The Japanese government's intellectual property export strategy, known as the "Cool Japan" strategy, proposed its latest amendment in 2024. It identified translation as the primary obstacle in developing overseas markets for Japan's ACGN (Anime, Comic, Game, Light Novel) industry. To address this challenge, industry players are actively investing in AI translation technology, sparking controversy within the translation industry.
Linguistic and cultural challenges in ACGN translation
Japanese ACGN translation faces several difficulties, primarily due to linguistic factors. As an isolated language influenced by various foreign tongues, Japanese differs significantly from other languages worldwide. Beyond the official language used in government documents and national education, numerous commonly used dialects exist.
ACGN often incorporates dialect references or wordplay based on puns, creating a cultural gap that makes it challenging for AI to accurately interpret these nuances. Industry practices also contribute to translation difficulties. Mainstream commercial ACGN typically operates on a weekly update system.
According to a Nikkei report, Shogakukan, a major Japanese ACGN publisher, takes an average of seven days to translate manga manually. Using AI for initial translations with subsequent manual revision can reduce this process to three days. Purely AI-driven translation without proofreading could potentially make other language versions available simultaneously with the Japanese release.
To combat rampant piracy, time is of the essence
Timely translations are crucial in curbing overseas piracy, which is rampant in the Japanese ACGN industry. The absence of localized versions or significant delays between Japanese and localized releases leads to unauthorized translation activities by local Japanese-speaking fans and even profiteering by illegal entities, infringing on Japanese publishers' copyrights. Rapid official releases in multiple languages could significantly reduce related losses.
Nikkei reports that the export value of Japan's cultural content, centered around ACGN, is approximately JPY4.8 trillion (about US$30 billion). A survey by Japan's Content Overseas Distribution Association (CODA) estimated that losses from overseas copyright infringement of Japanese content ranged from JPY395.2 billion to 831.1 billion in the same year, roughly 8% to 17% of the export value.
Al translation advancement and challenges
In response to these challenges, Japanese AI translation startup Mantra, founded in 2020, has received investments from major publishers Shogakukan, Shueisha, and Kadokawa. By incorporating LLMs to assist in analysis, the company has reduced the error rate for Japanese to English translation to 1.6%. Other companies, such as internet firm CyberAgent and film production company Orange, are also investing in AI translation technology.
Despite the low 1.6% error rate, the Japan Association of Translators (JAT), with nearly 800 members, argues that AI still lacks the ability to process linguistic subtleties and consider different cultural contexts. A TV Tokyo report indicates that the ACGN industry acknowledges this limitation, which is why human translation remains valued, with AI translation primarily used for online weekly manga requiring timely release.
Finding a balance between AI speed and human translation accuracy remains an unresolved challenge. According to Nikkei, Japan's cultural content export value in 2012, before the establishment of the Cool Japan Fund, was about JPY1.4 trillion. The industry has since maintained an average annual growth rate of 13%. The revised Cool Japan strategy in 2024 aims to sustain this growth rate and achieve over JPY20 trillion in cultural content exports centered around ACGN by 2033.
