When software structure meets translation: material mediation in English-Chinese game localisation

When you play a video game, the words you read on the screen are not just static text pasted onto a picture. They are living parts of a machine. Behind every sentence, from a simple menu label to a complex story description, sits a hidden layer of software instructions. These invisible codes tell the computer when to insert a changing number, how to color a word, or where to wait for a player to press a key. This invisible layer is the bridge between the game's logic and the language the player understands. For decades, researchers have studied how translators adapt stories and cultural references for new audiences, but they have rarely looked closely at how these hidden software instructions survive the journey from one language to another. The question is not just whether the meaning changes, but whether the machine's ability to function remains intact when the words around it are completely rewritten.

A researcher set out to map this invisible bridge by examining the translation files of three popular open-source strategy games: OpenTTD, The Battle for Wesnoth, and 0 A.D. They gathered a massive collection of over 60,000 pairs of English and Chinese text strings. In this study, the English text served as the original source, and the Chinese text as the target, allowing the team to see exactly how the hidden instructions were handled in the translation process. The researcher was looking for a specific phenomenon they call material mediation. This is a way of describing how a software file acts as a mediator, holding the connection between a game's function and the words used to describe it. They wanted to know if the software structure could remain continuous and stable even when the visible language changed completely.

To do this, the researcher treated the translation files like a scientific specimen. They stripped away the visible words to look at the technical markers—the code-like instructions embedded in the text. These markers might be commands to insert a comma, a variable that pulls in a number like a score or a distance, or a selector that changes a word from singular to plural. The researcher compared the English instructions against the Chinese ones to see which ones were kept, which were changed, and which were left out entirely. They also built a system to classify what each piece of text actually does in the game, such as labeling a button, giving a tutorial instruction, or naming a historical figure. To ensure their findings were accurate, two independent experts reviewed a sample of the data, and the researcher even tested specific cases by running the actual game to see if the translations worked correctly in real time.

The results revealed a surprising level of stability in the software's skeleton. Out of nearly 8,000 text strings that contained these hidden instructions, more than 90 percent retained the exact same sequence of commands or the same functional pattern in the Chinese version. This means that even though the words changed, the software structure maintained a high degree of ordered correspondence in the localisation files. For example, a command to insert a number for a distance in kilometers might look different in the Chinese file, but the instruction to insert that number and the instruction to add a space after it were preserved exactly. The software structure acted like a sturdy frame that held the new language in place, though the study notes that this statistical retention does not by itself establish semantic equivalence, translation quality, or runtime correctness.

However, the study also found that the visible language often had to reorganize itself to fit the game's needs. In English, a sentence might use a specific grammatical marker to indicate that a noun is plural, like adding an "s" to the end of a word. The software file would contain a hidden instruction to select that plural form. In Chinese, nouns do not change for number in the same way; instead, the language uses a different system involving numbers and classifiers. The researcher found that in these cases, the hidden instruction to select a plural form was often omitted in the Chinese file. This was not a mistake or a missing piece. Instead, the Chinese translation reorganized the sentence around a different grammatical structure that did not need that specific instruction. The software function remained the same—the game still knew how many passengers were on a train—but the way the file expressed that fact changed to match the rules of the Chinese language.

The researcher also looked at how interface instructions were rewritten. In English, a button might say "Open list," using a command verb. In the Chinese translation, the same button might say "Display list," focusing on the result of the action rather than the action itself. The hidden software code that linked the button to the list remained unchanged, but the human-readable words shifted their focus. In some cases, a single English verb was split across several coordinated items in Chinese, or a complex instruction was condensed into a short phrase. The study showed that translators and the software system work together to keep the game playable. The software provides the fixed points—the values, the colors, the keys—while the human language flows around them, adapting to the grammar and style of the target culture.

One of the most important findings was that you cannot judge a translation's quality just by looking at the words or by counting how many characters are in a sentence. The researcher found that a short Chinese label might be a direct equivalent to a long English one, or it might be a completely different kind of instruction that just happens to be shorter. Without checking the hidden instructions and testing the game, it is impossible to know if a short label is a clever simplification or a broken function. The study supports the use of localisation files as an evidence source to show how technical continuity can coexist with grammatical and functional reformulation, but it does not by itself establish semantic equivalence, translation quality, or runtime correctness. To understand game translation, you must look at the file as a whole, seeing how the technical markers and the human words work together. The software structure provides a continuous thread that holds the game together, even as the language changes from one side of the world to the other.

This work matters because it changes how we think about translation in the digital age. It shows that translation is not just about swapping words; it is about navigating a complex system where language and code are deeply intertwined. The researcher demonstrated that the software file itself is a crucial part of the translation process, acting as a stable foundation that allows language to change without breaking the game, based on an analysis of public text. By separating the technical instructions from the visible words, they were able to show that high levels of stability exist in the machine layer, even when the human layer is completely reimagined. This understanding helps developers and translators create better games for global audiences, ensuring that the magic of the game works just as well in Chinese as it does in English, not by copying the words, but by respecting the structure that holds them together.