You are on page 1of 21

The death of the translator in machine translation

A bilingual poetry project


Tong King Lee

The Hong Kong Polytechnic University

This paper explores the notion of the death of the Translator, inspired by Barthes formulation of the death of the Author. It argues that the death of the Translator is caused by a loss of human agency in translation and is therefore most clearly exemplified in machine translation. Based on an avant-garde bilingual poetry project by a Taiwanese poet, the paper demonstrates that machine translation can produce unexpected new meanings through unpredictable routes of semantic and syntactic divergences from the source text. The poets use of transparency as physical medium and of machine translation as mediator raises the following questions: does translation actually allow us to read through a source text? If so, to what extent is such translation transparent? How should we even come to terms with the concept of transparency with respect to the meaning of a literary text in translation? The paper argues that in the bilingual project in question, machine translation plays the crucial function of bringing the readers attention back to the target language by way of delaying/blocking comprehension, hence rendering the corporeality of the target language transparent. Keywords: Hsia Y, death of the translator, machine translation, dissemination, diffrance

0. Introduction In his famous proclamation of the death of the author, Roland Barthes debunked the long-standing myth in literary criticism that meaning,1 seen as some sacrosanct entity, resides in the originator of an utterance. Preparing the epistemological grounds for post-structuralist thinking, Barthes proposed the displacement of the author as the source of meaning, positing instead that it is language which speaks, not the author; to write is, through a prerequisite impersonalityto reach that
Target 23:1 (2011), 92112. doi 10.1075/target.23.1.06lee issn 09241884 / e-issn 15699986 John Benjamins Publishing Company

The death of the translator in machine translation

93

point where only language acts, performs, and not me (Barthes 1977: 143). A crucial motif here is that of impersonality. Due to the inherently intertextual nature of all texts, the authority of the author2 as the origin from where meaning is derived dissipates within the interwoven threads from multiple sources (which themselves come from other, multiple sources). It is in this sense that writing is seen as impersonal or depersonalised, since its meaning cannot be attributed to a singular entity:
[L]inguistics has recently provided the destruction of the Author with a valuable analytical tool by showing that the whole of the enunciation is an empty process, functioning perfectly without there being any need for it to be filled with the person of the interlocutors. Linguistically, the author is never more than the instance writing, just as I is nothing other than the instance saying I: language knows a subject, not a person, and this subject, empty outside of the very enunciation which defines it, suffices to make language hold together, suffices, that is to say, to exhaust it (ibid.:145; emphasis in original).

If every utterance is nothing more than a fundamentally linguistic moment (as opposed to an author/speaker communicating to a reader/listener via the medium of language), subjectivity is, in theory, detachable from the act of writing itself. Barthes theory thus prompts a radical rethinking of the centrality of authorial subjectivity in the determination of literary meaning, and of the extent to which the author may exercise jurisdiction over the interpretive possibilities of a work. If one extends Barthes formulation on the dethronement of the Author beyond a single language/text into the discursive space of translation, can we similarly pronounce the death of the Translator, who is, of course, a kind of author in his/ her own right? And if we are to make such a pronouncement, what theoretical implication might it have for our understanding of literary meaning? The idea of the translators demise seems to contradict more recent sociological studies of translation, wherein the voice and agency of the translator as an active player in the negotiation of meaning is foregrounded and indeed celebrated (see Wolf 2007 ad passim). Yet just as Barthes negation of the authors subjectivity has led scholars to critique the subjectivity of the reader (e.g., P. Chen 2009: 181), it may be argued that if the author can be dispossessed of his/her agency in the production of meaning, as the many strands of post- theories in Translation Studies have attested, the subjectivity of the translator may similarly be called into question. The death of the translator is most directly caused by a loss of human agency in translation, and is therefore most clearly exemplified in machine translation (notwithstanding the fact that humans are responsible for the intricate programming behind all machine translations). Often seen as a poor substitute for the human translator and used only for the most mechanical of tasks, machine translation can at times produce unexpected new meanings through unpredictable routes of

94

Tong King Lee

semantic and syntactic divergences from the source text. What do we make of such unintended meanings produced by the unconscious machine? Since writing is essentially a kind of language performance, according to Barthes, translation, as a form of writing, too must be regarded as a discursive activity wherein only language acts. If Barthes elimination of authorial control over interpretation heralds the death of the Author, an elimination of the translators subjectivity through machine translation signals the death of the Translator. This leads us to a consideration of how machine translation as language performance leads to the death of the Translator, as well as the implication of this for an understanding of meaning in translation. The present paper is a case study of an English-Chinese bilingual poetry project in which machine translation is a central mechanism. The project in question is an avant-garde literary endeavor in bilingual poetry writing which employs the Internet and a computer translation application. Being the first of its kind, at least in the Chinese world, the project is itself a literary experiment. Besides theorising the writing/translation process in the project, this paper further seeks to participate in and extend the projects play with the machine translator, with an eye on tracking the path of dissemination that meaning undertakes. In the following, the paper first discusses the effects of machine translation on literary meaning in Hsia Ys bilingual enterprise, drawing on post-structuralist theories on writing/translation. This is followed by an experimental play with Hsias engagement with the machine translator through the use of multiple translation programs. The paper concludes by addressing the theoretical significance of machine translation for a nuanced understanding of the meaning of meaning. 1.

The machine translator in Hsia Ys Pink Noise

In 2007, the contemporary Taiwanese author Hsia Y (b.1956) published a collection of thirty-three bilingual poems by the name of Fenhongse zaoyin/Pink Noise, taking Chinese literary avant-gardism to a whole new height. As suggested by the title of the collection pink noise being a technical term used in audio engineering to denote a type of spectral density the concept underlying the genesis of this work is noise, belletristic noise (wenzi zaoyin) to be exact. In an interview (Hsia 2008, see also Bradbury 2008), Hsia revealed that the inspiration for her collection came from noise and low-frequency acoustic art CDs, which prompted her to contemplate the possibility of applying noise aesthetics to the written word. The creative process started with the poet plucking random English phrases3 from resources in cyberspace, specifically, from endless chain of blogs and the many websites that popped up when hyperlinks in spam were clicked. These phrases

The death of the translator in machine translation

95

were then pieced together in such a way that they resemble the visual form of poetry (in the poets own words, to to fit them together until they clicked like a music box), and then fed into the translation application in Sherlock 4 to generate corresponding Chinese versions. Based on the co-text/context in the machinegenerated translations which can turn out to be quite different from that in the corresponding English poem as a result of collocational shifts the poet then tweaked the original English phrases and ran them through Sherlock again to produce a new translation. The cycle was repeated a number of times for each set of poems. Finally, Hsia juxtaposed the intertwined English and Chinese texts to form parallel texts, creating what she called, in quotation marks, a volume of translated poetry (fanyi shi).5 The material form of Pink Noise embodies the innovative spirit of the poet. Hsia printed her poems not on paper but on transparency sheets. Each set of parallel poems is printed in two colours: the English poems in black, and the Chinese (translated) poems in pink. The choice of colour is intentional: the machine translations, which are stringently literal and thus highly unreadable, constitute verbal noise to the ear of the reader who has been inculcated with the doctrine of fluency. They are thus printed in pink to recall the technical jargon of pink noise, echoing the title of the book. The poems are also lineated in such a way that the English poems are left-justified, while their Chinese translations are right-justified. The visual impact of such a layout is stunningly postmodern. It reminds us of Kress and Van Leeuwens (1996: 23032) notion of inscription, which pertains to the multimodal effect afforded to a text by the material qualities of the medium (such as paper) used. For Pink Noise, when the reader collapses and holds together the transparency pages on which a set of parallel poems is printed, the English and Chinese lines superimpose on and partially overlap each other, so that it is possible to perform a parallel reading of each line in the two languages (and in two colours as well). In doing so, one is literally reading the Chinese translation through the English text, as the transparency on which the translated poem is printed lies beneath that on which the English poem is printed. It is this transparency and reading through that Pink Noise prompts us to think about. Does translation actually allow us to read through a source text? If so, to what extent is such translation transparent? How should we even come to terms with the concept of transparency in respect to the meaning of a literary text in translation? 1.1 From the death of the Author to the death of the Translator The process by which Pink Noise is conceived places the constructs of authorship and translatorship on the margins. Hsia Y does not profess to be the author as the term is conventionally understood of the English poems in her collection.

96

Tong King Lee

As the strings of words that constitute the original poems are randomly taken from the Internet, the origin of each work, if any at all, is varied, fragmented and heterogeneous. This patching together of disparate jigsaw-like phrases that eventually do not quite form an organic unity (neither is such unity desired by the poet) is reminiscent of a similar technique used by Hsia in her third project Moca wuyi mingzhuang (Rub Ineffable). The latter work was basically created out of Hsias second poetry collection Fuyushu (Ventriloquy), specifically by dismantling (by way of physically cutting out) individual phrases from Fuyushu and permutating (by way of pasting) them to form new poems. This pastiche technique was once again enacted in Pink Noise, except that the poet carried it further this time by taking bits and pieces of linguistic material from cyberspace. The following example is the first stanza of the first poem in the collection:
Brokenhearted time and ordinary daily moment How fucking creepy is that? So different and sweet A promise awaits us At the limits of the mystical glow If we must die We will need those rhyming skills Some people are born with Others develop (Hsia 2008)

The fragmented nature of the source texts has immense implications for the (in) determinability of literary meaning in Pink Noise. Each line in the original poems bears the trace of an unrecorded past source (which is by no means necessarily literary in nature), and yet differs from its past usage through its recontextualisation in the poem. Thus, each line in an English poem may take on a certain sense in its source, wherever that is. When uprooted from its original site of occurrence, a phrase or clause can acquire a potentially new sense as it is re-planted into the poems co-text, this co-text being made up of other phrases that are similarly derived from various digital sources in an almost casual manner. The meaning of each line, then, cannot be determined or pinned down as if it were something discrete and concrete, encapsulated and locked within the English signifiers. Meaning is rather an emergent entity that is in perpetual flux, always arising as a consequence of an opposition to other (absent) concepts, as opposed to being a presence in and of itself (Davis 2001: 15).Yet there is no final interpretation as such. The text remains open even when seen as a whole, for the juxtaposition of various English lines to form a poem is but an illusory act on the part of the poet to create the formal image of poetry, as it is conventionalised. In the hands of Hsia, strings of

The death of the translator in machine translation

97

words from different sources come together and cohabit, often uneasily, within the discursive space of the poem to form an intertextual site where meanings encounter, clash and coalesce to form new senses. Critics may suggest that indeterminacy of meaning is characteristic of all poetry, and cannot thus be seen as a distinctive feature of Pink Noise. While this may be true, it should be noted that such indeterminacy is deliberately foregrounded by the poet in a very material way that is not often seen in other poetry collections. By originating the individual phrases in her English poems in an un-patterned manner from myriad digital sources, Hsia relinquishes her authorial subjectivity and consequently much of her control over meaning. She has enacted the death of her perceived role as author by adopting a literary mode that seeks to bring to the fore the intertextual network within which language is embedded and meaning disseminated. In so doing, Hsia produces what she has variously called anti-poems (fanshi ), non-poems (feishi ) and pseudo-poems (weishi ), refusing to settle on a fixed name and reminding the reader that the word poem must be interpreted within quotations (Hsia 2008). Hsia did not stop here in her experiment with meaning. A critical stage in the creation of Pink Noise was the use of the translation application in Sherlock. The products of machine translation are invariably characterised by foreignised syntax and disjunctive collocations, which contribute to a high degree of rupture in linguistic flow. Meaning in translation proceeds and stops intermittently, baffling the Chinese reader to the extent that the reader has to consult the corresponding English poem frequently to gain access to the Chinese translation.6 Of interest to us is the fact that a line in an English poem is often reincarnated into something quite different in its machine-translated version. To take an example, the following is the Chinese translation of the English stanza cited above, with the back-translation given by Google Translate:
Translation by Sherlock: (Hsia 2008)

98

Tong King Lee

Back-translation by Google Translate:


Heartbreaking moments of the times and the general daily What sex is it creeping? Very different and sweet We promised to wait for Limits of love in the mysterious In the bright light, like Gods glow If we must die We will need those skills in rhyme Some people are born with Others showing

The title of the English poem contains the phrase heartbreaking time, which is rendered as lingren xinsui de shidai, literally heartbreaking times/milieu/era. The machine translation thus changes the temporal perspective of the title from a more delimited and discrete frame, suggesting the time when the event of heartbreaking takes place, to a more macroscopic and abstract one, suggesting a milieu where hearts are broken. Interestingly, the back-translation by Google Translate reverts the Chinese word shidai (times, milieu, era) back into its discrete meaning with the word moments, thus tying the English back-translation (and, by extension, the Chinese translation) back to the English original. The signifier, time, has thus looped back onto itself, with a diversion in-between via a double translation (first from English to Chinese, and then back from Chinese to English), illustrating the unpredictable trajectory of meaning that a signifier may undertake through machine translation. The first line of the English poem How fucking creepy is that? uses the expletive as an adverb, quite typical of low-register English vernacular, with no particular reference to the denotative meaning of the word sex. A human translator with a minimal degree of familiarity with English would arguably have understood and rendered the emotional load of the word, as opposed to its denotative sense. The machine translator, however, disregards such emotional meaning and goes straight for the literal sense: zenme xingjiao shi rudong what sex is creeping? (see back-translation). The adverb fucking has thus been concretised into the Chinese verb xingjiao, which denotes sexual intercourse. The adjective creepy in the English text undergoes a similar semantic mutation, its meaning being transformed from that of scary to that of moving slowly along the ground (rudong). If one tries to make sense of the barely grammatical Chinese sentence, it could mean something like what kind of sexual intercourse is a creeping motion? The literal rendition performed by the machine translator has unexpectedly protruded the sexual element that is at most only latent in the English line, bringing

The death of the translator in machine translation

99

an interesting turn to the interpretive potential of the poem. Such concretisation of sexual innuendos through machine translation occurs in a number of poems in the collection (Yeh 2008: 174). The above example shows that the machine translator has assumed a curious kind of unconscious autonomy of its own, enabling the original English poem to branch off in a way that a human translator, bounded by the ethics of translation, would not tolerate. By delegating the work of translation to a computer program, Hsia has relinquished yet another form of subjectivity that of the translator, and she does so for a reason. To Hsia, the machine is the ideal candidate for the role of the translator by virtue of its carefree mindlessness and tendency to render a source text literally. A machine translator is superior to the human translator because it makes no commitment, and like any lethal lover, it announces from the very beginning that it is not to be trusted (Hsia 2008). This unconsciousness of the machine translator has been exploited by the poet to create a poetical aesthetics that challenges conventional norms of literary production and reception. This aesthetics is best encapsulated in Hsias description of the linguistic nature of machine translation: words keep coming but it doesnt move forward. Nor does it take you anywhere; it persists in place even as it relentlessly crumbles, sentence by sentence it crumbles, when suddenly it arrives somewhere (ibid.). This is the aesthetics of rupture, of noise, which can only be realised by the machine, for the human translator has a professional tendency, albeit to varying degrees, to somehow smooth out the meaning of the source text for target readers. The machine translator, in contrast, has such utter unconsciousness as to ignore the meaning, both denotative and connotative, loaded into the signifiers of the source text by convention and tradition. It is this unconsciously unethical stance on the part of the machine that underlies, or even motivates, Hsias attempt to deconstruct meaning by way of creating lettristic noise. This very special type of noise allows a translation to break away from the control of the source text, generate novel interpretations (suddenly it [the word] arrives somewhere) and gain its own meaning potential. This autonomy of meaning in translation is achieved at the expense of the loss of human agency; adapting Barthes (1977: 148) formulation, the birth of noise aesthetics must be at the cost of the death of the Translator. 1.2 Machine translation and Hsia Ys physical desire for words The common metaphors used to conceptualise translation point to its communicative function, for example, the bridge metaphor and other metaphors relating to the communication model (see St. Andr 2010 ad passim). What if one were to turn things around and posit translation as a non-communicative, or even mis-

100 Tong King Lee

communicative, act? To Walter Benjamin, the hallmark of bad translation is communicability: any translation which intends to perform a transmitting function cannot transmit anything but information hence, something inessential. (Benjamin 1923 / 2000: 15). Within the literary circle in Taiwan, Hsia has been known for her unorthodox, non-communicative modes of expression, the distinctive features of which include non-referentiality and anti-narrativity.7 Hsias disregard for the interpretability of meaning and the grammatical acceptability of her poems makes her a controversial figure. Nowhere is this more obvious than in Pink Noise, where the poet solicits the help of the machine translator to evolve a brand of poetics that challenges the shockproof faith in the communicability of meaning (Greer 1991: 153). Most of Hsias poetry collections, with the possible exception of her fourth work Salsa, display, and indeed flaunt, her usual disposition to create jarring linguistic effects. This disposition climaxes in Pink Noise, in which Hsia undertakes a more ambitious attempt to release meaning from the constraints of its signifiers in a drastic way, in order to achieve the absolute liberation of language, a liberation theology/languages theological liberation (Hsia 2008). Specifically, Hsia enacts her ideal mode of writing by recourse to the machine translator, which contributes to the discursive thrust of Pink Noise by exacerbating the non-communicability of meaning. Such non-communicability is first and foremost realised in the form of ungrammaticality and anti-narrativity. Consider the following poem, taken from the collection:
Things seem to get worse before they get better When, from a long distant past, nothing subsists After the people are dead, after the things are broken and scattered She poised herself on the balance beam gracefully And he waited with his fingers poised over the keys Whos ready to remind us Amid the ruins of all the rest Everything vanishes around me And works are born as if out of the void Ripe, graphic fruit falls off My hand has become an obedient instrument flying of [sic] a remote will

Translation by Sherlock:

The death of the translator in machine translation 101

Back-translation by Google Translate:


Bad things seem to get better in their When, from the distant past, long-form, no living After the man is dead, the thing is broken and dispersed after the To maintain a balance her own beautiful balance beam He waited with his fingers to maintain a balance in the keyword Who is ready to remind us Among the ruins in all the rest Everything around me disappeared And work as if born outside the valid Mature fruit fall images My hands become a remote control will be subject to the instrument flight

The back-translation departs considerably from the Chinese translation (the original in this second translation), but it might give non-Chinese readers some sense of the syntactical disjuncture that obtains in the Chinese text. Some clauses manage to cling onto the fringe of logicality in the translation, despite being cast in a heavily exotic, foreignised syntax. For instance, the two clauses in the second line, zai ren shi si de zhihou, zai shi shi canpo he qusan zhihou are awkward constructions in Chinese, but if the copula shi (translated from are in the English original) is dropped, the Chinese clauses do generally correspond to their English counterparts, on the linguistic level, that is. Other clauses, however, mutate into unrecognisable chunks of words as if juxtaposed by chance. The fourth line, And he waited with his fingers poised over the keys becomes a nonsensical sentence that is impossible to decipher with the logic of Chinese grammar: ta dengdai le yu ta de shouzhi baochi pingheng zai guanjianzi. While the original English line is a hypotactic sentence that hinges on the preposition with, the machine translator changes this preposition into the Chinese conjunction yu (and). But as the two clauses cannot be easily interpreted as having any conjunctive relation with each other, the sentence becomes paratactic. If one forces an interpretation on the ungrammatical sentence, the translated line might read He is waiting to keep a balance on the keyword together with his finger. To the rational reader trying to imbue the sentence with a determinate meaning, this reading would be unsound,

102 Tong King Lee

not least because it suggests that the subject he is waiting do something together with his finger, as if his finger were a separate entity with its own subjectivity. Such nonsensicality is aggravated by the fact that the machine translator (or machine poet, as Hsia calls it) has rendered keys as guanjianzi (keywords) in Chinese. The word key(s) is polysemous in English; within the co-text of the English poem, it could refer to the keys of a piano, considering the conventional collocation fingers poised over the keys. (It is worth noting, however, that this meaning is not finalised in and of itself, but a consequence of diffrance, that is, it emerges through its intertextual relationship with other signifiers as well as contexts. That is why it is not possible to pin down the meaning of key with any confidence; any tentative meaning that we subscribe to the word is the result of our understanding of how the word is used and what it means in other contexts (cf. Davis 2001: 15).) If we tentatively set down keys of a piano as the referent of the English word keys, the word undergoes yet another transformation through machine translation. The sense of keywords in the Chinese text is a figurative extension of the literal meaning of keys. Machine translation has thus accidentally uncovered the metaphorical meaning underlying the word keys. The path taken by the word keys in this example demonstrates what Derrida called the retentive and protentive characteristics of meaning (Derrida 1967/1974: 47). A signifier does not refer to something already present and fixated within itself, but is related to something other than itself, thereby keeping within itself the mark of the past element, and already letting itself be vitiated by the mark of its relation to the future element (Derrida 1972b/1982: 13). The word keys derives its sense from other elements within its co-text as well as from previous uses of the word in other contexts (the retentive aspect). The same word also allows itself, through machine translation in this case, to be extended into its (future) incarnation as guanjianzi or keyword in Chinese (the protentive aspect). But what sense does it make to speak of poising his fingers over the keywords, which is what the Chinese translation reads literally? Nonsensicality troubles the translation here again, burdening the reader with the task of interpretation, and it is exactly this non-sense/ nonsense that the poet is trying to force readers to come to terms with. Machine translation, through its intractably literal translation or, perhaps more aptly, trans-mutation and sheer ignorance of grammaticality and idiomaticity, churns out unintentional imagery, such as that of human fingers poising over abstract keywords, whatever that is supposed to mean (indeed, if it is intended to mean anything at all!). This generates an abrupt aesthetics that challenges the readers poetic sensibility and his/her sense of security as regards how meaning is coherently communicated through a linguistic medium. Other ungrammatical clauses in the translated poem include zai suoyou xiuxi zhizhong feixu, corresponding to Amid the ruins of all the rest (Line 6)

The death of the translator in machine translation 103

and bingqie gongzuo shi chusheng haoxiang zai wuxiao waimian, corresponding to And works are born as if out of the void (Line 8). In the first of these lines, the word rest (meaning remainder) in the English poem is translated into xiuxi (to take a rest/break) by the machine, and the noun feixu (ruins) takes the position of a verb in the Chinese context, rendering the translated phrase ungrammatical. It is up to the reader to fill in some verb (such as biancheng to turn into) in front of feixu in order to produce a valid verb-object structure. Literally, this unacceptable Chinese line might read to ruin amid taking (all the) rest, where rest refers to a state of inactivity or repose. (It must be emphasised though that this is one out of several possible readings.) The reading does not make sense according to our conventional way of looking at meaning, but it is also by virtue of this ungrammaticality that a new image is conjured up in the mind of the reader: how could a state of rest become a context within which ruins are located? How is this translated Chinese line, if at all interpretable, intertextually and interlingually related to the original English line? Here the exotic form of the poem evokes new interpretive possibilities. In a similar vein, Line 8 in the English poem is interpretable, insofar as the sentence conforms to established rules of grammar, though this in no way suggests that its meaning is closed. The Chinese translation, in contrast, distorts the syntax and translates void into wuxiao (invalid, see back-translation). This grammatical distortion, coupled with the literal rendition of void, makes the translated sentence incomprehensible by conventional logic. Once again, if we force an interpretation onto the line notwithstanding its ungrammaticality, it is literally as if work comes about outside the invalid. Interestingly, the same Chinese line is turned by Google Translate into And work as if born outside the valid (see back-translation). If one juxtaposes the original English line, the machine-generated Chinese translation and the machine-generated English back-translation, an interpretive tension ensues. What might this line actually mean, if it is intended to mean anything at all? Does it mean that work comes out of nowhere (original English), that work that has come about is valid [outside the invalid] (Chinese translation) or that work that has come about is invalid [outside the valid] (English back-translation)? Or does this semantic plurality suggest that meaning is a fluid entity that follows a meandering course, especially as it threads through different languages? Ungrammaticality, the trademark of machine translation, is often seen as a justification as to why machines cannot surpass human translators when it comes to translating complex discourses. In Pink Noise, however, this perceived flaw becomes a discursive resource that the poet taps into to embody the materiality of language. The more ungrammatical and nonsensical a translated poem sounds, the greater the extent the reader is compelled to focus on the linguistic sign per se, rather than on the construct of meaning, which is a very slippery notion in

104 Tong King Lee

Pink Noise. This is not to say that meaning is out of the whole picture; meaning, as it were, is still the motivation for the readers pursuit in his/her reading, but no longer assumes (or, more accurately still, assumed by readers to assume) a concrete form within the discourse of the poetry collection. Meaning is here comprised of traces, that is, the textual relation between a contingent sense that a word or phrase might possess in an English poem and its various senses as used in other (absent) contexts, as well as the relation between this contingent sense and its potential sense(s) in Chinese translation. If each English poem in Pink Noise is a multidimensional space in which a variety of writings, none of them original, blend and clash, a tissue of quotations drawn from the innumerable centres of culture (Barthes 1977: 146), translation then is a site of flux where the multitude of strands from each poem are proliferated into exponentially more strands of meaning in the target language. Meaning, rather than being established by the word, is disseminated through the word as traces of traces (Derrida 1972a/1981: 26). Such traces are theoretically invisible and therefore indeterminable by the lay reader. In Pink Noise, however, Sherlock produces automatic translations that defy syntactical conventions to such an extent that the baffled reader cannot but succumb themselves to the mercy of the physical manoeuvres of the signifier itself. This is the very intention of Hsia Y: to problematise the reading process and hence draw the readers attention back to the Chinese script and to the way meaning is refracted rather than reflected across language barriers. The ungrammatical clauses in the translated poems make for difficult reading, which is exacerbated by the physical form of the book that renders the reading process extraordinarily tedious.8 Such difficulty protrudes the signifier as the single most important entity in Hsias creative work, outside of which there is literally nothing. In one of Hsias earlier poems (Hsia 1997), the poet writes of a physical desire for words (dui wenzi de rouyu zhiai ), which can be taken as the poets declaration of her disposition in writing. The erotic overtones evoked by the word rouyu (literally desire for the flesh) here suggest the poets extreme affinity with the material form of the Chinese language, including its script and sound.9 Such affinity has motivated Hsia to expose the skin and flesh of the Chinese language by highlighting its visceral form while breaking down its conventionalised meaning. In this light, the production of ungrammatical discourse in Pink Noise can be seen as Hsias attempt to test the malleability of the Chinese language in the hope of extending its horizons (Hsia 2008), and this desired ungrammaticality can only come about with the death of the (human) translator and deployment of the machine (translator):
This [machine] translator is so preoccupied with fidelity, so infinitely faithful, that it radically estranges everything. The original and translation are supposed to

The death of the translator in machine translation 105

complement each other and create a shared meaning, yet all those faithful, declarative fragments often take on a different shape when glued together. Note that every sentence in the source text has a clear structure and is thoroughly translatable, whereas in the adjacent translations, every word and phrase is a linguistic entity, with all the characteristics of language in its totality, each driven to engage in an equivalent exchange and winding up physically united but spiritually apart (or is it spiritually united and physically apart?), cleaving like a shadow yet drifting farther and farther apart, estranged beyond recognition but with every detail infinitely magnified (Hsia 2008; my emphasis).

The ungrammatical constructions produced by the machine create a cacophony of words that estrange readers, jolting them out of their comfort zone of reading and challenging their taken-for-granted beliefs about the stability and sensibility of meaning. The paradox of unity in form and disparity in sense embodied in translation (physically united and spiritually apart) is Hsias desired outcome in her experimental use of the machine in literary writing. By establishing the products of machine translation as a mirror image of the English poems, Hsia is at once reveling in the logographicity of the Chinese language and mocking the humandiscursive construct of meaning. To traditional critics such audacious attempts are nothing more than a casual game with language, a deliberate attempt to deconstruct/destruct the Chinese language. From a theoretical perspective, however, Pink Noise is significant in its foregrounding of the textual qualities of language including its dissemination of meaning through difference and deferral in not just literary writing but bilingual literary writing. The corporeality of the target language comes into high relief through machine-rendered foreignisations, which in turn provoke a delay, or deferment, in comprehension, as the reader seeks to negotiate and reconcile the traces of differences between translation and original. 2. Proliferating diffrance through multiple machine translations The preceding section illustrated how the machine translator can sometimes contort the semantics of a line of poetry, and that a double machine translation (a first translation followed by a back-translation) can bring the meaning either back to its original sense or further down the path of deviation. This prompts us to contemplate the possibility of having the machine disseminate meaning at free will. What if one were to participate, and indeed extend, Hsia Ys language experiment by proliferating meaning several times over using the machine translator? This section looks at how meaning can take on different turns when put through several translation applications simultaneously. To achieve this, the method of pluralist reading (Yeh 2010) is used: the English poems in Pink Noise were fed

106 Tong King Lee

into three translation programs, namely, Google Translate, Systran and Reverso to generate three corresponding Chinese versions of each poem. The English poem, Sherlocks Chinese version published in Pink Noise and the three other translations were then juxtaposed and compared. The results are expectedly haphazard. Many of the translated pieces are highly unreadable, with the usual traits of foreignised collocations and syntactical distortions. However, the translations should not be too quickly dismissed as a mass of linguistic junk, an outcome of unconscious play by the machine. Some semantic turns undertaken by the machine translator are interesting, not because their unconventional forms might offer some kind of linguistic entertainment, but because they have implications for how we conceive the relationship between meaning and machine. As an example, the title of Poem 29 Those misty memories seem awfully good has several virtual reincarnations. While Sherlock and Systran render misty into the concrete image of baowu or light mist in Chinese, Google Translate chooses (as if it had consciousness of its own) to translate the same word into the abstract mengmeng , a duplicative bi-syllabic morpheme that describes a state of cloudiness. Reverso stands out from the rest by simply refusing to translate: the word misty is left sticking out like a sore thumb in the translated Chinese title. The word senses in the line To make their senses come alive and feel good is translated as the concrete-sounding daoli (principles) by Sherlock and Systran, but as ganguan (sensory organs) and ganjue (feeling) by Google Translate and Reverso respectively. Each machine seems to provide its unique interpretation of the two keywords, and the different translations highlight/downplay different potential facets of the words for instance, the concrete and abstract dimensions of the signified concept that come together to form a more holistic picture. The intriguing thing about multiple machine translations is that each machine unpacks a word in unpredictably different ways. Each translation is neither right nor wrong in itself, and enters into combination with other words in the poem to create new collocations and therefore new senses. For instance, the title of Poem 9 In their near-human pleasure is literally translated by both Sherlock and Systran as zai tamen jinren de lequ . Google Translate, however, renders it as zai tamen jiejin renlei de lequ , which expands the abstract notion near-human into the sense of a physical proximity to human beings (jiejin renlei). Reverso takes an unexpected path: zai fujin de ren gaoxing de , which is a fragment that means the people in the vicinity are happy to The original English line, which itself has no determinable meaning, of course, splits itself into several potential readings in another language through the machine translator. Here Saussures notion of the arbitrariness of the link between signifier and signified comes into high relief, as meaning is left to

The death of the translator in machine translation 107

the unconscious operations of the machine, governed by the invisible hands of a computer program. Poem 18 is interesting as it dwells on one of Hsias perennial themes: sex. The line He and I made it our duty to screw as often as possible, direct and colloquial in English, takes a twist in machine translation. Sherlock renders to screw literally as ningjin (tighten): women de yiwu ningjin yue jingchang yue hao . The sense of tightness here, which is also seen in the version by Systran, is drawn from carpentry and unexpectedly conjures up the physicality of sexual intercourse if read against the English line. Google Translate and Reverso, on the other hand, turn the verb screw into a noun: luogan and luosi (both words referring to the object screw), thus dismantling the semantics of the English line and carrying the poem into another, non-sexual trajectory. It is no surprise that not all the translated poems are readable. Readability, however, is really besides the point here. Neither is this excursion with multiple machine translation merely play, though play is indeed part of the way in which meaning operates, evoking the sense of that which, by the spacing between the pieces of an apparatus, allows for movement and articulation (Derrida 1987/1992: 64). This spacing for movement and articulation is provided for by the arbitrariness of the machine. The results of this pluralist reading in machine translation show that signifiers can potentially assume a multitude of new senses via the machine. At times, this allows the translation to supplement the source text, while at other times the translation projects itself into a totally divergent semantic path. Machine translation thus provides a prime site in demonstrating the contingency and non-singularity of meaning and its slippery relation with signifiers. It also has ramifications for how we conceptualise literary meaning in translation. To what extent does translation communicate meaning? Is miscommunication then also a form of communication that is inherent in translation, just as misreading is part of the process of reading/writing by which authors enact an anxiety of influence (Bloom 1975: 36, 1997: 545)? If misreading is a strategy whereby a writer innovates under the influence of previous works, can miscommunication, as inherent in machine translation, also be construed as a discursive means by which a translation simultaneously derives itself and breaks away from a source text? 3. Conclusion: Machine translation and its manoeuvres in meaning Writing, as Barthes reminds us, is constituted by a multitude of traces, where meaning subsists as a volatile substance in constant movement with linguistic signifiers. Here the temporal continuity of meaning is broken, as the spatiality of writing is foregrounded, leading to an evaporation and systematic exemption of meaning:

108 Tong King Lee

In the multiplicity of writing, everything is to be disentangled, nothing deciphered; the structure can be followed, run (like the thread of a stocking) at every point and at every level, but there is nothing beneath: the space of writing is to be ranged over, not pierced; writing ceaselessly posits meaning ceaselessly to evaporate it, carrying out a systematic exemption of meaning (Barthes 1977: 147).

Potentially, translation brings the multiplicity inherent in writing to exponential levels by extending the space of writing into another linguistic system. Meaning, then, is ceaselessly posited and evaporated for yet another series of cycles in another set of linguistic signifiers. This proliferation of meaning is all the more exacerbated by literalism what Walter Benjamin calls transparent translation: A real translation is transparent; it does not cover the original, does not block its light, but allows the pure language, as though reinforced by its own medium, to shine upon the original all the more fully. (Benjamin 1923/2000: 21).10 Literalism, by virtue of its ostensible transparency, exposes the original text to misreading. In an uncanny way, Hsia Ys experiment with the machine translator reifies this notion of transparency through her use of transparency film as medium and of a computer program as linguistic mediator. While the material space within which Pink Noise is embedded brings readers attention back to the written word, machine translation participates in this by foregrounding the corporeality the ishness of Chinese. To Hsia, it is the machines tendency to be non-fluent that allows her to stretch the limits of the Chinese language. She is therefore of the opinion that [i]ts important that someone do this [referring to the use of machine translation in literary writing] now because this technology is only going to get better with each version. One day the translations are going to be so fluent, these cybernetic Sherlocks will turn into mediocre poets (Bradbury 2008: 38). The result is a foreignising (and, to some critics, disconcerting) breed of interlanguage, a misshapen, hybridised cross-fertilisation of English and Chinese that is reminiscent of Benjamins pure language. Through the arbitrary and contingent manoeuvres in meaning of the machine translator, a new poetical sensibility emerges in Pink Noise, where rupture in sense and syntax is replete throughout. A determinate meaning continually eludes the reader, as meaning arrives only at some tentative point, to be disseminated in-between discordant translated lines, ranging over the space of writing. This article further experimented with multiple machine translations, demonstrating the potential of meaning to be permutated across linguistic borders in manifold erratic moves. Hsias apparent fetishism with the materiality of writing reveals more than just an avant-garde endeavour. The concept of transparency created through Pink Noise is theoretically profound. To Venuti (2008), transparency points to the naturalisation of a source text through the use of fluent prose in the target

The death of the translator in machine translation 109

language. A fluent translation, accordingly, gives target text readers the illusion of having access to a source text directly. Under this regime of fluency, the intervening medium of translation is misleadingly perceived to be as transparent as a glass. Despite Venutis call for translators to adopt a foreignising strategy, fluency persists as a dominant criterion with which translators and critics judge the quality of translated work. The acclaimed Spanish-English translator Edith Grossman, for instance, asserts that
[r]epeating the work in any other way for example, by succumbing to the literalist fallacy and attempting to duplicate the text in another language, following a pattern of word-for-word transcription would lead not to a translation but to a grotesque variation on Borgess Pierre Menard, who rewrites his own Don Quixote that coincides word for word with Cervantes original, though it is considered superior to the original because of its modernity. Furthermore, a mindless, literalist translation would constitute a serious breach of contract. There isnt a self-respecting publisher in the world who would not reject a manuscript framed in this way. It is not acceptable, readable, or faithful, as the letters of agreement demand, though it certainly may have its own perverse originality. (Grossman 2010: 1011)

It is exactly through a mindless, literalist translation that is not acceptable, readable, or faithful that Hsia Y explores the notion of transparency with her own perverse originality. The use of transparencies deliberately creates the material space in which the English poems can be read through the machine translations. Yet the literal machine translations remain defiantly unreadable and hence opaque in terms of their meaning; they are as semantically non-transparent as they are materially transparent. If we juxtapose Hsias concept of transparency with Venutis, we see a difference in focus: for Venuti, transparency is a constructed textual quality that gives readers the false impression that they can see the meaning of a source text clearly through the intervening medium of translation. For Hsia, the concept of transparency does not pertain to the source text, but to the target language. The machine translations do not allow the reader to see through the source text, but rather bring the readers attention back to the target language by way of delaying or even blocking comprehension. The corporeality of the target language thus becomes transparent through literal translation. It is the physical transparency of the book that paradoxically becomes an illusion, since the meaning of the translated texts is for the most part rather opaque. Grossmans comments quoted above testify to the necessity of using the machine as mediator in Hsias project. Compared with the unconscious machine, human translators are, to varying degrees, subject to the governance of the fluency dictum and the constraints of translational ethics. One of the implications of

110 Tong King Lee

Hsias bilingual project is that it alerts us, in a very extreme way, to how translating for sense may suppress latent senses underlying the signifier. Literalism, arguably less practically useful as a translation strategy than semantic and communicative translation, thus has potential value in unlocking some of these deferred meanings by differing them with more conventional senses. And extreme literalism of the type demonstrated in Pink Noise has to come from the machine, which heralds the death of the (human) translator. Pink Noise thus throws a very different light on the machine translator, which in this case is not just an inferior substitute for the human translator. It is the mindless, literalist translation produced by the machine that allows meaning to be disseminated across language borders in the poetry collection, illustrating how the machine can contribute to the proliferation of literary meaning. Lastly, Hsias project also brings our attention to the importance of the mode of inscription, and this and the use of the machine as translating medium, raise questions about the politics of transcription, translation and reading (Huang 2010: 48) in contemporary Chinese poetics.

Notes
1. The word meaning used here and throughout the rest of the paper does not suggest that the meaning of the word meaning is unproblematic. As the primary objective of the following discussion is to illustrate the deconstruction/destruction of literary meaning in translation, all instances of the word should be interpreted within quotation marks. Other keywords that should be read likewise are original, source and target. 2. Here the etymological link between author and authority is telling of the symbolic power accorded to the former. 3. All the parallel poems in the collection are in English/Chinese except for one in French/ Chinese. For convenience, the source texts are generally referred to in this article as English poems. 4. Sherlock was a web tool created by Apple Inc. for the Mac OS that included a translation application (or channel). The software was officially retired in 2007. 5. My quotations of Hsias own words are partly taken from an interview entitled Poetry interrogation: the primal scene of a linguistic murder appended to the end of the book, as translated by Zona Yi-ping Tsou. As Pink Noise was printed without pagination, quotations from the book are herein cited without page numbers. 6. This is not to say, of course, that the original English poems have any determinate meaning. However, the English poems, by virtue of being nominal source texts and being generally readable in English, do constitute some kind of reference point for the reader of the translated poems.

The death of the translator in machine translation 7. Some scholars have thus associated Hsia with the Language Poets, a post-modernist poetry tradition in contemporary Anglo-American literature (see, for example, C.Chen 2009). Notwithstanding the fact that Hsia does not share the radical oppositional politics that is at the heart of the agenda of the Language Poets (Shetley 1993: 138), Hsias oeuvre does exhibit post-modernist/post-structuralist tendencies, even though the poet may or may not have been thoroughly conscious of or concerned with the theories behind what she is doing. 8. Here I borrow Yehs (2008: 177) apt description that [t]he process [of reading Pink Noise] is constantly interrupted (the reader has to turn to the preceding page to check the English original in order to understand the Chinese translation), prolonged (the reader has to pause to re-read the radically defamiliarized Chinese), delayed (the reader has to insert a sheet of plain paper between pages so as to be able to read the words on the page), and distracted (the readers face is reflected on the opposite page when s/he reads a poem and the reader catches the reflection within peripheral vision) (my emphasis). 9. In an interview, the writer expressed her obsession with the materiality of the Chinese language: Ive always loved those sentences that are rendered with a clumsy fidelity, those adorable literal versions that are virtually indifferent to Chinese grammar and all those second- and third-hand translations from Russian via English and Japanese and who knows what else. Chinese writing began as ideograms depicting the traces and tracks of birds and animals, the sound of wind, ripples on water. Even today, when I encounter a sentence pieced together or assembled in some quirky fashion my vision is awash with an intuitive pleasure in the ideographic nature of the Chinese writing system, its eccentric liberties, keen as animal instincts, evolving over time as if they knew no limits (Hsia 2008). 10. Here, of course, Benjamin and Barthes diverge with respect to the notion of origin. In Benjamins scheme, the original text is something of a lofty and reified nature, which literal translation should strive to shine upon all the more fully. Barthes, in contrast, repeatedly deconstructs the integrity of the original text and conceptualises it as a neutral construct.

111

References
Barthes, Roland. 1977. Image, music, text. Trans. Stephen Heath. London: Fontana Press. Benjamin, Walter. 1923/2000. Die Aufgabe des bersetzers. In Baudelaire: Tableaux Parisiens. Heidelberg: Richard Weissberg Verlag. Trans. Harry Zohn as The task of the translator. In Lawrence Venuti (ed), The Translation Studies Reader. London & New York: Routledge. 1525. Bloom, Harold. 1975. A map of misreading. New York: Oxford University Press. Bloom, Harold. 1997. The anxiety of influence: a theory of poetry. New York: Oxford University Press. Bradbury, Steve. 2008. Hsia Yu (an interview). Zoland Poetry 2. 3446. Chen, Chun-jung. 2009. Hsia Yus postmodern language poetry. Chung Wai Literary Quarterly. 38:2. 197227. Chen, Pei-yun. 2009. Translation, interpretation and will-to-power. Chung Wai Literary Quarterly. 38:2. 169195. Davis, Kathleen. 2001. Deconstruction and translation. Manchester: St. Jerome.

112

Tong King Lee Derrida, Jacques. 1967/1974. De la Grammatologie. Paris: Minuit. Trans. Gayatri Chakravorty Spivak as Of Grammatology. Baltimore & London: John Hopkins University Press. Derrida, Jacques. 1972a/1981. Positions. Paris: Minuit. Trans. Alan Bass as Positions. Chicago: University of Chicago Press. Derrida, Jacques. 1972b/1982. Marges de la Philosophie. Paris: Les ditions de Minuit. Trans. Alan Bass as Margins of Philosophy. Chicago: University of Chicago Press. Derrida, Jacques. 1987/1992. Laphorisme contretemps, in Psych: Inventions de lautre, Paris: Galile; trans. Niclas Royle as Aphorism countertime. In Derek Attridge (ed), Jacques Derrida: acts of literature. New York & London: Routledge, 41633. Greer, Michael. 1991. Language poetry in America: 19711991. Meanjin 50.1. 14956. Grossman, Edith. 2010. Why translation matters. New Haven and London: Yale University Press. Hsia, Y. 1997. Fuyushu [Ventriloquy]. Taipei: xianzaishi jikanshe. Hsia, Y. 2008. Fenhongse zaoyin /Pink Noise (Second edition). Taipei: Self-published. Huang, Yunte. 2010. Take it with a grain of MSG: reading translated literatures from other shores. In Carol Maier and Franoise Massardier-Kenney (eds.), Literature in translation: teaching issues and reading practices. Ohio: The Kent University Press. 4352. Kress, Gunther and Theo Van Leeuwen. 1996. Reading images. The grammar of visual design. London & New York: Routledge. Shetley, Vernon. 1993. After the death of poetry: poet and audience in contemporary America. Durham: Durke University Press. St. Andr, James (ed.). 2010. Thinking through translation with metaphors. Manchester: St. Jerome. Venuti, Lawrence. 2008. The translators invisibility: a history of translation (Second Edition). New York & London: Routledge. Wolf, Michaela. 2007. Introduction. The emergence of a sociology of translation. Michaela Wolf & Alexandra Fukari, eds. Constructing a sociology of translation. Amsterdam & Philadelphia: John Benjamins. 136. Yeh, Michelle. 2008. Towards a poetic of noise: from Hu Shi to Hsia Y. Chinese Literature: Essays, Articles and Reviews 30. 16778. Yeh, Michelle. 2010. Literature in identity formation: reading Chinese literature in translation. In Carol Maier and Franoise Massardier-Kenney (eds.), Literature in translation: teaching issues and reading practices. Ohio: The Kent University Press. 117135.

Authors address
Tong King Lee Department of Chinese and Bilingual Studies The Hong Kong Polytechnic University Hung Hom, KOWLOON, Hong Kong ctleetk@polyu.edu.hk

You might also like