Craft, Performance, And Grammars
Craft, Performance, and GrammarsTerry KnightDepartment of Architecture, School of Architecture and Planning,Massachusetts Institute of Technology, Cambridge, MA USAtknight@mit.eduAbstract. Recent interest in new, digital and computational ways of making hasbeen paralleled by rising interest in traditional making and craft practices. Mostefforts to merge digital and craft practices focus on the things produced, withattention to process only to the extent that it informs results. However, thesocio-cultural, aesthetic, and creative dimensions of a craft practice areexpressed in its performative, temporal aspects as much as in its products. Anew computational theory of making offered by making grammars points tonew possibilities for the study of temporal performance. In this paper, I usetraditional kolam pattern making in India as a case study to probe the potentialsof making grammars to represent craft performance, in contrast with the use ofshape grammars to represent craft designs. Different generative strategies arerevealed in the comparison.Keywords: Shape grammar, making grammar, craft, segmentation.1IntroductionRecent interest in new ways of making and in computational tools, technologies andtheories to support them has been paralleled by rising interest in traditional makingand craft practices. Researchers look to by-hand techniques and traditional materialsto advance digital fabrication with new materials. Conversely, hand-craftersexperiment with digital fabrication and new materials to expand the possibilities oftheir craft. “Digital craft” [1] is a phrase often used for work fusing made-by-hand andmade-by-machine methods. In architectural design, craft techniques such as sewing orweaving are emulated in fabrication processes [2], and traditional materials such asbamboo are combined with digital fabrication [3].Most efforts to merge craft sensibilities and practices with new makingtechnologies and computational strategies focus on the things produced (frombuildings to jewelry), with attention to making processes only to the extent that theyfacilitate or inform results. Relatedly, important socio-cultural dimensions of craft areoften left behind in the borrowing or emulation of craft techniques. Socio-culturalaspects of craft, along with aesthetic and creative ones, are expressed in the productsof craft practices. But, just as important or even more so, they are expressed in theperformative, temporal aspects of craft. Craft activities, weaving and calligraphy forexample (Fig. 1), are often public or communal activities – performances in time –
meant to be shared or viewed. They may be deeply imbued with unique culturalvalues and expressive of cultural identities. Understanding and making explicit, evenformalizing and computing, the performative aspects of a craft may help provide newinsights into its cultural dynamics as well as its creative and generative possibilities.However, the performative nature of craft – it’s embodied. improvisational, andtime-based qualities – may seem an uncomfortable fit with formal computation,especially computation of the digital kind.Fig. 1. Expert calligrapher Mohri Suzuki performing calligraphy (left, retrieved from [5]) andKenyan women weaving baskets (right, retrieved from [6]).A new computational theory of making offered by making grammars [4], though,points to new possibilities for the study of craft practice. Making grammars are anadaptation of shape grammars, a long-standing computational theory of design. In thispaper, I use traditional kolam pattern making in India as a case study to probe thepotentials of making grammars to understand and represent craft performance, incontrast with the use of shape grammars to understand and represent craft designs. Inparticular, I consider how making grammars might be used to express temporalaspects of craft performance. While I do not consider explicitly the socio-culturalaspects of craft, or kolam in particular, my work here suggests a computational basisfor socio-cultural and other inquiries.2From Shape Grammars to Making GrammarsShape grammars provide a unique, computational theory of design, one alignedespecially well with creative design practice. They are distinctive for their visualapproach. The rules of a shape grammar generate designs by computing directly withshapes made of basic spatial elements (points, lines, planes, and solids), rather thanwith symbols, words, numbers, or other abstract structures that represent visual shapesindirectly. Computations with shape rules involve seeing and doing. In each step of acomputation, the user can choose what shape to see and then what action, or rule, toapply next.Designing with shape grammars is thus a kind of performative, making activity.Shape grammar theory offers a natural basis for a computational theory of making.
Designing with shape grammars is about doing (drawing) and seeing with basicspatial elements to make shapes. George Stiny and I [4] have recently extended thisdefinition of designing to a definition of making: making is doing and sensing withstuff to make things. I summarize the details of our work here, beginning withinformal definitions of the terms we use. Doing is an action such as drawing, knotting,folding, typing, throwing, stomping, and so on. Sensing includes any one or more ofour senses. Both doing and sensing can be done with ‘‘tools’’. Tools might be ourbodies’ ‘‘tools’’ such as our hands or our eyes, or tools might be extensions of ourbodies, for example, pencils or eyeglasses. Both doing and sensing might includeactions or sensings by a machine as well as by a person. Stuff includes physicalmaterials like gases, liquids, or solids with properties that can be visual, acoustic,mechanical, geometric, and so on. A little more abstractly, stuff can be points, lines,planes, and solids. Things are finite objects made of stuff.With these definitions in hand, Stiny and I adapted shape grammars for computingor making shapes to making grammars for computing or making things [4]. The rulesof a making grammar are based on both the thing being made and a person’s sensoryinteractions with that thing. Thus, a making grammar is a theory of both theconstructive and the sensory aspects of a making activity. Of course, a makinggrammar, like any finite description, can never capture all aspects of a making activityand can only approximate it. The rules are limited to particular aspects of interest.A making rule has the general form M N where M and N are sensed things.More specifically, M and N are things with any sensory interactions indicated in someexplicit fashion. Details of these sensory indications may vary according to the thingsmade. The arrow denotes “replace with” in the usual, formal way. In terms ofmaking, though, the arrow stands for a particular doing and/or sensing.A making rule M N applies to a (sensed) thing T being made, when the makercan identify a copy of M in the thing T. Then the thing M can be changed into thething N. Depending on the thing being made, the formal definition of “copy” might bethe same as that for shape grammars, or it might be specific to the things computed.A making rule can be distinguished as either a sensing rule or a doing rule. Asensing rule represents a perceptual change in a person, through the person’s sensoryactions (moving hands, eyes, etc.) with a thing. It represents a change, shift, or(re)focus of attention in how a thing is perceived. A sensing rule A B says: If a(sensed) thing A is a part of a current (sensed) thing being made, then (re)grasp,(re)focus on, attend to it (with eyes, hands, nose, etc.) as shown by the (sensed) thingB. A doing rule represents a physical change in a thing through a person’s physicalactions (folding, drawing, etc.) with the thing. A doing rule X Y says: If a (sensed)thing X is a part of a current (sensed) thing being made, then do something to it asshown by the (sensed) thing Y.Separating sensing and doing in a making activity is subjective and represents aparticular perspective on that activity. Sensing and doing may sometimes beinseparable. In this case, a making rule may represent sensing and doingsimultaneously, in other words, a simultaneous change in a person and in a thingthrough the person’s sensory and physical actions.In our preliminary work on making grammars, Stiny and I gave an example of amaking grammar for knotting strings, a highly tactile making activity inspired by
khipu, the knotted strings made by the Incas as a physical recordkeeping andcommunication language. The knotting grammar generates single and multipleoverhand knots along a string. In this grammar, the things are knotted strings, the stuffis fiber, doing is knotting (looping, pulling, etc.), and sensing is touching (grasping,focusing attention, repositioning) with the hands. The rules include doing rules forknotting and sensing rules for touching or grasping, as well as a combined sensing anddoing rule. The grammar is a highly schematized version of an actual knottingprocess. But it is suggestive of the possibilities for making rules to encode temporalqualities of knot making. As Stiny and I noted in our paper, the rules capture naturalstopping or stable points in a continuous tying process. Readers can refer to [4] formore details of this example and making grammars in general.3Making Time and Making GrammarsA key feature of making is time. Craft practices, and making activities in general, arecontinuous, temporal events.Making sense of and participating in a craft practice not only involve structuring insome way the spatial aspects of the things being made, they also involve structuringthe temporal actions involved in the making of those things. Such structuring might beretrospective and deliberative – as in analysis – or it might be impromptu andon-the-fly – as in real time making. Shape grammars focus on the spatial qualities ofdesigns and the ways that rules structure designs by segmenting them spatiallythrough visual perception. Making grammars offer opportunities to explore thetemporal qualities of making things, and the ways that rules structure and segmentthings and their making both spatially and temporally.Formal studies of the temporal dimensions of craft and other creative makingactivities is sparse. Exceptions include a proposal for a grammar of human movementrelated to architecture [7], and some studies of movement segmentation andperception in dance [8, 9]. However, research on the temporal dimensions of routine,everyday activities and events, such as making a bed or drinking a cup of coffee, ismore prolific. Research of this kind is pursued in different fields and with differentobjectives. In philosophy, debates revolve around the formal concept of an “event”,including, for example, how an event is related to space and time and how it isrepresented [10]. In cognitive science and psychology, researchers investigate howpeople perceive and comprehend the continuous, fluctuating, multi-sensory stream ofactions in the world. A central idea here is that in order to perceive, understand, learn,predict, and act in the world, people discretize the continuous flow of events intotemporal chunks, each with beginning and end boundaries [11]. Analogies are madebetween event perception and object perception: both events and objects can besegmented and organized hierarchically into bounded parts. In computer science, AI,and robotics, researchers take a more functional and application-oriented approach.Here, the overall objective is to understand human motion in order to buildcomputational models (and robots) to recognize, represent and generate human-likeactions and gestures [12]. The temporal segmentation of motions into discrete,primitive actions is central to this enterprise, as well as the segmentation of the objects
and people engaged in motions. Linguistic approaches using rules and grammars forrepresenting actions are common [13, 14]. Notational systems from dance, forexample Laban movement analysis, are also sources for computational models [15]. Ingeneral, researchers aim for automated segmentation strategies that will be applicableacross diverse, human activities. At the same time, the temporal segmentation ofcontinuous activities is recognized as complex, ambiguous, and subjective [16].The idea behind making grammars intersects with some of the ideas above, inparticular the roles of spatial and temporal segmentations in understanding andgenerating events and actions. Making grammars, though, make no claims forgeneralized knowledge about the way people segment the world. The rules fordifferent making activities, and the ways rules segment making activities, can vary aswidely as the things studied and the people who make them. Also, making grammarsare directed toward creative human activities – not cutting an apple. Further, makinggrammars are limited to representations of things and how they are sensed. Othercontext, such as the maker or performer, the physical setting, and so on, which aresometimes included in computer science work, are not included in making grammars.A making grammar can describe and structure both the spatial and temporalaspects of things and their making through its rules. Making rules describe spatialaspects of things by segmenting things spatially in the same way that shape rulesdescribe spatial aspects of shapes by segmenting shapes spatially. More specifically, amaking rule defines a spatial segmentation of a thing when it is applied in acomputation. When a making rule M N is applied to a thing T, it segments T intoM and other parts of T.Making rules also suggest the possibility to define temporal segmentations of themaking of a thing. In a making rule M N, the replacement operation can beinterpreted as a doing or sensing action in time. A making rule defines a temporalsegmentation of the making of a thing when it is applied in a computation. When amaking rule M N is applied to a thing T to make another thing T’, it segments acontinuous making process into temporal breakpoints defined by T and T’. Moreover,the duration of the action represented by a making rule might be described byassociating a clock or timer with the rule as Stiny and I discussed in our introductorywork on making grammars [4].In the following section, I use the traditional craft of kolam making in India tosuggest how a making grammar might be used to study the performative, time-basednature of a craft practice in contrast to how a shape grammar is used to describe theproducts of a craft practice.4Kolam: Easy to Design, Hard to MakeKolam are traditional, ritualistic patterns made in the southeastern state of Tamil Naduin India. They are customarily made by women on the thresholds of their homes (Fig.2). Kolam making is a highly skilled practice, taught to girls from a young age by theirmothers or other female relatives. Traditional kolam, called kampi kolam, are abstract,but many contemporary patterns include representational or figurative elements.
Fig. 2. Woman making a kolam pattern from rice powder (retrieved from [17]).Kolam are made daily, early in the morning. After they are completed, they aremeant to be walked on so that by the end of the day they are worn away. Patterns arecreated with finely ground rice powder that is trickled in a thin stream from betweenthe fingers onto the ground. First, a regularly spaced grid of dots or pulli is laid out onthe ground. Then the rice powder is dropped to form a line that loops around the dots,intersecting itself repeatedly. Some kolam patterns are formed with one continuousself-intersecting curve that ends where it begins; others with multiple continuousclosed curves. Patterns generally have reflective or rotational symmetry. Fig. 3illustrates some traditional, abstract kolam patterns.Fig. 3. Some traditional, abstract kolam patternsKolam are rich with aesthetic and cultural significance. They are ritual markers ofthe space between public and private worlds, they pay homage to the goddess of theearth and good luck, and they are tied to various other value and belief systems. Theperformance of kolam making is an important aspect of kolam folklore and isconsidered the locus of creativity [18].Most studies of kolam focus on the completed patterns. Their formal properties, inparticular, have attracted the attention of computer scientists and mathematicians whohave analyzed the patterns in terms of graphs, combinatorics, array grammars, and
L-systems (for example, see [19, 20, 21]. Much less has been written about theexecution of kolam patterns. One notable exception is a study by the anthropologistAmar Mall [18], which describes the improvisational, creative, and social aspects ofkolam making. It is one of few studies that references first-hand reports from womenpractitioners. According to Mall, women may do some preplanning of a pattern,possibly on paper, but the performance is live without a plan in hand. The executionof a kolam patttern may rely to some extent on the memory of a plan, but only as abasic model or prototype to follow. It is otherwise improvisational. The practitionermay change a plan she had in mind on-the-fly to accommodate a mistake ormiscalculation, or to pursue a new idea triggered by the pattern in progress. Thematuration of an initial design concept or intention evolves with its material enactment[18:75]. Renate Dohmen also examines kolam making from a performativeperspective, but with less attention to the specifics of kolam execution and more to itssocio-cultural dimensions [22].Previous formal analyses of the structural and mathematical properties of kolamhave little bearing on the way the patterns are made as continuous lines drawn on theground. These include generative analyses such as the shape grammar I developed forgenerating kolam designs [23]. The calculation of abstract designs with this grammaris very simple. However, the calculation of a kolam in real time, by tracing linesthrough a grid of dots so that they close back on themselves, is another story.Designing a kolam pattern algorithmically is easy. Making one is hard, requiring yearsof practice.In the following three sections, I present three grammars to illustrate somedifferences between generating abstract kolam designs and making them – differencesbetween plans and performances. The first grammar is the shape grammar I definedpreviously. It generates designs by placing motifs in a modular grid, a strategy that haslittle relationship to how kolam are made. The second grammar is also a shapegrammar. It generates designs based on the concept of mirror curves. It begins toparallel the live process of making kolam. The third grammar is a making grammar. Itattempts to capture performative aspects of kolam making. The three grammars arenot illustrated here in full. Rules are given in just enough detail to convey the maingenerative strategies. For example, labels controlling rule applications are omitted.Also not shown are “pre-processing” and “post-processing” stages, including theinitial shape and rules for defining an organizational grid at the beginning of acomputation, and rules for erasing a grid and labels at the end.4.1A Modular Motif Shape GrammarThis shape grammar generates kolam designs by placing different motifs within amodular square grid. Motifs are placed within the grid to create continuous linepatterns with rotational or reflective symmetries. The grammar begins with asymmetrical grid of square cells. The specific reflective or rotational symmetry of thegrid is assumed to be indicated in some fashion, for example, by showing the point ofrotation and/or axes of reflection. Each cell of a grid is marked with a dot denoting apulli. The six rules of the grammar are shown in Fig. 4. Each rule places a differentmotif within a cell of the grid. Each motif touches the boundaries of a cell at one, two,
three, four, or
A key feature of making is time. Craft practices, and making activities in general, are continuous, temporal events. Making sense of and participating in a craft practice not only involve structuring in some way the spatial aspects of the things being made, they also involve structuring the temporal actions involved in the making of those things.
1 IntroductionOn categorial grammars and learnabilityLogical Information Systems (LIS)Categorial grammars and/as LISAnnex Plan 1 Introduction 2 On categorial grammars and learnability 3 Logical Information Systems (LIS) 4 Categorial grammars and/as LIS 5 Annex Annie Foret IRISA & Univ. Ren
Extending Lambek Grammars to Basic Categorial Grammars Wojciech Buszkowski Faculty of Mathematics and Computer Science Adam Mickiewicz University Poznan Poland Journal of Logic, Language and Information 5.3-4: 279–295, 1996 Abstract Pentus
formal grammars and the devices that are used to carry them out by briefly comparing Context-Free Phrase Structure Grammars with the framework object of our studying, Categorial Grammars (CG)1. Phrase Structure Grammars. In formal language theory a language is defined as a set of st
Part A: Art, craft and design education in schools and colleges 6. Achievement in art, craft and design 7 Teaching in art, craft and design 14 The curriculum in art, craft and design 25 Leadership and management in art, craft and design 33 Part B: Making a mark on the individual and institution 39.
Context-Free Grammars another grammar with precedence and associativity – and – are left-associative operators in mathematics –* and / have higher precedence than and – Grammar G 1 Source: Tucker & Noonan (2007) Context-Free Grammars parse tree for 4**2**3 5 * 6 7 Source: Tucker & Noonan (2007) Context-Free Grammars
Non-Context Free Languages Portions c 2000 Rance Cleaveland c 2004 James Riely Automata Theory and Formal Grammars: Lecture 7 – p.1/48 - Non-Context Free Languages Last Time Context-free grammars and languages Closure properties of CFLs Relating regular languages and CFLs Today
Inversion Transduction Grammars [Wu 1997] Head Transducer Grammars [Alshawi et al. 2000] Tree-to-string models [Yamada and Knight 2001], [Galley et al. 2004] Loosely tree-based model [Gildea 2003] Multi-Text Grammars [Melamed 2003] Hierarchical phrase-based models [Chiang 2005] Giorgio Satta Grammar-Based MT
LTGs are equal to Linear Inversion Transduction Grammars (Saers et al., 2010). To be able to in-duce transduction grammars directly from par-allel corpora an approximate search for parses is needed. Thetrade-off betweenspeed andend-to-end translation quality is investigated and com-pared to Inversion Transduction Grammars (Wu, 1997) and the .
