Haiku Whisper’ – an (inter)active audio installation
Poetry as performance is a diminishing art form, particularly in a society that is moving further into the separate physical spaces of individual media devices and away from shared art experience. It is suggested that this is particular manifestation of a wider disconnect with our auditory environment. Into this context is presented Haiku Whisper, a poetry generating system designed to create unique poetry in the Haiku tradition to an unsuspecting audience. Furthermore, the poetry is created in response to participant’s movement through physical space, and played back in real time; a pause in movement creates a pause in performance. It is hoped that such a system can reintroduce an overlooked art form and induce a heightened awareness of our personal soundscapes in a playful, creative way. This paper considers the academic imperative for the development of such a system, presents the technical aspects required for execution, and discusses the reception of the preliminary public installation.
Wall - sensor - Arduino - MaxMSP - speakers - wall. It's as simple as that.
Haiku - (俳ai句 haikai) is an ancient Japanese form of poetry, consisting of 17 on (also known as morai), which in English translates – almost - as 17 syllables, put together in three phrases of 5, 7 and 5. The essence of Haiku is in the juxtaposition of two images or ideas, and a kireji or ‘cutting word’ between them. The fundamental aesthetic quality of Haiku is that it is seen as internally sufficient and, despite its brevity, can bear consideration as a complete work. Furthermore, each of the three lines is considered to represent a conceptually complete thought or action. These two features contain potential for creative manipulation whilst maintaining the integrity of the tradition.
Haiku Whisper an interactive oral installation - with orality understood as both the act of speaking and listening  that generates and performs random yet structured Haiku poems to unsuspecting audience members as they move through a public space. It is an attempt to re-introduce an unsuspecting audience to a cultural aesthetic that has been losing ground in contemporary society, and by choosing Haiku as a platform seeks to reflect an increasing cultural interdiscursivity. Furthermore, it attempts to do so whilst posing the audience questions about how the understand their everyday auditory landscape.
Poetry as communication has existed throughout history but, has lost traction in recent times; as a form of entertainment is has become increasingly confined to taught environments and literary circles . Furthermore, as a society we are becoming increasingly removed from our immediate environments as we withdraw into our digital devices, becoming ‘alone together’ .
With these twin premises as the point of departure, the interactive installation Haiku Whisper was designed to investigate the latter with the use of the former. By using hybrid aesthetics that reflect the character of the modern generation, the hope is to make the audience question their assumptions of their everyday auditory environment. The primary intentions of the project are:
- To communicate a cultural aesthetic that has been losing ground.
- To create an interactive system that faithfully generates unique Haiku in a coherent, sensical form.
- To create a sense of ‘belonging’ to the interaction; the user should be aware of their role in the piece, even if their understanding does not amount to control.
A strong and varied body of interactive poetry generation systems has appeared in the recent past, with varied success. Tosa  developed the Interactive Poetry System in which human and computer agents collaborate through the exchange of poetic phrases, whereas Rokeby’s The Giver of Names  generates poems based on objects selected by participants and collated on a pedestal. Zhu et al  created Poetry Mix-Up, in which users create a ‘re-mixed’ form of poetry by communicating with the system via Short Messaging Service (SMS), bringing a social communication value to the interaction and grounding it in modern culture. With Hitch Haiku  Tosa et al present system that generates Haiku based on participant selected phrases from a famous Japanese essay (1000 Books and 1000 Nights), translating the outcome into English. Although Haiku Whisper is a less complex – and therefore arguably less nuanced – generation system, it’s strength and advantage lies in the ability to connect the participant to their physical and audio space in a manner that has not been addressed before.
Oliveira defines four categories of poetry generation systems, defining Haiku Whisper as ‘template based’, in which systems fill templates that carry syntactic and/or rhythmic constraints . Template based systems are acknowledged as the least complex, but suit the design requirements adequately. A more nuanced approach may be suitable for later elaboration.
A series of imbedded Passive Infra-Red (PIR) sensors register movement through and Arduino microcontroller. This in turn triggers the random selection of audio files contained within a Max/MSP patch, feeding back through a multi-channel audio system to a series of paired speakers matched with their corresponding sensors. The PIRs and speakers are embedded within the external walls of a busy London street.
The Max Patch
There are three PIR/speaker units, accounting for the three lines constituting a Haiku. Distances between three and four meters separate these. When triggered, the corresponding sound file plays, and lasts between two to three seconds. This correlates with the time it takes a fairly fast paced pedestrian to reach the next PIR, thus triggering the following Haiku stanza in time and rhythm with the last. Therefore if the participant walks continuously, they will hear they will experience their uniquely generated haiku spoken to them as they move through the space. In contrast, if they are to stop or pause, so will the performance, logically only resuming as they do.
The sensors are HC SR501 PIRs, selected for their high sensitivity, diminutive size, and low cost. Furthermore, they are equipped with two variable resistors to calibrate both re-trigger delay and range and are also highly compatible with the Arduino microcontroller. The HC SR501 has two IR detecting lenses; when a body passes it first intercepts one lens, causing a positive differential range between the two halves, thus triggering the microcontroller. The sensors were placed in custom built cases and mounted in 200mm steel tubes to reduce the sensitivity from a 110° field to a band of 50cm at a range of 4m.
Haiku is a language of abbreviation; it is an art form that infers a wider story from the context it creates, allowing for great leaps of imagination and disparate thought processes. As such, a database was created that would reflect and embrace the inter-changeability and idiosyncratic possibilities of the form. The preliminary database used 160 original recordings of genuine, published Haiku mixing both translations from Japanese and those composed in English. Earlier tests experimented with online digital generators, but the results were of a low quality and as such rejected. These original recordings, performed by a professional actor, proved invaluable in ensuring clarity and quality of reproduction, particularly in consideration of the high levels of background noise encountered when the installation was active. Furthermore, Manurung  affirms that a poetic text must hold the three properties of meaningfulness, grammaticality and ‘poeticness’; the database is selected to reflect this.
The technical aspects of the system work well, with a 90% successful trigger rate and no conflict between triggers/sound files being activated concurrently. The only failing with the system at this time is that it only works fully when the participant walks in one direction (in this case East to West) i.e.: files are in grammatical order - 1st, 2nd, 3rd – when triggered from Left, Middle, Right. The system still produces rhythmically complete Haiku when triggered from either direction, A change to the software could reconfigure this, so that the stanzas are triggered in order of poetic occurrence rather that which PIR sensor they are assigned to, but this brings forth new issues concerning both re-setting and multiple participants in the space at the same time. It is therefore proposed that a unidirectional installation is preferable to such a potentially chaotic one.
In terms of actual poetic output, the system produces complete, yet not always coherent, Haiku poems. There is an obvious element of abstraction and surrealism to the output, but this was fully expected and even encouraged in the database selection. In addition, due to the extreme brevity inherent in Haiku, it is by definition abstract and relies heavily on symbolism for it’s impact; when there are so few words to available, what is inferred is often more important than what is said . In addition, the structure of Haiku lends itself to the type of manipulation undergone in Haiku Whisper. As stated above, traditional Haiku entails the juxtaposition of two contrasting images or ideas – contained in the first and last phrases – and a kireji or ‘cutting word’ between them, which defines the manner in which the juxtaposed images are related and interpreted. This is generally manifest in the middle phrase. Therefore, it is perfectly feasible to switch phrases between Haiku whilst maintaining the poetic integrity and tradition of the form. As such, Haiku Whisper should be viewed as a success in terms of producing output that can confidently be identified as Haiku.
The installation was a site-specific design for placement on Mile End Road, East London, E1 4NS. As such, the installation was activated and observed for 2hrs on Sunday 17th February 2012 from 1-3pm. During that time, 32 people became participants in the piece; 17 single, six pairs, and one trio. Their physical reactions were observed and recorded.
Of the 32 participants, all were observed to react in some form or other. The smallest reactions involved simply turning the head in surprise or expectation of finding someone close by who had spoken the words themselves. Other reactions included a slowed or quickened pace, stopping entirely, inspecting the wall in which the audio units were embedded, and calling out to question who was there. An initial shock was observed on several occasions, as participants were seen to jump or turn sharply when the first stanza played.
The majority of participants understood that the audio was coming from the wall of the building, and those that were in pairs or more we often heard to be discussing as much. Of the 32, those in pairs (or more) were seen to be more inquisitive as to the source of the noise (although this is undoubtedly due to other social factors and not the inquisitiveness or lack thereof of individuals), and three pairs and the trio investigated long enough to discover that it was in fact their movements that was causing the audio to play. Of those, two pairs then spent several minutes interacting with the piece, deliberately triggering responses to signs of amusement.
Whilst all the above is presented as preliminary observational data, it undoubtedly provides insight into both the effectiveness and reception of the design.
Haiku Whisper is presented here as an investigative installation for reconnecting both with a diminishing cultural aesthetic and our own urban audio environments. The tool with which it does so consists of a hidden interactive Haiku generator, developed to perform a unique, coherent Haiku to an unsuspecting audience in a public space in reaction to their movement. Inasmuch as the intention was to create both a system and a reaction, it is a success. When analysed further, there is certainly room for further exploration. It was a stated aim for the work to create a sense of belonging in the audience, even though they would originally not be aware of their role in the installation, and this was only partially successful. The nature of being a ‘hidden’ work dictates that a majority of participants would not engage fully with it. Perhaps the next incarnation should find a middle ground between being completely hidden and explicitly displayed.
- Dietze, S. 2002. Ten Dreams of Technology. Leonardo 35. Vol.5. Pp. 509-522.
- Hiraga, M. 1999. ‘“Blending” and an Internpretation of Haiku: A Cognitive Approach’. Poetics Today. Vol. 20, No.3, Metaphor and Beyond: New Cognitive Developments. Pp. 461-481.
- Manurung, H. 1999. ‘An Evolutionary Algorithm Approach to Poetry Generation’ in Proceedings of 1st International Workshop on Literature in Cognition and Computing (1999).
- Olivera, H.O. PoeTryMe: A versatile platform for poetry generation. CISUC ‘08. University of Coimbra, Portugal. 2008.
- Ong, W. ‘Orality and literacy: The Technologising of the Word’. London, UK. Routeledge. 1971.
- Schloss, W.A. and Stammen, D. ‘Ambient Media in Public Spaces.’ Sophia Antipolis Microelectronics Conference’08. (SAME ’08). Vancouver, British Columbia, Canada.
- Tosa, N. 1998. Interactive Poem. In ACM SIGGRAPH 98 Conference Abstracts and Applications (SIGGRAPH’98). ACM. New York, 300.
- Tosa, N.,Obara, H., and Minoh, M. ‘Hitch Haiku’. 2007. Proceedings of the 2nd International Conference on Digital Interactive Media in Entertainment and Arts (DIMEA’07). 6-7. Perth, Australia.
- Turkle, S. 2011. Alone Together: why we expect more from technology and less from each other’. Basic Books, New York.
- Zhu, K., Ranasinghe. N., Edrisinghe, C., Noel, O., Fernando, N., and Cheok, A.D. ‘Poetry Mix-Up’. ACM Computers in Entertainment. Vol. 9, No.2, Article 8. July 2011.