7.3 Hands
Hands are the most prominent articulators of Sign Language, and
have many degrees of freedom and articulatory possibilities. They are
represented in SignWriting with complex graphemes which encode in
their graphical attributes the different features of the hand. In the
corpus they are assigned the CLASS=HAND
, and all four
other tags are used: SHAPE
, VAR
,
ROT
and REF
.
7.3.1 SHAPE
The first feature of the hand is its “shape” or configuration: how
the fingers are bent and placed to form a unique shape that acts as a
unit. Graphically, the SHAPE
tag is roughly the outline
of the grapheme, mainly the strokes representing the fingers. Fingers
are very flexible, so there are a great many possible configurations
that the hand can adopt, and SignWriting strives to provide symbols
for every one of them. However, not all of them are in use in every
sign language. There is also allophonic variation, meaning some
different finger configurations are perceived to be the same hand
shape for native signers, and so there can be vacillation and
inconsistencies in their transcription.
In this corpus, hand shapes are labelled according to the phonology
of Spanish Sign Language, not phonetically, so some different symbols
are tagged with the same SHAPE
. The phonological base for
our labeling is to be published in a forthcoming article. Since there
is not a standard notation for hand shapes across languages, we use
our own ASCII-based notation which is derived from the previously
mentioned phonology. For users not interested in the underlying
linguistic theory, these labels can be assumed to be arbitrary strings
uniquely identifying the different configurations. Currently, 72
different hand SHAPE
s can be found in the corpus.
7.3.2 VAR
Apart from finger configuration, hands can rotate in the three dimensions of space, which complicates their transcription in the flat page. SignWriting uses a combination of graphical features to represent hand orientation, encoded in the remaining tags for hands in this corpus.
The first of them is the VAR
iation. Graphically, it
encodes the “alteration” of the basic shapes encoded in the previous
label. This variation can happen in two ways. First, the body of the
hand can be filled with different patterns of black and white. White
represents the palm, and black represents the back of the hand, as
viewed from the point of view of the signer. The fingers can also be
detached, meaning the orientation is horizontal.
Fill variation is encoded with the letters w
(white),
b
(black), and h
(half). Finger detachment
is encoded by prepending the letter h
(horizontal) to the
tag. This gives six possible VAR
tags:
w |
| b |
| h |
|
hw |
| hb |
| hh |
|
There is also the possibility that a hand grapheme has a “black
left” and “white right” fill pattern. This is encoded as
h
or hh
, and treated as a graphical
reflection (see below for the REF
tag).
7.3.3 ROT
To complete the graphical representation of hand orientation,
HAND
graphemes can also be rotated around their center.
This rotation is not continuous but rather has 8 possible values,
encoded in this corpus using the notation for the cardinal directions.
The hand is considered to be pointing along is distal axis, that is,
the straight line from the forearm to the fingertips when they are
fully extended.
N |
| NE |
| E |
| SE |
|
NW |
| W |
| SW |
| S |
|
N |
| NE |
| E |
| SE |
|
NW |
| W |
| SW |
| S |
|
7.3.4 REF
As a last transformation, HAND
graphemes can appear
“mirrored” in SignWriting. Mirroring of a grapheme is not reflective
of any one phonological feature, but rather a graphical attribute that
can be used to convey different meanings. For example, right and left
hands are mirror images, so the corresponding graphemes can be
mirrored to better identify each of them. “Black”
VAR
iants are also often mirrored, to better iconically
depict the hand as it would be seen by the signer.
Therefore, the meaning of reflecting a grapheme has to be extracted from the context, and can not be deduced from the isolated grapheme at all. This also means that there is not a phonological criterium to decide on a “normal” form of a grapheme, so the criteria chosen in this corpus may seem arbitrary. However, they are chosen to maximize graphical homogeneity and predictability, which can help in the computational treatment of SignWriting.
Reflection is codified in the REF
tag, which can take
the values n
(not reflected) or y
(“yes”,
reflected). To decide whether an instance is reflected, the following
algorithm is used:
- Always, reflection must be decided from “
N
orth’ rotation. If a grapheme is rotated, it must first be (mentally) set upright. - If the
VAR
ish
orhh
, the variant with the black on the right isn
, and the one with black on the left isy
. - If the
VAR
isw
orb
, attention must be paid to the fingers. If they are in the same position as the unreflectedh
VAR
iant, then they are themselvesn
ot reflected. In other words, theh
VAR
iant decides, and theb
andw
ones copy it. - If the
w
ob
VAR
iant is not identical to theh
one, attention is paid to the flexion of the fingers. If they bend to the left, theREF
isn
, otherwise it isy
. In the case of the single little fingerSHAPE
, where the finger bends to one side but curls to the other, then
ot reflected grapheme is that where the finger is to the far left (w
hite right hand). This step can also help decide theREF
for otherw
orb
graphemes without having to look up theh
version. - Horizontal
VAR
iants follow the same pattern as vertical ones.
The algorithm above is also important because when dealing with
handwritten SignWriting, such as this corpus does, there can often
appear “non-normative” uses of graphemes which are however
understandable and need to be annotated. In any case, since pictures
convey graphical information better than words, the following tables
present some examples of REF
tags.
n |
| n |
| n |
|
y |
| y |
| y |
|
n |
| n |
| n |
| n |
|
y |
| y |
| y |
| y |
|
| | | | | | | |
| | | | | | | |
| | | | | | | |
7.3.5 Ambiguous graphemes
Sometimes grapheme SHAPE
s are symmetric, meaning that
the REF
lected versions end up being graphically
identical. In this case, REF
is always taken to be
n
. In a few cases, rotation can also be ambiguous (for
example the closed fist, which is a square). In this case, the first
possible ROT
in this sequence is chosen:
N
\(\to\)NE
\(\to\)E
\(\to\)SE