Chinese AD scriptwriters often adopt a "paternalistic" tone, providing "explicitness" (explaining cause-effect) and "subjective comments" (interpreting metaphors or emotions) to ensure the audience "properly" understands the film’s art or moral lessons.
  • Information Density: Chinese AD carries a significantly higher cognitive load. On average, Chinese descriptions used more words per "quotation" (coded segment) than Spanish ones.
  • Cultural Distance and Naming: When a cultural reference is familiar (e.g., a monument in Rome for a Spanish audience), it is named directly. When it is "distant" (the same monument for a Chinese audience), it is often reduced to a generic description (e.g., "a street").
  • Linguistic Tools (Chengyu): Chinese scriptwriters frequently use chengyu (four-character idiomatic expressions). These are highly efficient, allowing them to convey complex emotional or visual states (like "fascinated as if drunk") very concisely, which is a unique advantage for the time-constrained nature of AD.

Key Results

  • Content Similarity: The "what" of AD (who, where, when, what) is highly consistent across both languages, supporting the idea that translation is a viable starting point.
  • Statistical Divergence: Significant differences (p < 0.05) were found in eight specific areas, including psychology (Chinese AD describes internal states much more), explicitness, and subjective comments.
  • Character Fixation: Chinese AD tends to name characters immediately upon their first appearance, whereas Spanish AD often waits until the character is named within the film's dialogue to avoid "spoiling" the narrative.
  • Quantitative Load: In seven out of eight films, the Chinese AD was longer in word count than the Spanish version, despite the films having the same duration.

Limitations and Future Work

The authors acknowledge several hurdles:

  • Corpus Size: Eight films is a relatively small sample size for broad generalizations.
  • Historical Context: The AD scripts used were not all recent; practices in both countries may have evolved since these scripts were written.
  • Directionality: None of the films in the study were originally Spanish, which might affect how the Spanish AD was constructed.

Future Directions: The researchers propose a reception study to see if Chinese users actually prefer the high-information, interpretive style or if they would benefit from the more streamlined Western approach. They also suggest that "localization guidelines" are essential—translation alone isn't enough; the scripts must be adapted to fit the cultural expectations of the target audience.

Summary updated

Abstract

This article presents a corpus-based study comparing AD in Chinese and Spanish, as a first step towards testing the feasibility of AD translation from Spanish into Chinese. A bilingual multimodal corpus consisting of eight films with AD in both languages was first compiled, and then a labelling scheme was designed for the corpus annotation with ATLAS.ti. The results from a quantitative and qualitative analysis of this corpus show that there are many similarities in content selection between Chinese and Spanish AD, which justifies our proposal of translating AD scripts from Spanish into Chinese to increase AD in China. However, we also found that Chinese AD carries a higher information load than Spanish AD. There are also differences in the general approach to AD, in the sense that Chinese scriptwriters tend to intervene or interpret more than their Spanish counterparts. Therefore, such differences also suggest the need for a certain degree of localisation if AD is translated from Spanish into Chinese instead of created ex novo.
BibTeX
@article{liu2022made,
  author = {Yuchen Liu and Helena Casas-Tost and Sara Rovira-Esteva},
  title = {{Made in China versus Made in Spain. A corpus-based study comparing AD in Chinese and Spanish}},
  journal = {The Journal of Specialised Translation},
  year = {2022},
  doi = {10.26034/cm.jostrans.2022.088},
}
Figure