With the rapid advancement of digital communication, graphic new media platforms have evolved into core carriers for the overseas dissemination of fine traditional Chinese culture. Platforms including WeChat Official Accounts and Weibo feature low operating costs, strong interactivity, rapid information transmission and long-term communication effects, which have made them mainstream promotional channels in the new media landscape ([3]). Among these platforms, bilingual WeChat Official Accounts stand out with systematic content, regular updates, standardized column settings and persistent communication effects. They break the limitations of fragmented, ephemeral short-video communication and serve as a vital medium for the long-term overseas promotion of TCM culture. Integrating text and images, bilingual TCM WeChat Official Accounts continuously release systematic content such as popular science on TCM theories, intangible cultural heritage herbal processing techniques and seasonal healthcare regimens, offering overseas audiences a critical window to understand traditional Chinese concepts of health.

However, TCM culture boasts highly composite attributes and contains numerous cultural-loaded concepts with no direct equivalents in Western languages, which poses tremendous challenges to cross-linguistic translation. Current translation practices for bilingual TCM WeChat Official Accounts fall into two extreme modes. Most official accounts adopt generative AI machine translation to translate texts rapidly. Though productive, such translations frequently suffer from inconsistent terminology, distorted cultural connotations and prohibited medical promotional claims, which severely damages China’s international academic reputation. A small number of high-quality accounts adopt full human translation and proofreading. While their translations are polished, this mode incurs excessive labor costs and lengthy delivery cycles, failing to meet the demand for frequent, routine content updates of official accounts.

Given the above operational dilemmas, the inherent drawbacks of single-mode translation cannot be resolved. The translation industry urgently requires an innovative translation system that balances efficiency, textual quality and compliance. Against this background, this paper delineates the applicable boundaries between Large Language Models (LLMs) and human translators in processing TCM texts, and explores the implementable pathways of Human-AI Collaborative Translation (HACT) for bilingual TCM WeChat Official Accounts, aiming to provide operable theoretical support for the digitalized, large-scale global outreach of TCM culture.

The TCM knowledge system is a unique theoretical medical framework rooted in thousands of years of Chinese civilization, and an integral component of fine traditional Chinese culture. Distinct from Western modern medicine founded on anatomical research, TCM integrates philosophical speculation, folk healthcare and medical ethics. Such multi-layered cultural attributes directly complicate its translation for international audiences.

First, philosophical concepts in TCM feature linguistic incommensurability. Core TCM doctrines originate from the theories of Yin-Yang and Five Elements as well as the harmony between humanity and nature, highlighting the dynamic balance between human bodies and the natural world alongside treatment based on syndrome differentiation ([1]). Core terms such as qi, blood, wind and dampness lack fully corresponding signifiers in Western linguistic systems. Literal translation easily distorts overseas audiences’ cognition. For instance, the TCM pathogenic factor “wind pathogen” may be misinterpreted as natural wind in the physical sense.

Second, TCM contains diversified clinical expressions and intervention modalities. TCM covers multiple therapies including herbal medicine, acupuncture, tuina massage and qigong (life-energy cultivation practice). Its descriptions of disease mechanisms and therapeutic principles (e.g., soothing the liver and regulating qi, activating blood circulation to dissipate blood stasis) are highly condensed. These terms function both as medical jargon and heavily context-dependent culture-loaded words, requiring translators to possess both medical expertise and cross-cultural interpretative capabilities.

Third, Western audiences hold divergent perceptions of TCM wellness philosophies. TCM closely connects daily diet with disease prevention and treatment, embodied in concepts such as medicinal and edible homology and seasonal healthcare. The international communication of TCM culture is not merely one-way export of medical knowledge, but a cross-cultural dialogue on lifestyles. Translators must overcome pragmatic barriers to convert this unique wellness philosophy into universally accessible shared health wisdom.

The integration of traditional TCM culture with new media represented by WeChat Official Accounts effectively reduces reading obstacles brought by obscure ancient medical texts. The TCM industry bears strong professional specificity. Translators engaged in TCM external publicity translation must master three core capabilities: linguistic and cultural conversion, medical professional identification and foreign language expression. They also need systematic command of TCM cultural connotations, basic TCM theories and clinical practices. Centered on TCM cultural communication, translation work requires collaboration with industry specialists and academic scholars to accurately define the professional characteristics and communication positioning of TCM English translation ([5]). As long-form graphic contents, bilingual TCM WeChat Official Accounts exhibit distinctive traits that differentiate them from other external publicity texts:

Human-AI Collaborative Translation (HACT) refers to an innovative cooperative mode where human translators and intelligent systems such as LLMs interact dynamically to realize in-depth collaboration in translation generation, optimization and decision-making. “Its core feature lies in shifting optimization from single translation segments to full-process intelligent assistance, redefining translators’ roles from mechanical editors to senior decision-makers and quality controllers” ([4]).

The HACT model demonstrates high theoretical compatibility for TCM official account external translation. On the one hand, it balances translation productivity and textual quality. Leveraging massive corpora, AI rapidly completes basic semantic conversion and processes large volumes of repetitive basic popular science content. On the other hand, it compensates for machines’ deficiencies in cross-cultural pragmatic competence and compliance awareness. Human translators can concentrate their energy on interpreting culture-loaded terms, restructuring logic in Chinese-style long sentences, and eliminating potentially illegal medical promotional language.

The open-source general large language model DeepSeek LLM is adopted to generate initial AI translations in this paper. Before translation, the WHO International Standard Bilingual Glossary of Traditional Chinese Medicine is imported as a fixed constraint glossary. The model is forced to adopt standardized translations from the glossary rather than literal translations during output.

Unified basic prompt template:

Task: English translation of popular science texts for bilingual TCM WeChat Official Accounts

Constraints:

1) Strictly follow standardized terms from the WHO TCM glossary;

2) Avoid absolute medical claims such as “cure”, “radical treatment” and “significant curative effect”;

3) Adopt formal English popular science writing style and revise Chinese-style run-on sentences;

4) Add standardized explanations for unique TCM pathogenesis and health preservation concepts.

Targeting pain points in official account operation, this paper constructs a four-dimensional closed-loop workflow: Corpus Co-construction—Intelligent Initial Translation—Collaborative Restructuring—Expert Review & Validation.

1) Pre-translation Stage: Professional translators, together with licensed TCM practitioners, sort out high-frequency terms from WeChat Official Account articles, supplement them to the WHO terminology database, and generate a dedicated bilingual glossary for import into the model.

2) While-translation Stage: Input the full text of each official account article and load the fixed prompt template and terminology database. Intelligent Initial Translation + Collaborative Restructuring LLMs are deployed to generate batch translation drafts; Translators revise the translations against the four evaluation criteria. Human translators conduct hierarchical editing to resolve logical discontinuities, literal conceptual translation and cross-cultural cognitive gaps.

3) Post-translation Stage: Expert Review & Validation TCM professionals conduct final audits focusing on herbal compatibility and pathological mechanisms. Finalized texts are fed back into the database to enable continuous fine-tuning of the AI translation model, forming closed-loop feedback circulation.

Typical corpus extracted from bilingual TCM culture WeChat Official Accounts is analyzed below across lexical, syntactic and pragmatic dimensions to elaborate the practical pathways of HACT strategies.

Corpus Sources, Screening Criteria and Sampling RulesSources of WeChat Official Accounts

Three official bilingual TCM WeChat Official Accounts in China that have run for more than three years and target overseas audiences are selected, namely China Traditional Chinese Medicine, Huayi Network and Journal of Traditional Chinese Medicine.

Screening Criteria

Text type: Popular science articles about solar term health preservation, medicinal and edible homology, and basic TCM theories;

Excluded contents: Clinical diagnosis and treatment, prescription drug advertisements, and pure activity notices;

Complete Chinese-English parallel texts without obvious literal machine translation traces, with a word count ranging from 300 to 800 words.

Sample Scale and Control Group Setup

A total of 15 articles are obtained through stratified random sampling as research corpus. The control group consists of pure machine translations generated solely by DeepSeek LLM for the same texts, while the experimental group refers to HACT collaborative translations revised by human translators.

Unified Four-Dimensional Evaluation Criteria (Applied Consistently to All Cases in This Paper)

Terminology Accuracy: Compliance with WHO international standard TCM terminology without literal mistranslation or conceptual confusion;

Logical Coherence: No comma splices or run-on sentences in English; clear causal and coordinate logic, and consistent textual references;

Cultural Adaptability: Supplementary cultural explanations for unique TCM concepts to eliminate comprehension barriers for overseas readers;

Compliance: Avoidance of absolute curative-effect expressions to meet the regulatory requirements of overseas medical publicity.