American Lawyers and Japanese School Students:
A comparative analysis of the localizations
of Ace Attorney and Persona
Maloe de Goeij
A master thesis submitted in partial fulfilment for the
degree of Master of Arts
in
Translation Studies
University of Amsterdam
Supervisor: Dhr. Dr. E.R.G. Metz
2
T
ABLE OF
C
ONTENTS
1 Introduction ... 3 2 Theoretical Framework ... 6 2.1 Translation Theories ... 6 2.1.1 Frame Semantics ... 9 2.1.2 Japanese-English translation ... 12 2.2 Video games ... 142.2.1 Video Game Localization and transcreation ... 15
2.2.2 The influence of the internet ... 16
3 Methodology ... 19
3.1 Data scope and acquisition method ... 19
3.2 Data analysis method ... 20
4 Case study ... 21
4.1 The games and game series: Ace Attorney and Persona ... 21
4.1.1 The story ... 21
4.1.2 The gameplay ... 22
4.1.3 The localizations ... 23
4.2 Presentation of the data ... 23
4.3 Analysis of the data ... 26
4.3.1 “False” frame match ... 27
4.3.2 “False” frame mismatch ... 30
4.3.3 Frame mismatch: option or servitude?... 33
4.3.4 Frames differ, slightly to greatly ... 37
4.3.5 Ambiguous cases ... 42
4.3.6 Puns and jokes ... 43
4.3.7 Names ... 49
4.3.8 Japanese language specific expressions and words ... 52
4.3.9 Translation mistakes ... 59
5 Conclusion ... 61
References ... 65
Appendix ... 67
Glossing abbreviations ... 67
Frame created for this research... 67
Research data from Phoenix Wright: Ace Attorney ... 68
3
1 I
NTRODUCTION
To the uninitiated, video games may seem like a trivial form of pastime or a waste of time. Since their emergence and rise to popularity, people have even wondered whether (some types of) video games could be detrimental to our health. However, no matter what your opinion on video games is, it is undeniable that – especially in the wealthier countries of the world – video games are more relevant and present in our lives than ever before. Whether it is in the form of the millions of apps on smartphones, the newest developments in virtual reality gaming, or the rising popularity of
edutainment (education in the form of entertainment), fact is that the video game industry is
thriving.
However, it is not just these new areas of the video game industry that have enjoyed success in recent years. Now markets have started to recover from the 2008 global economic crisis, both home entertainment and handheld video game consoles are back on the rise and captivating audiences all over the world. As a result of this, when video game giant Nintendo released their latest console, the Nintendo Switch (which is actually a hybrid between a home and a handheld console) in March 2017, the company could not keep up with the demand for the system. For months stores around the world dealt with stock shortages because of the unexpected popularity of the Nintendo Switch (Martin 2017). Despite this, the company managed to sell an impressive 7.63 million units of the console worldwide by September 2017 (“CONSOLIDATED FINANCIAL
HIGHLIGHTS” 2017, 8). Considering this, it is not surprising that the academic world has finally started to notice that there is a significant lack of research on most aspects of video games (Newman 2008, 8).
Video games are no longer a simple, one-dimensional phenomenon. Unlike the first video games which usually consisted of a limited menu interface and a 2D image relaying the often simple and repetitive gameplay to the player, modern video games exist of an array of different assets, varying from life-like 3D models to lengthy clips of spoken dialogue and vast world maps the player can explore. These physical assets can oftentimes be experienced with multiple players, both live as well as online, in co-op or versus format. This multitude of facets and aspects make modern video games a relevant subject for a variety of different types of research.
The majority of pre-2000 research on video games focuses on the (mental) health risks they could potentially pose to players (Ricci et al. 1998; Funk 1992; Scott 1995). However, since the start of the new millennium, scholars have started to recognize video games as a multi-faceted medium that can be studied from various points of view and is relevant to multiple fields of study. Some of these include sociology (Bryant and Vorderer 2006; Gabbiadini and Greitemeyer 2017), educational
4 studies (Khine 2011; Ru-Lan Hsieh, Wen-Chung Lee, and Jui-Hsiang Lin 2016; Gee 2003) and – as I will be doing in this thesis – translation studies.
Translation studies itself is a relatively young field of study which has mainly concerned itself with literary translation between Indo-European languages, and only recently started to adopt multi-disciplinary studies and methods (Costales 2012, 391). One of these new fields of study focuses on the translation of video games, or as it is called within the video game industry, video game localization.
Video game localization is a complex process (which I will elaborate on in chapter 2) and it poses a wide variety of translation problems to the translator. As Bernal-Merino (as cited in Czech 2013, 5) states in his 2007 article, while video games do not include any text types that are not also found in other media (e.g. computer software, films and series), the challenges involved with translating a video game are indeed very different from other media. Translators have to deal with complex storylines that may be extremely fragmented and presented in a non-linear format, context in the form of pictures and video they may not have access to at the time of translation and making sure not to change too much or too little during translation to avoid backlash from video game fanatics. This last matter of deciding to significantly change an aspect of a source text during translation is something that is quite uniquely prevalent to video game localization, and within the industry it is commonly referred to as transcreation.
While loyalty to the source text – whether this is loyalty to the information presented or loyalty to the style – is usually considered the most important factor for assessing the quality of a translation, within video game localization this is sometimes considered to be less relevant. Instead the localizer is often given the freedom to change, add or remove any number of cultural references, jokes or even entire scenes or levels in a video game in order to give the player the best possible experience (Hevian 2007, 309).
From this, I hypothesize that using this strategy during the localization of a video game has a significant impact on multiple aspects of the final product such as humor, characters’ personalities, the flow of conversations and cultural references. Studying the effects of transcreation in detail and determining what aspects of a target text are most affected by it, may offer valuable information to translators. If a translator has a better insight as to what translation problems transcreation may cause or help to solve, how it might change their translation and whether that is what their audience desires, they will be able to make a more informed choice as to whether they want to employ it as a translation strategy or not.
In my bachelor thesis (de Goeij 2016, 16) on the challenges in video game localization I touched briefly on transcreation as a translation strategy and its significance within video game
5 localization. However, because of its importance I will study it in more detail in this thesis and attempt to determine and analyze in exactly what way the translated text is affected when translators decide to transcreate.
Consequently, the main research question for this research is: “How does transcreation as a translation strategy in video game localization affect the translated linguistic assets?” To be able to give an adequate answer to this broad question, I will first attempt to answer two sub-questions: 1. “How does transcreation as a translation strategy in video game localization affect general text meaning?” 2. “How does transcreation as a translation strategy in video game localization affect specific linguistic target text elements such as cultural references and humor?” Another aspect of transcreation is the adjustment of images and animations, but since this thesis is written from a translation studies perspective, that is beyond the scope of this research.
To find an answer to these research questions I will perform a case study on two Japanese role-playing games (hereafter JRPGs) from two video game series of which one is known for its extremely transcreated English translation, while the other is known these days for being kept as close to the original as possible. These video games are respectively called Phoenix Wright: Ace
Attorney – Dual Destinies and Persona Q: Shadow of the Labyrinth. I will gather data from both
games and compare the original Japanese versions to the localized English versions and then analyze the results to see whether there is any significant difference to the amount or type of changes that occurred during translation.
However, before I do this, I will first illustrate the academic context in which this research takes place by giving an overview of the main fields of study and the most relevant theories that concern this research.
6
2 T
HEORETICAL
F
RAMEWORK
In order to establish which theories and facts form the foundation for this research, this chapter will provide an overview of several fields of study and discuss secondary sources that are of particular relevance. In presenting the secondary sources and theories, I will attempt to highlight their qualities and flaws and in what way they are useful to this research. The two main subjects of this research, video games and translation, will be the main topics at hand in this chapter. During the introduction of these topics, I will also present definitions of the terms used in this thesis.
The first part of this chapter consists of a short overview of the theories within the field of translation studies that are the most easily applicable to the translation of video games. I will then elaborate on the theory that I will use in the methodology of the case study on the two video games. Similarly, I will dedicate a separate section to the few works that have been written on the subject of Japanese-English translation, since most existing translation theories have been written with a focus on Indo-European languages. On the topic of video games this chapter will serve to introduce the concepts the reader needs to be familiar with to have a better understanding of the analysis of the case study data in chapter 4.
2.1 T
RANSLATIONT
HEORIESEver since scholars first started to write theories about translation practice and product and started to debate what the qualities of a ‘good’ translation were, one of the central conversations has been about literal translation as opposed to free translation (Munday 2016, 31). Literal, or ‘word-for-word’ translation refers to the concept of translating a text from one language into the other while transferring the meaning of each individual word directly. Free, or ‘sense-for-sense’ translation refers to the concept of translating the meanings that are conveyed by the words of the source text (hereafter ST), instead of translating the individual words.
This juxtaposition of free versus literal is very relevant to the topic of transcreation, since transcreation claims to be the freest type of translation and even claims to surpass translation altogether. It is also the very core of this research, since I will be studying how ‘free’ the translation of a very transcreated video game is, as opposed to one that was translated in a more literal way. However, when discussing this topic in depth, these two concepts are too vague to be functional in a case study methodology.
One could beg the question “What exactly counts as literal translation?” For example: The word ‘cheese’ seems easy to translate literally, but is the Dutch word ‘kaas’ really the literal
7 translation of the word ‘cheese’, if without any further context or clarification an American person imagines sliced and pre-packaged cheddar and a Dutch person imagines a large block of Gouda cheese, and both of these people think of wildly different connotations to the words? On the other hand, one can ask whether it is truly possible to grasp the true sense or meaning of a ST as a whole or even parts of it. Most of the time there is no telling whether another person would interpret a ST the same way, or whether a translator is interpreting it the way it was intended by the author. If this is not possible, then what ‘sense’ is a translator translating, if not the literal meaning? In short, this means that though it is a useful guideline and base concept, these uncertain factors make the free versus literal debate less useful in a more detailed case study.
A more detailed theory that is often mentioned in research on video game localization (Costales 2012, 392; Mangiron and O’Hagan 2006, 20; Gillberg 2013, 3–4; de Goeij 2016, 6) is
Vermeer’s skopos theory. This theory presumes that every ST has a specific purpose, or ‘skopos’, and that this purpose dictates how the text should be translated. Six rules form the core of this theory: “1. A translational action is determined by its skopos. 2. It is an offer of information in a target culture and target language concerning an offer of information in a source culture and source language. 3. A target text does not initiate an offer of information in a clearly reversible way. 4. A target text must be internally coherent. 5. A target text must be coherent with the source text. 6. The five rules above stand in hierarchical order, with the skopos rule predominating.” (Munday 2016, 127)
Though the skopos theory has been rightfully criticized for not being applicable to all types of text since the skopos of a text is not always as clear as the theory presumes (Munday 2016, 130), it is a theory that is very easily applicable to video game localization. The reason for this is that the main purpose of a video game is simple: from a consumer’s perspective it is to be as entertaining as possible and from a game developer’s perspective it is to sell as much as possible. These two
perspectives are closely connected because if the game is not entertaining, it will most likely not sell well. It is also noticeable that within this theory, coherence with the source text has a very low priority, which is again easily applicable to the practice of video game localization.
However – like I argued in my bachelor thesis – while the skopos theory is useful to describe and make sense of video game localization as a process, it does not give much insight on the more minute details within a ST or target text (hereafter TT). This is why even though I argue that it is an important theory to keep in mind, it is not the right theory to apply when studying translation product on the sentence or word level.
A slightly more practical theory is Toury’s descriptive translation studies method. This method presumes that a TT has a certain place and status within the target culture and that this is
8 the determining factor in what kind of translation strategy is used. To be able to study translation based on this hypothesis, he constructed a three-step process that can be applied to a ST-TT pair and then repeated on other texts to be able to draw significant conclusions about the translation
strategies used in a group of similar texts.
This three-step process is as follows: “1. Situate the text within the target culture system, looking at its significance or acceptability. 2. Undertake a textual analysis of the ST and the TT in order to identify relationships between corresponding segments in the two texts. Toury calls these segments 'coupled pairs'. This leads to the identification of translation shifts, both 'obligatory' and 'non-obligatory'. 3. Attempt generalizations about the patterns identified in the two texts, which helps to reconstruct the process of translation for this ST-TT pair.” (as summarized in Munday 2016, 175)
Though this three-step guide seems very practical, and Toury stresses the importance of using the ‘coupled pairs’ during the analysis, he offers many examples but no clear definition of what these ‘coupled pairs’ are and how exactly one is supposed to compare them (Toury 2012, 115–30). Furthermore, the concept of a ‘translation shift’ is not detailed enough when discussing
transcreation. There is a big difference between changing a ST verb into a noun in the TT and removing an entire sentence and replacing it with something with an entirely different meaning, but these two could both be marked as translation shifts.
One extremely detailed and practical method for studying translation product is Vinay and Darbelnet’s theory consisting of seven identifiable translation procedures belonging to two overall translation strategies. As explained in Munday (2016, 88–92), this method prescribes that the researcher first identifies the ‘units of translation’, which are ST and TT pairs of words or word groups that correspond to one another. Then, each unit of translation is analyzed and assigned one or more of the translation procedures to determine what kind of change (if any) has occurred during translation.
One strategy is direct translation, which is comparable but not identical to the concept of word-for-word translation I introduced in the beginning of this section. Direct translation can be split up into three translation procedures: 1. Borrowing (transferring the source language (hereafter SL) word directly to the TT without changing it) 2. Calque (literally translating different parts of the SL word to create a neologism) 3. Literal translation (translating literally without changing the grammatical, denotative or connotative meaning).
The second strategy, oblique translation, is comparable to sense-for-sense translation and it can be split up into four translation procedures: 1. Transposition (translating a part of speech with a different part of speech) 2. Modulation (changing the semantics and point of view of a word or
9 group of words) 3. Équivalence (replacing an idiom or a saying in the ST with a target language (hereafter TL) idiom or saying with the same meaning) 4. Adaptation (changing cultural items in the ST and replacing them with items from the target culture which might have similar connotations).
A notable addition to this theory is the distinction between what Vinay and Darbelnet call
option and servitude (as cited in Munday 2016, 93). While the translation strategies and procedures
explained above can tell the researcher what changes were made during translation, the concepts option and servitude can help to identify whether these changes were voluntary or not. In other words, it can help answer the “why” question. Servitude occurs when a modulation or transposition was unavoidable because of structural differences in the two languages, while option occurs when changes are made based upon the translator’s own preferences.
Due to the fact that this theory is very detailed and attempts to reveal even the smallest changes that occur during translation, I elected to use it in the methodology of my bachelor thesis. Using Vinay and Darbelnet’s model, I analyzed a set of data from the video game Pokémon Omega
Ruby and its official translation. However, during my analysis I ran into several issues pertaining to
the application of the theory to the data at hand. Roughly 22% of the data I analyzed, resisted categorization according to Vinay and Darbelnet’s model (de Goeij 2016, 22). Of that 22%, a large portion consisted of very ‘free’ translations that deviated so extremely from the ST that none of the categories in Vinay and Darbelnet’s model applied (de Goeij 2016, 25). Additionally, determining when multiple words were one single unit of translation and when they were actually multiple shorter units of translation was a rather subjective process filled with trial and error.
A different problem with the use of Vinay and Darbelnet’s model was that a large portion of the changes I found were likely to just be a result of the differences between the Japanese and English languages (de Goeij 2016, 27). While identifying these differences can be interesting in research on these two languages, it merely skews the data when attempting to determine what kind of changes in translation a specific translation strategy causes. For this reason, I decided against using this theory for the current research and attempted to find a theory, model or method that could (for the most part) eliminate the interference caused by the fact that Japanese and English are two very differently structured languages.
2.1.1 Frame Semantics
While searching for a practical theory or model for the analysis of my case study data that could be applied to both the Japanese-English language pair and video game localization, I encountered an article that analyzed English to Japanese translations using the Berkeley FrameNet project
10 (Hasegawa, Lee-Goldman, and Fillmore 2014). The FrameNet project is a database that is publicly accessible through the URL “https://framenet.icsi.berkeley.edu”, that uses the theory of frame
semantics to categorize the English language and make connections between form and meaning
visible. In their article Hasegawa, Lee-Goldman and Fillmore (2014, 194) concluded that this
database is versatile and abstract enough to be used on multiple languages, including languages with a basic structure that differs from English, like Japanese. Additionally, a version of FrameNet
specifically for Japanese has also been created.
The basic concept of frame semantics was derived from the general concept of frame that is used in cognitive sciences (Fillmore and Baker 2015, 792). In this context ‘frame’ is a word used to express the bundles of concepts humans use to make sense of the world around them. A frame is a set of pieces of knowledge and beliefs that ‘belong together’. They are formed by, and to a certain extent shape our experiences and influence the way we think about and remember things. Most of the time these frames, though present, are so integrated into our daily lives that they are not part of a conscious thought process (Fillmore and Baker 2015, 796). For example, when we think of selling something, there are many concepts we automatically associate with that action: The seller (yourself), the buyer, the object that is sold, the money that is received in return and sometimes even the time or place at which the action is completed. All of these concepts are part of one particular cognitive frame.
Many frames are universal to human beings, simply because we all have similar bodies that are subjected to similar experiences (e.g. pain, warmth, tiredness), live in and travel to similar environments (e.g. houses, towns, forests, etc.) and perform similar interactions with others (e.g. talk, fight, joke, etc.) (Fillmore and Baker 2015, 792). However, there are some frames that are specific to certain (or multiple) cultures and time periods. A person from the 15th century would not
know what a computer is and what concepts are associated with it, while someone from the
Netherlands might not understand the implications of a Japanese person using a face mask (which is usually because they have a cold and do not want to infect others).
While in general ‘frame’ is a cognitive concept and not a specifically linguistic one, in frame semantics (as well as in this thesis) this term is used to describe a cognitive frame that is linked to a lexical unit (hereafter LU). That is, frame semantics (and with it FrameNet) is based on the principle that words, when placed together with other words and processed by a competent speaker of the language, bring to mind, or in other words, evoke certain frames (Hasegawa, Lee-Goldman, and Fillmore 2014, 171). It is this focus on context and meaning (instead of pure linguistics or grammar), while still enabling the researcher to look at translation on the word and sentence level, that makes this theory uniquely suited to this research.
11 To make frame semantics a more practical and useful theory, a team of researchers at the University of Berkeley started a project they called FrameNet. As stated above, FrameNet is an online database that is dedicated to defining and mapping frames and showing the links between them. In the database’s frame index, each frame is first given a name. Then, in the frame’s index entry, a definition of the frame is added and a set of frame elements (hereafter FEs) is assigned to it. These FEs are the different concepts that belong to the frame and they consist of core FEs (those elements that are center to the frame and are usually present when referring to the frame) and non-core FEs (elements that can belong to the frame but can also easily be left out). Finally, a set of frame relations is added, in which the frame is connected to other frames in terms of perspective, degree of specificity and causative relations, among others.
For example, the frame Committing_crime is defined as: “A Perpetrator (generally intentionally) commits a Crime, i.e. does something not permitted by the laws of society.” This frame’s core FEs are Perpetrator and Crime, and its non-core FEs include Explanation, Frequency, Instrument, Manner, Means and more. Some of the frame to frame relations are that it inherits from Misdeed, it is inherited by Abusing, Kidnapping and Theft and it precedes Criminal_Investigation. In this thesis, the frames and their FEs are presented in different fonts and written with capital letters for the purpose of clarity.
As becomes apparent from this explanation, by using the frame index, one can investigate the core meaning of a set of words without interference from grammatical structures. This means that FrameNet acknowledges that while two words might have different grammatical properties, they can call to mind very similar situations (Hasegawa et al. 2011, 109). That is, they evoke the same frame.
However, like most theories (and especially one that is so new), frame semantics has flaws and blind spots which need to be kept in mind while using FrameNet as a model for research. Firstly, while many frames (like the previous examples of selling and committing a crime) are based around an action, there are frames that are based around objects or states of being. This often causes frames in a single sentence to overlap and have the LUs of the sentence belong to multiple frames. For example, in the sentence “Martha carried a bag full of sand.” one identifiable frame pertains to the action Bringing (‘carried’) of which the FEs are Agent (‘Martha’) and Theme (‘a bag full of sand’), while another identifiable frame pertains to the state of being Fullness (‘full’) of which the FEs are Container (‘bag’) and Contents (‘sand’). In this case the FEs present in the Fullness frame only cover ‘a bag full of sand’, while the rest of the FEs ‘Martha carried’ fall outside of the frame. Meanwhile, FEs included in the Bringing frame cover all LUs of the sentence. Therefore, one could argue that Bringing is the major frame in this sentence. However, as sentences get longer and more complex,
12 more frames will be included, which could possibly cause more overlap. This could complicate the process of sentence analysis because it might be unclear which frames are the major frames. Looking at every single frame in the sentence would technically be possible, but because of time constraints that goes beyond the scope of this thesis.
Secondly, most previous research on frame semantics that uses FrameNet has focused on the analysis of informative texts (Friberg Heppin and Toporowska Gronostaj 2014; Hasegawa, Lee-Goldman, and Fillmore 2014; Hasegawa et al. 2011). Because of the lack of previous research, it is still unclear how useful exactly the model is for the analysis of texts with a purpose other than to offer information to the reader. For example, novels or even film scripts, which may rely heavily on jokes and plays on words for their entertainment value. Since the art of making puns relies on the presence of ambiguity of meaning, it might be difficult to determine the frames in such cases.
Thirdly, while the FrameNet model is very useful to determine ST-TT loyalty (by comparing the presence or absence of frames and frame elements, and thereby comparing the core meanings of ST and TT), loyalty to the source text is not preferred in all translations. Depending on the skopos of the text, the accuracy of a translation could even be largely irrelevant (Hasegawa, Lee-Goldman, and Fillmore 2014, 195). This shortcoming is of course no problem for the current study since its very purpose is to attempt to determine how similar a transcreated text is to its original, and not to assess the quality of the translations.
Finally, at the most detailed level of meaning, this model does not provide much nuance. A hamburger and a bowl of miso soup are interchangeable with the word ‘food’ in a sentence like “She quickly finished her food”. A frame-based analysis would not find any difference in frame or FEs, no matter whether the sentence concluded with ‘food’, ‘hamburger’ or ‘miso soup’. However, I will be using this model for just the part of my case study that is aimed towards finding similarities and differences in the overall meaning of sentences. I am therefore confident that the second part of the case study analysis will make up for this shortcoming.
The second part of my case study will focus on the smaller details of the translation that frame semantics will unlikely be able to pick up on. Since it is likely that the types of changes found in the data will be very diverse, no one methodology or theory would be applicable to all parts of the data. Therefore, I will perform a qualitative analysis by taking examples and sorting them into categories. I will then gloss each example and discuss the changes made during translation in detail. 2.1.2 Japanese-English translation
Despite its importance in the modern-day translation industry (because of the large amount of computer hardware and software that is exchanged), the Japanese-English language pair suffers
13 from a severe lack of proper research in the field of translation studies. Traditionally, western
translation studies have focused mainly on Indo-European language pairs, most of which use the same alphabet and are (partly) typologically and ancestrally related (Wakabayashi 1991, 414). This means that most existing translation theories were created for the analysis of, and tested mainly on languages like English, Spanish, French and German. As I mentioned in the first part of this chapter, I found in my bachelor thesis (de Goeij 2016, 27) that this causes compatibility problems between translation theories and languages like Japanese. The theories sometimes over-explain what is simply a typological difference between two languages, and fail to explain some other translation practices unique to language pairs with different ancestry.
There is one author who has produced (or co-produced) the few pieces of academic literature that presently exist that focus purely on Japanese-English translation, and that is Yoko Hasegawa (Hasegawa, Lee-Goldman, and Fillmore 2014; Hasegawa 2012). Hasegawa’s Routledge
Course on Japanese Translation (2012) is the only academic book I have managed to find that
discusses various existing main-stream translation theories in relation to the Japanese language, and comments on their limited applicability. However, other than her work on the development of FrameNet, she has not created a translation theory that is better suited to explaining Japanese-English translation phenomena.
In terms of language typology, English is classified as a mainly analytic language (which means that it relies mainly on separate words and word order to convey meaning), while Japanese heavily favors agglutinative structures (which means that it mainly conveys meaning through the inflection of words) (Hasegawa 2014, 3). This basic difference – combined with other factors like the fact that Japanese tends to use onomatopoeia to describe the manner of actions, while in English the manner is usually integrated into the verb stem (Hirose 1981, 1) – means that changing parts of speech and grammatical structures during Japanese-English translation is impossible to avoid. Additionally, the grammaticalization of respect (Hasegawa 2014, 255), the existence of benefactives (Hasegawa 2014, 164), and the common usage of sentence final particles (Hasegawa 2014, 293), are all integral elements of Japanese that are difficult to convey in English and thus are forcibly omitted in translation. In Vinay and Darbelnet’s terms, it means that servitude to the source language is much more frequent with the Japanese-English language pairs than with typologically related languages.
While there is a lack of substantial research on Japanese-English translation, I have found many instances where scholars make some mention of a translation difficulty related to this language pair in works on other subjects such as video game localization. In her article on transitive and intransitive constructions in Japanese and English, Luk has found that while Japanese prefers
14 sentence constructions that describe actions as though they happen spontaneously, English prefers constructions that directly involve the actor performing the action (Luk 2014, 779). In her article on the prioritization of entertainment value in video game localization, O’Hagan acknowledges that Japanese-English translation is very challenging because Japanese uses ideographs which can contain a lot of meaning in relatively few characters compared to English (O’Hagan 2009, 154). Gillberg (2013, 16) reiterates this and adds that even the two phonetic alphabets of Japanese, hiragana and
katakana, do not match up with English in terms of meaning nuance.
2.2 V
IDEO GAMESAs stated in the introduction of this thesis, it was not until relatively recently that the academic world has started to consider video games as a worthy study subject. Contributing to this belated conclusion is the fact that new media like video games are usually met with initial resistance and suspicion by older generations (Hung 2011, 2). By now, video game studies has grown out to be a multidisciplinary field of study that covers subjects of a wide range, from the economics of the video game industry to the portrayal of gender in video games (Newman 2008, 8). However, most of these sub-categories of video game studies bear very little relevance to this thesis and thus will not be discussed in detail. In this section I will introduce and define the basic concept of a video game for the purpose of clarity.
The answer to the question “what is a video game?” might seem clear-cut at first glance, however with the extremely wide range of genres of electronic entertainment that presently exists, it is important in a thesis like the present one to clarify what the term ‘video game’ refers to exactly. Firstly, to be considered a video game, the medium must be electronic. Secondly, electronic
simulators and training programs that have education as their sole purpose (such as military operation simulators and driving simulators) are not considered video games in this thesis; games must have an element of entertainment to them. Lastly, this thesis only considers a piece of software to be a video game when there is some form of negative consequence when the player does not succeed at the task the software presents them with. This means that interactive visual novels in which the player merely decides which way the story progresses (i.e. it is impossible to fail), do not classify as video games according to the definition in this thesis. In summary, what is considered a video game in this thesis is: “A piece of software for an electronic medium, the main purpose of which is to entertain the player, and which offers some kind of negative consequence when the player fails.”
15 The above definition includes many of the currently existing genres of video games, though genre itself is a complex concept within the video game medium. While older video games were easily divisible into a set of unambiguous genres like platformer, shooter or racing, many modern video games defy such categorization due to their multitude of gameplay and design facets (Costales 2012, 392). For example, one of the game series used in the case study of this thesis, Phoenix
Wright: Ace Attorney, is (in)famous for defying the norms of what a video game is normally
considered to be and is especially hard to put into one of the existing genres (Gillberg 2013, 9). It is almost like an interactive visual novel in the sense that there is no action-based gameplay involved, and it is very text heavy. However, it is indeed a video game because when the player selects the wrong answer, they are given a penalty.
2.2.1 Video Game Localization and transcreation
Like with the term ‘video game’ it is necessary to briefly explain what is considered to be ‘video game localization’ in this thesis. Video game localization, as it is defined in this thesis, is the process of adapting textual and non-textual assets of a video game to make it suitable for markets in different countries and cultures. That is, unlike translation which refers to a concrete action of translating a string of text from one language into another, video game localization is a process that involves all kinds of experts including managers, graphic artists, voice actors, translators, editors and quality controllers. A ‘localization’ of a video game then, is the resulting product of this complicated process.
While it has been getting more attention in recent years, video game localization is still not a subject that has been studied in depth in a lot of instances. Many of the articles that focus
specifically on video game localization serve mainly to introduce the basic concepts and challenges of the localization process and do not go into detail (Costales 2012; Bernal-Merino 2007; Czech 2013; Mangiron and O’Hagan 2006; Hevian 2007). There are a few academic works that are
exceptions to this. Costales (2014) discusses how translators of super hero video games have to find a balance between being loyal to the original version of the game and being loyal to the previously existing translations of the comic books, performing a case study on the localization of the game
Batman Arkham Asylum. Mangiron (2013) describes the various unique aspects of the use of
subtitles in video games, though her article doesn’t include any specific case study.
Many of the works I have mentioned above emphasize that in order for a localization to be successful, it should “retain the look and feel of the original” while providing players with an
experience as though the video game was developed in their own language (Mangiron and O’Hagan 2006, 14). To achieve this, all elements in the game that make the player aware that this game was
16 not originally intended for them and in the process ruin immersion, should ideally be removed or replaced (Chandler and Deming 2012, 110). This transformation of the source material that goes beyond accurately translating of a body of text is what Mangiron and O’Hagan (2006) have dubbed
transcreation. In practice, matters that most often restrict transcreation are budget (every image or
animation that is altered costs more money) (Mandiberg 2015, 240) and franchise loyalty (a player expects a character to have the same name throughout a series of games and between books, movies, games and other media) (Bernal-Merino 2007, para. 5).
Two particularly relevant works on the subject of transcreation are two theses, one by Mandiberg (2015) on the localization industry’s cultural responsibilities towards the audience, and another by Pelletier-Gagnon (2011) on the American video game localization industry’s attitude towards Japaneseness. Both of these works essentially discuss the strategy of transcreation to deal with culturally specific ST elements, though Pelletier-Gagnon does not use this specific term. Mandiberg heavily argues that using transcreation for the purpose of removing culturally specific items and situations is a fallacy that only serves to dumb-down the audience by refusing to expose them to foreign cultures. Therefore, he does not consider this to be proper translation practice and calls for the industry to stop erasing cultural elements from video games during localization. Pelletier-Gagnon, however, acknowledges the drawbacks of transcreation but also highlights its benefits and does not focus as heavily on moral judgement. However, unlike the present study, neither of these researchers look at this phenomenon more closely than scene level. Since the purpose of this research is to analyze in detail what happens to the TT when transcreation is applied to help localization specialists to make decisions that suit their purpose, I do not consider it my place to make a value judgement on this matter in this thesis.
2.2.2 The influence of the internet
The impact the internet and internet culture has had on the video game (localization) industry, is not often mentioned by scholars when discussing video game localization. Because of this lack of
thorough research on this subject, it is difficult to make any definitive assertions, but I will now state two main hypotheses which may help to further make sense of the possible results of my case studies in chapter 4. Both of these hypotheses relate to how the worldwide gaming community has evolved and educated themselves and one another through the use of the internet, and – despite what the academic world has said about video game localization practice – how this might have started to affect the practice of transcreation.
17 Like James Newman (2008, 12–13) discusses in his book on the audience’s interaction with video games, even single-player video games usually are not enjoyed in a complete vacuum in this day and age. The internet allows a player to look at reviews before even purchasing the game, discuss their experience of playing games with other players on the other side of the world, make and share artwork, fiction or even fan-made games based upon commercially distributed video games, and much more. Some video game fans even end up making the discussion of video games their fulltime job, like many video game reviewers, let’s players (people who record themselves while playing video games) and video game theorists on the video streaming website YouTube have done. Many video game companies also sponsor these YouTube stars (often on the premise that they give their video game a positive review), which shows that the industry itself acknowledges the importance of the internet community.
Though the localization industry might not always be swayed by everything a fan community wants, one of the first times the gaming audience did visibly affect a video game company’s attitude towards localization was when developer Square Enix changed their localization process from outsourcing to in-house in 1998 after the backlash from fans who played the poorly translated Final
Fantasy VII (O’Hagan and Mangiron 2013, 180). Since then, the internet has not only provided the
video game community with more and more easily accessible ways to communicate with the industry, but it has also allowed them to educate themselves on the process of localization. In some cases, the video game developers even contact their audience directly in the form of surveys. These surveys are usually sent by e-mail and pose questions to the audience on the subject of preferences and opinions on previously released games and what they would like to see in future releases.
Pelletier-Gagnon (2011, 73) mentions that sometimes video games that are heavily
transfigured during localization are met with resistance because the audience feels betrayed by the product they received. I hypothesize that this is related to how much more acutely aware the audience is of which video games they play were not originally produced in their language, and what might have been lost or added during localization than they used to before the internet was this widely used. Given the fact that the video game industry clearly does pay attention to their audience at various levels, it would then be logical that they adapt their localization strategies to this more aware audience, perhaps deciding against transcreating video games as much as they used to. This might also become apparent in the results of the case study of the current thesis.
Additionally, the internet has also made it much easier for audiences from all corners of the world to educate themselves and each other on different cultures. When you want to know what a
magatama is or how udon noodles are made, a quick Google search will provide more and better
18 part of daily life. Therefore, the industry might not consider it necessary to use transcreation as often as it did before. The translator of the very first Phoenix Wright: Ace Attorney game, Alexander Smith (as cited in Mandiberg 2015, 245–46) has stated that – for the reason stated above – he might not have decided to change miso-soup to ‘hamburgers’ at all if he had translated the game in this day and age.
19
3 M
ETHODOLOGY
3.1 D
ATA SCOPE AND ACQUISITION METHODSince the case study in this thesis will involve two individual games, it is important that a similar amount and type of data will be retrieved from both games. This will ensure that the data will be as representative for each of the games as possible, and therefore that the results of the case study will be as relevant as possible. To achieve this, I will take approximately 25.000 words, comparative to roughly 38 pages of transcript from each of the two video games.
I am aware that this amount of data is only a small part of the text of each of the video games, but due to the fact that applying the FrameNet analysis method is a time consuming and laborious process, analyzing any larger body of text is beyond the scope of this thesis. However, I am confident that this amount of data will be enough to produce reliable results. The analysis method used in this study ensures that, despite the relatively small amount of data, the results of the study will be significant and meaningful due to its scrutiny of cognitive meaning on the sentence and word level.
Because the purpose of this research is to attempt to uncover how exactly transcreation affects the localized versions of video games and because the scope is fairly limited, it is imperative that the data used is of relevance. Because of this, I have chosen to exclude certain parts of the in-game text from my case study data. I will exclude menu text because other restrictions (such as spatial constraints) can apply to this type of text and skew the data. I will also exclude any tutorial text that teaches the player the mechanics of the games and the functions of the buttons on the console, since that type of text is likely to always be translated very literally because its sole purpose is to instruct the player.
To convert the in-game text from the video games into a file that can be edited during analysis, I will transcribe the text of the original Japanese video games from so called let’s plays (videos of people playing video games, showing the game console screen) on YouTube. I will then use my own Nintendo 3DS console to play the games in English and transcribe the text from them. This method is the most suitable, because in both Ace Attorney: Dual Destinies and Persona Q, the in-game text that the player is shown depends on the choices they make. Therefore, playing the games in the same way and making the same choices as the person does in the let’s play video will maximize the chances that the appearing text will match.
20
3.2 D
ATA ANALYSIS METHODAs explained in chapter 2.1.1 of this thesis, the main method of analysis used in this case study will be frame semantics. More specifically, the practical method called FrameNet, developed by a team of researchers at the University of Berkeley.
After gathering the required amount of data from the two video games and their localizations, I will analyze each utterance (a coherent piece of text uttered by one character) by identifying the major frame or frames in both the Japanese original and the translation and
comparing whether those frames match or not. If the frames do not match, I will note whether this is because the frame in question is missing in the original, whether it is missing in the translation, or whether they are simply different from each other. In other words, if one utterance contains three frames in the original version but only two frames in the translation, I will mark one frame as missing in the translation. The other two frames will then be marked as ‘match’ if they match or ‘different’ if they do not match. If the frames do match, I will analyze them further and note whether most FEs also match, some FEs match while others do not, or even (nearly) none of the FEs match at all. If an utterance is not coherent (for example if the character merely says “Uhm.”), and thus cannot belong to any frame nor is able to be analyzed with any other analysis method, I will not take that utterance into account.
In the event that I find a significant shift in the translation that FrameNet is unable to pick up on – in other words, there is a difference in meaning, but the analysis result is ‘Frame match’ and ‘FE match’ – I will also take note of this and analyze and discuss it further in chapter 4.2 of this thesis. Should I find utterances with significant meaning that do not belong to any existing FrameNet frame, I will create my own frame and define it for the purposes of this research. To keep the analysis as objective as possible I will only create new frames when absolutely necessary and add a list of the frames I created to the appendix of this thesis.
After analyzing all the data in this manner, I will present the results in a quantitative manner and attempt to draw several generalized conclusions from it. Then I will further analyze and discuss a selection of examples, divided into categories based on what kind of change occurred during
21
4 C
ASE STUDY
4.1 T
HE GAMES AND GAME SERIES:
A
CEA
TTORNEY ANDP
ERSONABoth of the video games used for this case study are part of ongoing and long-running series that are popular in both Japan as well as the US and Europe. The Ace Attorney series kicked off with its first instalment Phoenix Wright: Ace Attorney (Japanese: 逆転裁判 [literally: Turnabout Trial]) in 2001, though the game was not localized into English until 2005. The video game used in this case study is
Phoenix Wright: Ace Attorney – Dual Destinies (Japanese: 逆転裁判 5 [literally: Turnabout Trial 5]). It
is the fifth main entry in the Ace Attorney series and was released for the Nintendo 3DS in 2013.
Persona, which is actually a spinoff series of another game series called Shin Megami Tensei,
released its first game Revelations: Persona in 1996 and was localized in that same year. The game from the Persona series used in this case study is Persona Q: Shadow of the Labyrinth which was released on the Nintendo 3DS in 2014.
These two particular games were chosen because they have many factors in common such as being released on the same platform, having release dates quite close to one another, and being text-heavy Japanese role-playing games. This is likely to limit the differences found in translation caused by factors irrelevant to this study such as differences in spatial constraints (due to screen size for example) or technological differences. At the same time, these two games seem to differ greatly on the subject this thesis is researching in detail: transcreation as a localization strategy.
4.1.1 The story
The Ace Attorney series revolves around the main character Phoenix Wright (Japanese: 成歩堂 龍一 [Naruhodou Ryuuichi]), who is a defense attorney. In most of the games the attorney faces the most ridiculous and seemingly hopeless cases, usually involving a murder of some kind. In Dual Destinies, a mysterious bomber attempts to take justice into their own hands by blowing up a courtroom. At the trial for that bombing, another bomb is set off, injuring Phoenix’s good friend and colleague Apollo Justice and implicating another friend of theirs as the culprit. Since Apollo is too injured to appear in court, it is up to Phoenix and his psycho-analyst co-worker Athena Cykes to clear their friend’s name and find the real culprit in the process.
22 Unlike Ace Attorney, every game in the Persona series usually introduces new characters and even a completely different universe in every game. The common denominator in the series (aside from gameplay elements) is the concept of characters using ‘Personas’. In the games there are certain people who have a Persona, which is a manifestation of their psyche in the form of a mythical being or monster. These Personas can be used in battle to perform certain physical or magical attacks. Persona Q is actually a spin-off of Persona 3 and Persona 4 combined, in which the player chooses to play as the main character from one of those games. The chosen main character ends up in an unknown other dimension with his friends. In this dimension time and space seem to be different and the group of friends is trapped in a strange school where a cultural festival is being held. They cannot leave the school and they cannot communicate with the school’s students. They then meet a mysterious boy and girl named Zen and Rei, who have lost their memories and are also trapped in the school festival dimension. To get out of the other dimension and back to their own world, they have to conquer labyrinths and fight monsters called Shadows, all the while trying to retrieve Zen and Rei’s memories.
4.1.2 The gameplay
In Ace Attorney, most of the time the player watches passively as the characters interact with each other. There are two gameplay modes within the game. One occurs when a trial is in process, while the other occurs when the characters are collecting evidence for the case. During trials, the player reads the conversation and is sometimes prompted to make a decision, answer a question, or present the correct piece of evidence in order to try and prove the defendant’s innocence. If the player chooses wrongly, the judge will call the main characters out on their mistake and will start to lose faith in their arguments. If the player makes too many mistakes during one trial, the judge will declare the defendant guilty and the player will have to restart the trial. During the investigation process, the player chooses where the characters should go and who they should talk to in order to gather enough evidence to be able to win the next trial. In this part of the gameplay the player cannot lose and is simply forced to keep searching until all the required evidence has been collected.
The gameplay in Persona is much more like a classic JRPG. While there are still a lot of conversations for the player to read (or listen to, since most of the conversations in this game have full voice acting in both the Japanese and English versions), in between the conversations the player has to navigate what is usually referred to as ‘dungeons’ in the video game community. These dungeons are extensive multi-story labyrinths with enemies attacking in the form of random ambushes the player cannot see coming. When the player is engaged in battle in this fashion, they
23 must select what each of the characters will do (e.g. fight, use magic, use an item, defend
themselves). Doing this in a strategic manner – keeping in mind the allies’ and enemies’ weaknesses and strengths – will make it more likely that the player wins. If the player loses a battle, they have to start again from the last time they saved the game.
4.1.3 The localizations
Though both Ace Attorney games and Persona games used to be translated in such a way to remove as many of the Japanese elements in the games as possible, the localization strategies for Phoenix
Wright: Ace Attorney – Dual Destinies and Persona Q: Shadow of the Labyrinth are wildly different.
This is because starting from Persona 3, translation strategies for the Persona series were adjusted to maintain Japaneseness instead of erasing it, most likely because the growing Persona fanbase consisted largely of cultural enthusiasts who appreciated these foreign aspects. Logically, it would be much harder to implement such a change in the Ace Attorney series since those games are part of one linear story with the same protagonist instead of introducing a new story and protagonist with every new game like Persona. Suddenly changing Phoenix Wright’s nationality from American to Japanese in the localization of a new game would likely be met with resistance from fans of the series.
Some of the differences caused by these different strategies are easy to see at first glance. In Persona character names are not changed in the localized version, whereas in Ace Attorney all characters including the main characters get a new name during localization (I will elaborate on these changes in section 4.3.7). Meanwhile, the foreignizing strategy of the Persona translation is so extreme that many of the honorifics – which are an integral part of the Japanese language – are maintained. From this, one can conclude that the Persona localization team expects the audience of this game to have a large amount of previous knowledge of Japanese culture.
4.2 P
RESENTATION OF THE DATAFor both of the games in this case study, the data that was collected, covered the first section of the first chapter in the respective stories. In the entirety of the data from Phoenix Wright: Ace Attorney –
Dual Destinies 1218 separate frames were identified. In the data from Persona Q: Shadow of the Labyrinth 1038 separate frames were identified. Figure 1 shows how many of the frames from the
24 data from each game matched and in what way in concrete numbers, while Figure 2 shows how the frames from the two games compared in terms of percentages.
Figure 1 – frame count in Ace Attorney and Persona
Figure 2 – frame percentages in Ace Attorney and Persona
From figure 2 in particular, it becomes clear that the rate of frame matches is much higher for Persona than it is for Ace Attorney, with about 82% of frames in the Persona data matching up in
0 100 200 300 400 500 600 700 800 900
Ace Attorney Persona
Frames differ E missing J missing Frame only match Partial match Full match
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Ace Attorney Persona
25 some way and only around 67% of frames in the Ace Attorney data doing so (see figures 3 and 4 for the exact percentages). In other words, the core meanings of the sentences in Persona’s translation are often much closer to the meanings of the original sentences than is the case in Ace Attorney’s translation, a result that reflects the games’ reputations.
Figure 3 – Frame percentages in Ace Attorney
Figure 4 – Frame percentages in Person
57.31% 9.11% 0.82% 11.90% 1.07% 19.79%
Ace Attorney
Full match Partial match Frame only match J missing E missing Frames differ
76.97% 5.49% 0.19%2.79% 1.73% 12.81%
Persona
26 Though in both data pools over half of the frames fully matched up, this percentage is much higher still in Persona’s translation than in Ace Attorney’s translation. From this we can conclude that it is likely that transcreation results in meaning being less literally conveyed. This is also
reflected by the fact that partial matches were actually more frequent in Ace Attorney than Persona. In such cases the meaning of the concept underlying the sentence will have been the same, but certain FEs were added, missing or different, often causing the translation to invoke a different feeling than the original. What significance these percentages hold when compared to other video game localizations or even compared to other media and text genres is impossible to conclude as long as there is no further research undertaken in a similar manner.
In both data pools there were very few cases where the frame matched but none of the other elements within the frame did, just as there were relatively few cases in which the translation was missing some meaning that was present in the original without substituting this meaning with something. From that second fact one could generalize that it is possible that transcreation is not very likely to cause parts of an original text to be omitted without compensation. Conversely, a significant difference that could be a result of transcreation affecting the translation, is that almost four times as many frames were added in Ace Attorney’s translation as there were in Persona’s translation. This means that meaning that was not present in the original text was added to the localized version.
Finally, a substantial amount of the data in both of the games (though this percentage is much larger in Ace Attorney) consisted of frames that were different. However, this was also one of the areas of the data where frame semantics does not offer as much information as would be preferable, since it merely indicates whether the frame is different or not, and not how different it is. I will elaborate on this in chapter 4.3.4.
Though this quantitative representation of the data is a good way to get an idea of the general ways in which transcreation could affect the translation of a video game, it says less about the specific ways and details in which the frames and FEs match up or differ. To shed light on these differences, nuances and cases where frame semantics was unhelpful to identify changes in translation, I will discuss a number of examples in the next section.
4.3 A
NALYSIS OF THE DATAIn this section I will discuss a wide variety of examples found in the data in more detail with the purpose of gaining further insight of how transcreation as a strategy affects a localized product.
27 Since the types of changes between the original and the translation that were found, differ so greatly, it is extremely difficult to apply any single theory to all of them. Therefore, I will simply provide the Japanese sentence of each example with a roman alphabet transcription, a gloss and a more literal translation, before discussing the changes made during translation in detail. Due to the limited scope of this thesis, it is not possible for me to detail the definition of all the frames named during the discussion of the examples. However, since the definitions and other details of all the frames are available at “https://framenet2.icsi.berkeley.edu/fnReports/data/frameIndex.xml”, the database can be used as a reference in case of any uncertainty or confusion.
Though I will not be looking at many instances that resulted in a full frame match – simply because most of those cases are clear-cut and therefore not very interesting to investigate further – I will firstly provide some examples in which the translation is still rather different despite the frames and FEs matching. Similarly, I will also highlight some cases where the frames differed, even though the meaning of the sentence remained roughly the same. I will then look at how many of the frame mismatches were due to servitude to the English language and how many were by voluntary choice of the translators, and I will take a closer look as to how significant the differences between original and translation were when the frames were marked as different. Finally, I will discuss some cases in which it was difficult to determine whether frames and FEs matched or not.
After discussing frame related examples, I will move on to non-frame related examples that require a closer look without frame semantics being involved. These examples will involve the translation of things like puns, jokes and proper names. Additionally, I will examine how each of the games’ translations handles uniquely Japanese language items like honorifics and kanji (ideographs with inherent meaning). Finally, I will highlight the few translation mistakes that I have found in the data.
In each of these following sections I will compare the examples from the two video games and attempt to conclude from them whether the differences or similarities found could have anything to do with transcreation being the translation strategy or not. I am aware that many of the examples could belong to more than one of the categories I have divided them into, but I have attempted to sort them into the category that they are the most suitable to.
4.3.1 “False” frame match
When I refer to a frame match as “false”, this means that though the same frame is used in the original and the translation, the cognitive meaning that the translated sentence (or a large part of the sentence) evokes, is actually quite different. These cases made up an insignificant part of the
28 data, and as such I consider them to be exceptions. However, since FrameNet is a little tested
analysis method, I would like to highlight its shortcomings in the interest of refining and improving it. As the examples below will illustrate, I have found that a “false” frame match is often caused by the fact that some frames can include a very wide range of meanings.
The following example sentence from Ace Attorney is preceded by a conversation in which the defendant and Athena Cykes’ friend Juniper Woods, says that she has given Apollo Justice a lotus root, because she heard that lotus root can help you see into the future.
ココネ：（生だったら、さすがに食べられなかったみたいだけど）
Nama da-ttara sasuga=ni tabe-rare-na-katta mitai da
Raw COP-CON as.one.would.expect=LK eat-POT-NEG-PST seems COP
kedo but
Lit: ‘But if it was raw, it’s logical that it seems [he] was unable to eat it.’ Athena: (And if that’s true, I guess he must not have eaten them…)
The main frame in this sentence is Conditional_scenario, since this is the only frame of which the FEs cover all of the LUs in the sentence. In this frame there is a possibility that has a consequence associated with it. In this sentence the frame Conditional_scenario is represented with the LUs ‘dattara’ and ‘if’ in Japanese and English respectively. The Profiled_possibility is ‘nama’ in Japanese and ‘that’s true’ in English, and the Consequence is ‘sasuga ni taberarenakatta mitai da kedo’ in Japanese and ‘I guess he must not have eaten them’ in English. The parentheses present in the original Japanese and official translation indicate thought instead of speech.
The Japanese sentence offers the fact that the lotus root was raw as the reason Apollo did not eat the lotus root, while the translation refers to the fact that Apollo was not able to see into the future and therefore could not have eaten the lotus root. If I was to speculate about the reason for this change, it is likely that while it is general knowledge in Japan that you do not eat lotus root raw (even though it can be eaten raw), in America this might not be as logical.
The next example, also from Ace Attorney, involves a witness called Ted Tonate who is an expert in bomb defusing, who explains that defusing bombs is not easy.
バラシマ：マグロの解体ショウとはわけがちがう！おわかりでショウ？ Maguro =no kaitai shou towa wake=ga chiga-u
29
O-wakari de-shou
POL-awareness COP-POL
Lit: ‘[It is] something different from a tuna dissecting show! [You] know [that], right? Tonate: It’s not like disassembling a toaster, you know!
In both the original Japanese and the English translation, the main frames in this utterance are Similarity and Awareness (though the second one I will not discuss since it is a normal match). The Similarity frame covers both expressions of similarity and negations of similarity, as is the case in this example. The frame is represented by ‘wake ga chigau’ in Japanese and ‘not like’ in English. The remaining FE is Entity_2 (‘Maguro no kaitai shou’ and ‘disassembling a toaster’ respectively).
In this case it is quite obvious why the translator made the decision to change this cultural element in the translation. Public displays of dissecting fish are not something an American or European audience would be familiar with, so it was simply replaced with a toaster. Once again, because this frame covers a wide range of meanings, and the basic structure of the sentence remains the same, FrameNet is unable to pick up on functional equivalents. It is useful to note that in Persona such cases of false frame matches due to the presence of realia did not occur. Culturally specific items such as the dish ‘takoyaki’ were maintained in the translation.
In the data from Persona there were even less cases of “false” frame matches, despite the higher rate of frame matches in general. I deduce that this lower percentage of “false” matches is caused by the more direct translation strategy, causing more translated sentences to exactly contain all elements from the original Japanese sentence. There were cases where there was a small change made in a sentence that FrameNet could not detect, but it was difficult to assess whether these cases were truly a “false” match. This is illustrated in the following example.
？？？：入っちゃダメ…その中、怖いのがいるよ… Hai-ccha dame
Enter-SJV do.not
Sono naka kowai no=ga i-ru=yo
That in scary thing=NOM be-NPST=SFP
Lit: ‘Don’t go in… in there, [there] are frightening beings…’ ???: Don’t go inside… There are monsters in there…
The first part of the sentence is a normal frame match, where the identifiable major frame is Path_shape. The second part of the sentence is also a frame match, yet the meaning is not
