Content uploaded by Onno Crasborn
Author content
All content in this area was uploaded by Onno Crasborn on Feb 20, 2024
Content may be subject to copyright.
Global Signbank manual
Version 3
Radboud University, Centre for Language Studies
Onno Crasborn, Inge Zwitserlood, Els van der Kooij & Ellen Ormel
February 2024
Acknowledgments
The creation of Global Signbank and the writing of this manual has been a long-term effort.
The authors would like to thank the many deaf and hearing colleagues that have contributed
to its conception and production, and SUN Qian for designing the Signbank logo. We
gratefully acknowledge the funding of the Dutch Research Council NWO (projects
277-70-014, 276-70-012, 380-70-008, 360-70-500), NIH (grant R01 DC-013578), and NTNU
(Norwegian Sign Language Signbank Project).
1
Contents
1. Introduction...................................................................................................................... 4
2. Using Global Signbank in Combination with ELAN....................................................4
2.1. ECV........................................................................................................................................................4
2.2. Lexicon service................................................................................................................................ 5
3. Before adding a new sign to the lexicon......................................................................... 5
4. Lemmatisation guidelines................................................................................................5
5. Basic properties................................................................................................................ 7
5.1. Header fields.....................................................................................................................................7
5.2. Word class (Part of speech)........................................................................................................7
5.3. Tags.......................................................................................................................................................7
5.4. Notes.................................................................................................................................................... 8
6. Glosses............................................................................................................................... 8
6.1. Annotation ID-gloss in Dutch & Annotation ID-Gloss in English: glosses used on
the gloss tiers in ELAN................................................................................................................................ 8
Determining the gloss.............................................................................................................................9
6.2. Lemma ID Gloss: the gloss that unites Annotation ID Glosses into a single
lexical item....................................................................................................................................................... 9
6.3. Non-lexical signs............................................................................................................................. 9
7. Morphology.....................................................................................................................10
7.1. Sequential morphology (compounding)............................................................................10
7.2. Simultaneous morphology.......................................................................................................10
1.7.2. Simultaneous phonetically bound compounds..................................................... 10
2.7.2. Blends......................................................................................................................................11
8. Form description............................................................................................................ 11
8.1. Handedness.................................................................................................................................... 12
8.2. WD and WP.....................................................................................................................................13
8.3. Handshape...................................................................................................................................... 13
8.4......................................................................................................................................................................15
8.5. Letters and numbers.................................................................................................................. 15
8.6. Handshape change.......................................................................................................................15
8.7. Relation between articulators................................................................................................ 16
8.8. Location............................................................................................................................................17
8.9. Virtual object..................................................................................................................................19
8.10. Contact type...............................................................................................................................20
8.11. Movement direction...............................................................................................................20
2
8.12. Movement shape......................................................................................................................21
8.13. Relative orientation................................................................................................................21
8.14. Orientation change................................................................................................................. 23
8.15. Repeated movement.............................................................................................................. 24
8.16. Alternating movement.......................................................................................................... 24
8.17. Phonology other...................................................................................................................... 24
8.18. Fixed combinations (Also called ‘compounds’): no phonological description..
24
8.19. Mouth........................................................................................................................................... 24
8.20. Phonetic variation...................................................................................................................24
8.21. Combinations of values in a single field........................................................................ 25
9. Semantics........................................................................................................................ 26
9.1. Iconic image....................................................................................................................................26
9.2. Named entity..................................................................................................................................26
9.3. Semantic field................................................................................................................................ 26
10. Minimal pairs............................................................................................................. 27
11. Relations to other signs.............................................................................................. 27
12. Relations to foreign signs...........................................................................................27
13. Notes............................................................................................................................ 27
14. Importing data............................................................................................................27
14.1. Batch import of videos and images................................................................................. 27
14.2. Batch creation of new lemma and sign entries.......................................................... 28
14.3. Batch changes to existing signs.........................................................................................28
14.4. Batch changes to existing lemmas................................................................................... 28
15. Users, Permissions and Access Rights...................................................................... 28
16. References................................................................................................................... 29
3
1. Introduction
Global Signbank is a lexical database that stores datasets documenting various sign
languages. This document functions as a guide on how to add new signs to Global Signbank,
and explains how the various fields in the database should be interpreted and used. Some of
the functionality in the user interface is also explained. The text was written based on the
NGT dataset in Global Signbank, and all examples are ID glosses from that dataset, which is
publicly accessible through the Public View. Registered users who are logged in will see more
information about each sign. Access to the NGT dataset (or other datasets that have been
set to ‘published’ in the backend) as a researcher can be requested through the Datasets
menu interface.
In the area of phonology, additional values for certain fields can be added based on
linguistic differences with NGT. For some non-phonology fields, not much attention has yet
been paid to the possible values (like for Semantic field), and revision or enrichment may be
needed based on a more well-developed view of these properties.
2. Before adding a new sign to a dataset
It is tempting to keep adding new signs to a dataset during corpus annotation, but when a
dataset contains a few thousand signs, many forms will already be there under a different
gloss than expected or with different translations than the one that is appropriate for the
current corpus context. So before adding a new sign to a dataset, make sure that it is really
not yet in Signbank:
1. Obviously, first check the already existing Annotation ID Glosses;
2. Then, use the field Translation equivalents to find existing signs by their semantics,
thus checking whether a form is not already present under a different translation or
with a related meaning.
3. If the sign can be translated in multiple ways, but you can’t find the most obvious
translation(s) using the Annotation ID Gloss field or the Translation equivalents field,
search existing items by their phonological properties.
Feel free to be pragmatic about including glosses in the dataset for manual activities that
may not strictly-speaking be words or word forms: if it facilitates your corpus annotation (by
having these glosses export to the ECV, and appearing in the pop-up list in your gloss-tiers in
ELAN), include these possibly non-lexical constructions or morphemes in Signbank as a sign.
For the NGT dataset, we did just that: we included various types of pointing signs,
not all of which may be lexicalised forms (like PT:up for pointing upwards), and grouped
them under a single Lemma ID Gloss PT.
Similarly, we added glosses for the classifier predicate constructions by combining
the predicate root (like BE, AT, PIVOT) with the handshape, linked by a ‘+’; these in turn are
grouped under the lemmas BE, AT, and PIVOT, respectively. So for instance, there’s an
Annotation ID Gloss BE+B for a flat hand classifier construction without movement, that
could be used to mean ‘there’s a car positioned at that location in space’. See the Corpus
NGT annotation guidelines (Crasborn et al. 2020) for more information.
4
And finally, we’ve added all hand alphabet letters of NGT as signs (#A, #B, et cetera),
as in corpus dialogues they are often combined with a mouthing to express a concept.
By not checking the ‘In dictionary’ field, these glosses can be hidden from public
display, but still facilitate the glossing workflow in ELAN.
3. Lemmatisation guidelines
Global Signbank promotes a strict form-based classification of signs: each different
phonological form gets its own Annotation ID Gloss; when several sign forms are
phonological or morphological variations of the same head word, they can be grouped
under a shared Lemma ID Gloss. Different morphological forms of a sign can thus be added
as separate Annotation ID Glosses and linked by using the same Lemma ID Gloss entry.
In the NGT dataset (and in annotating the Corpus NGT), we have chosen to be very
conservative here: not every possible verb inflection receives its own entry based on
differences in the location or movement in signing space. We do distinguish first person
forms with contact to the body, as ‘± contact’ is a phonological form distinction that we can
encode in the phonological description of an Annotation ID Gloss in Global Signbank. Glosses
for such forms have suffix :1 (e.g. AAIEN:1/STROKE:1, ROEPEN:1/CALL:1).
Different word classes (N, V, A, …) with the same phonological form are likewise
documented under a single gloss, and then assigned multiple word classes in the corpus
annotations: it is debatable whether in NGT there are indeed different words for ‘sit’ and
‘chair’, and thus we include ‘sit’ and ‘chair’ as different senses for ZITTEN-A/SIT-A, for
instance. The actual Dutch/English word is not crucial, as it’s just an ID label: it could equally
well be ZITTEN-STOEL-A/SIT-CHAIR-A, if that is easier to ‘read’ in annotation files, for
instance, or less confusing to annotators. (Note that in other sign languages there may well
be systematic form differences distinguishing nouns from verbs, like repetition of
movement. That could then be a motivation for creating two distinct Annotation ID Glosses.
But if it is a morphological alternation with predictable forms, it could equally well be argued
that it’s better to include them under a single gloss.)
Inherently reciprocal signs are assigned a Dutch and an English gloss like other signs
(DISCUSSIEREN/DISCUSS). Signs that also have a non-reciprocal variant (AANVALLEN/ATTACK
vs. AANVALLEN/ATTACK-reciprocal get the suffix –ELKAAR/–EACHOTHER).
GEVEN-ELKAAR/GIVE-EACHOTHER could thus be distinguished in ELAN annotations from
‘give towards two locations’. They should be assigned an identical Lemma ID Gloss.
In case there’s already two annotation glosses that are really one lemma, follow the
following (sometimes conflicting) rules for selecting the unified gloss:
- Choose the most generic or high-level concept (FACT vs. PROOF)
- Choose the gloss that’s closest to the iconic image (ELECTRICITY vs. THUNDER)
- Choose the specific gloss when this is in high-frequent use (EXPERIENCE-D vs.
EXPERT)
- Choose a Dutch noun, verb, or adjective/adverb rather than an ejection (MWOH or
MWA vs. TWIJFEL/DOUBT)
- Mouthing and emotional facial expression should be ignored
(HART-BONST/HEART-BEATS vs. BANG/AFRAID)
5
- Choose a Dutch word rather than a neologism even if the meaning is not perfectly
matching (LEF/HAVE-GUTS vs. QUAZ/QUAZ).
At the time of writing this version of the manual, Annotation ID Glosses have only one string
of translation equivalents, that are not grouped in different senses; this will be made
possible in the course of 2023. The present state has encouraged us to include ‘double’
entries for a single phonological form when two possible senses of a sign are very distinct:
signs like BATTERY and DOCTOR-A are now two different entries, even though the
phonological form is identical. They are marked as ‘homonyms’ in the panel Relations
between Signs, and Signbank can generate a list of them through the menu Analysis >
Homonym List. We have yet to decide what to do with such homonyms when signs can have
multiple senses: should they be merged under a single gloss with two senses, or stay
separate entries? We could choose to merge the above examples under a single Annotation
ID Gloss BATTERY-DOCTOR, say, and list two different senses of that form, or we could keep
them separate because the senses are not at all related.
Each dataset manager can make different decisions here, of course. For future
cross-linguistic study of datasets, we recommend following the NGT practices.
Figure 1. Some possible relations between glosses
A recurring point of discussion is the lexicalisation of classifier constructions: many forms in
spontaneous signing can be analysed as a classifier construction. Only when the form is
stable (recurrent) enough and has a meaning that is not fully predictable from its parts, it
would be entered into a dictionary. Thus, the NGT sign THEE/TEA can be fully analysed as a
classifier construction in terms of its form, but it is commonly used with a meaning that has
specialised from ‘repeatedly dipping something small in a cylindrical container’ to ‘tea’. For
that reason, it is listed in the lexicon. Not all cases are as clear as that, and intuitions from
more than one signer will be helpful.
6
4. Basic properties
4.1. Header fields
Each sign must receive at least a Dutch and English Annotation ID Gloss. The Lemma ID Gloss
will initially be identical to the Dutch Annotation ID Gloss. Section 5 is devoted to glossing.
Translation equivalents in Dutch contains a list of possible translation equivalents,
separated by commas. Inflections (of verbs and nouns, and adjectives) are not listed as
separate entries.
Annotation instructions contains further notices to annotators that can help them in the
selection of the right gloss.
To facilitate further processing of the sign, we use Tags (section 4.3). These largely refer
to workflow issues, but we also use a tag ‘Radboud sign’ that are often locally invented signs
that interpreters will want to learn before doing an assignment in our group. The need for a
description of the phonology can be observed by the absence of a value for the first
phonology field (Handedness).
4.2. Word class (Part of speech)
Although some signs can only be used in a particular grammatical function (e.g. referential),
many signs have more possibilities. For each sign, the function(s) (part of speech) for which
it may be used can be specified, with the following abbreviations:
●Noun for referential use, i.e. when it labels an entity (concrete or abstract) and can
be predicated over (e.g. BOEK/BOOK). For concrete concepts, as a test, if sign X can
be bought or looked at, it can function as an N.
●Verb for expression of an action, process, or state, i.e. when it labels an event in
which one or more entities are involved. (e.g. KOPEN-A/BUY-A).
●Noun or verb for signs that are ambiguous or that can be used for both.
●Adjective for modification of nouns
●Interjection for signs expressing a comment or feedback, that cannot be used with a
Noun or Verb (e.g. QUATZ)
●Particle
4.3. Tags
Some codified information is marked by the use of a limited number of tags. These refer to
workflow issues for editors. We currently have the following options:
●Add phonology
●Check morphology
●Check phonology
●Check semantics
●Picture: wrong frame
●Picture: wrong sign
●Sign: approved
●Sign: proposed
●Sign: reconsider gloss
7
●Sign: remove
●Video: missing
●Video: refilm
●Video: tech problem
●Video: wrong sign
The tag ‘Sign: to discuss’ merits specific attention, as it is typically used for
lemmatisation purposes. It can mean that a similar form already exists in the dataset, or
that the form is perhaps not lexicalised but should rather be considered a classifier
construction without a lexicalised (more specific) meaning.
The tag ‘Sign: proposed’ currently has an alternative under the panel Publication status: the
checkmark for ‘Proposed new sign’. This redundancy will be removed in the future. Feedback
on what works best is welcome. The colour-coding of rows in the list views will need to be
adapted accordingly.
4.4. Notes
Any remarks about the sign. We currently have the following types:
●Note
●Private note
●Phonology
●To do
●Suggestion for other gloss
●Discuss between annotators
●Etymology
This mixed bag partly refers to workflow issues. In contrast to tags, notes require some
further explanation. Others may be specific for a certain research project. It would be good
to evaluate their use after some time.
5. Glosses
5.1. Annotation ID-gloss in Dutch & Annotation ID-Gloss in English: glosses used
on the gloss tiers in ELAN
Manual signs that enter the lexical database receive a unique Dutch gloss and a unique
English translation. The English translations need not necessarily have a one to one
correspondence to the Dutch glosses and their variant suffixes (-A, -B, etc.), in the sense that
they need to reflect the meaning of the sign. For example, the Dutch gloss WEG with suffixes
-A, -B, and -C have three different English translations as ROAD, AWAY, and GONE. The three
homonyms ‘weg’ in Dutch thus each have three very different translations. This also works
the other way around, so always check for already used English glosses to see if you opt for
another English gloss or to add –A/B’s.
8
Determining the gloss
As also outlined in the annotation conventions of the Corpus NGT (Crasborn et al. 2020), we
use the following rules:
●In general, a gloss is the Dutch (and English) word or phrase that covers the most
general meaning of the sign.
●When a gloss needs more than one word, these are separated with hyphens (e.g.
BEGRIJP-ER-NIKS-VAN/MISS-POINT)
●When there is more than one sign for which a particular gloss is appropriate, they are
distinguished by letters (e.g. OUDERS-A, OUDERS-B/PARENTS-A, PARENTS-B)
5.2. Lemma ID Gloss: the gloss that unites Annotation ID Glosses into a single
lexical item
Some lexical signs are only specified for a subset of phonological features, lacking e.g.
location and/or articulator features. When combined with a morpheme expressed by such a
feature, it forms a lexical sign with a related, somewhat more specific meaning. The signs
with a more specific meaning get an Annotation ID Gloss, and share the same Lemma ID
Gloss. For example, the entries with the annotation glosses KEIZERSNEDE (CAESARIAN) and
BORSTOPERATIE (CHEST-OPERATION), sharing all features but the (meaningful) location,
have the Lemma gloss OPERATIE/SURGERY. The entries with the Annotation glosses
INSTEKEN-A/-B/-C/-D (PUT-IN-A/-B/-C/-D), sharing all features except for (meaningful) hand
configuration features have the Lemma gloss INSTEKEN (PUT-IN).
To be revised after writing the Lemmatisation guidelines (section 3)
5.3. Non-lexical signs
Signbank currently contains a few ID-glosses that do not refer to lexical signs and/or do not
have a consistent form, but are labels for manual behaviour we want to be able to annotate
on the gloss tiers in annotation files anyway, and describe in Signbank. Examples are:
Dutch gloss
English gloss
Usage instruction
imitatie
imitation
Label for all kinds of cited manual behaviour. No fixed
form. Specify on meaning tier.
self-touch
self-touch
self-touching action that is not meant to be part of the
message that is expressed, e.g. rubbing nose, scratching
arm, etc.
All hand alphabet
letters, prefixed by
#
All hand alphabet
letters, prefixed
by #
By including these items in Signbank, they appear in the ECV in ELAN, facilitating the
entry of glosses. They are not described in terms of phonology or semantics.
9
In addition, the NGT dataset in Signbank contains Annotation ID Glosses for the root
and handshape in classifier constructions, such as MOVE+T, BE+1, AT+money, and
PIVOT+S, where the first part is the root of the construction expressing movement or
position, and the part after the + character is the handshape. A similar solution has been
chosen for size and shape specifiers, by using annotation glosses like SHAPE+B.
6. Morphology
This panel contains three tables that allow the specification of two types of morphological
information about the sign: sequential and simultaneous.
6.1. Sequential morphology (compounding)
One can indicate here whether a sign is sequentially composed of two or more lexical signs.
The sequential units are marked as the first, second etc. element in the compound by
selecting the relevant qualifier in the list below.
●N/A
●Compound part 1
●Compound part 2
●Compound part 3
●Compound part 4
Also, they are linked to the lexical entries for those signs. If present, they can be selected
from a list of ID-glosses that is generated by Signbank. If a composing sign has not yet been
entered in Signbank, one first needs to propose a new ID-gloss for it, and subsequently
select it from the list.
A complex form does not receive a phonological description; the form of the composing
signs is specified in their respective Signbank entries. After selecting a compound part from
the following list, a reference must be made to the annotation ID gloss for the composing
part. After confirming this list, the compound will automatically appear in the morphology
panel of the composing part in a table called ‘Appears in’.
6.2. Simultaneous morphology
1.6.2. Simultaneous phonetically bound compounds
When signs are composed from morphemes (or FMUs, ‘form-meaning units’; see van der
Kooij, Zwitserlood & Crasborn, 2023; Zwitserlood, van der Kooij & Crasborn, 2023) below the
level of the sign, e.g. number of selected fingers, meaningful locations, etc., this is indicated
here. The morphemes are selected from a list that can be maintained in a separate area of
Signbank called ‘morphemes’. The abstract meaning of the morpheme is automatically
provided, but one needs to additionally specify the meaning of the morpheme in that
particular sign. The signs in which FMUs occur also receive a full phonological description.
10
2.6.2. Blends
Here, it is possible to indicate whether a lexical sign consists of more than one sign that are
combined to form a single syllable, as in the NGT sign WEEKEND-A. Both compositing signs
must be recognizable in form and meaning. Like for the sequential compounds, it is possible
to select the signs from which the blend is composed. In contrast to the sequential
compounds, a phonological description is provided here. Note that, since many signs are
root compounds, it is sometimes difficult to decide whether a single-syllable complex sign is
a root compound or a blend.
7. Form description (phonology)
This set of fields describes the form of the entries, i.e. of the citation forms of the
Annotation ID-glosses. As stated earlier, the basis for the distinctions in this panel are taken
from the Dependency Model of sign phonology (van der Kooij 2002, van der Hulst & van der
Kooij 2021), but also contain some surface-level properties (Crasborn & van der Kooij 2023;
van der Kooij, Zwitserlood & Crasborn, under review). The form descriptions in Global
Signbank are ultimately a compromise between phonological specifications of the entries,
their morphological build-up, and user-friendly search possibilities.
First, phonology does not specify form information that is predictable from human
articulatory possibilities and/or other form elements within a sign. However, it is convenient
to be able to search for both specified and non-specified form elements, such as neutral
space, in a database. Second, the large numbers of morphologically complex (monosyllabic)
signs in NGT necessitate an additional set of form elements to the set of pure phonological
features. Thus, while monomorphemic signs only need a single (relative) orientation feature
value, morphologically complex signs often need two. Third, like in most other databases of
signs, we have opted to use ‘handshape’ as shorthand for sets of articulator features, since
most sign language users are used to this parameter.1Finally, the form component of
Signbank has been constructed on the basis of phonetic form information with single
parameter value sets. Although we have adapted this component to some extent, it
appeared not to be possible to apply serious changes, such as allowing form descriptions of
bisyllabic signs in the time span and budget we were allowed.
Sections 7.1 to 7.19, following the panel for form description in Signbank, explain our
approach to each element and the solutions to exceptional forms. All values for each field
are prespecified and can be selected from drop-down lists to avoid typos.
1The phonological features of each handshape are provided in Signbank [Search>search handshape], but it
is not possible to search on them.
11
Figure 1 Form description panel in Signbank
7.1. Handedness
The following values are available to specify the number of hands in a sign:
●1 One-handed sign. Signs that are made with one hand only (e.g.
BAAS-A/BOSS-A). This includes signs that are made on the arm, up to the wrist (e.g.
TIJD-A/TIME-A).
●2a Two-handed asymmetrical sign. These are signs in which the weak hand
functions as the location (e.g. ACCEPTEREN-C/APPROVE).
●2s Two-handed symmetrical sign: both hands move and have the same
handshape (e.g. AANDACHT-A/ATTENTION-A). (Note that the hands may be a
non-identical locations , as in ZIEK-C/SICK-C, although this occurs only occasionally)
●2t Two-handed non-symmetrical sign in which the hands move in tandem: both
hands move, but they are not mirrored in the sagittal plane and the handshapes
can be different. In these signs, the handshapes may differ, and the hands typically
move in contact with each other (BEGELEIDEN-A/GUIDE-A).
12
●X Handedness not applicable. There can be several reasons to choose this
value, e.g.:
o Signs that are made with only non-manual articulators
o Generic glosses like [NAAM]/[NAME] (used for anonymising EAF files in the
Corpus NGT), or self-touch/self-touch, used for annotating meaningful
self-touching behaviour
o All classifier predicate annotations (see section 3 above and the Corpus NGT
annotation guidelines; Crasborn et al. 2020)
7.2. WD and WP
Determining the number of hands in a sign is not always straightforward, especially when it
is encountered in a corpus of (informal) data. Many two-handed signs can be articulated
with one hand (weak drop; WD), such as AL/ALREADY (but not KRUIWAGEN/
WHEELBARROW), and vice versa (weak prop; WP), e.g. HOEVEEL/HOW-MUCH (but not
ZON-A/SUN-A). Signbank provides the option to indicate whether a sign allows variation in
handedness or not, using the values ‘true’ of ‘false’ (‘neutral’ is the Signbank default). Note
that Signbank does not automatically exclude impossible values, like WD ‘true’ with
one-handed signs.
7.3. Handshape
In the fields Weak Hand and Strong Hand, the handshape(s) of signs is indicated in terms of
a short code. There is a field for the strong hand, that should be filled in for almost all signs,
and one for the weak hand, that is left empty for one-handed signs. Strong and weak hand
in symmetrical signs (Handedness = 2s) always receive the same value.
In terms of phonological analysis, handshapes as a unit do not exist: a sign has an
active articulator, which consists of the selected fingers and their configuration. Each
handshape that can be selected for Weak Hand and Strong Hand is spelled out in terms of its
distinctive features (and some other redundant properties, such as the number of selected
fingers independent of which they are) in a separate database, which can be accessed
through the menu Signs > Search Handshape (url
https://signbank.cls.ru.nl/dictionary/handshapes/), or by clicking on the handshape value
for a sign that has already been documented.
As stated in section 3, predictable articulation variation should not lead to new
entries. For example, we do not distinguish separate signs based on predictably differing
positions of the thumb or bending of the fingers at the MCP joints. Such variations are taken
together, as can be seen in the table of handshapes at the above url. If a variation is not
predictable or in case a sign has a strikingly deviating handshape, this can be indicated in the
field Phonetic Variation.
In case of a handshape change, only the initial handshape is described in the
Handshape strong and/or Handshape weak field (e.g. O in EMAIL-B, S in
SOMS-C/SOMETIMES-C), except for sequences of two unrelated handshapes in one sign (e.g.
TELEVISIE-A/TELEVISION-A. In such cases, two values, separated by a ‘>’ character, are used
(e.g. T>V). (See also section 7.5 on Handshape Change.)
13
Currently, the list of handshape values contains 96 items, including the idiosyncratic
combinations like B>L. used in one fingerspelling-derived NGT sign. The names are currently
partly based on Dutch conventions developed in the 1980s. An up to date list can be found
in Global Signbank by searching for a handshape (menu Signs, Search Handshape). Note that
dataset managers can select which values to show and hide for fields like these, so that only
values that are pertinent to their sign language are available to editors.
7.4. Letters and numbers
For (re)search purposes, it might be handy to have an easy way to find signs containing
elements of the manual alphabet or numeral handshapes. Therefore, Signbank offers the
possibility for adding a checkmark to that effect for such handshapes, to the right of the
handshape value.
Initialized signs contain handshapes that refer to a manual form of the letter that
corresponds to a translation equivalent, as BELGIE-B/BELGIUM-B and FAMILIE-A/FAMILY-A
(#F). Also, name signs often include a manual letter (e.g. #O in ONNO-CRASBORN). Note that
it is not always clear whether the handshape of a particular sign simply coincides with or
represents a manual letter corresponding to a translation equivalent. For example, in
AANPASSEN/ADAPT, the handshape of the strong hand may refer to the letter A, but it may
also refer to one entity represented by the thumb moving towards another entity
represented by the thumb – or it may simply be coincidental (being a random, unmotivated,
phonological form), of course.
Likewise, handshapes that are numerals, as in 6-B, can receive a checkmark for
‘number’.
In cases where it is not clear whether the handshape represents a letter or a
numeral, signs do not get checkmarks for ‘letter’ or ‘number’.
7.5. Handshape change
Handshape changes apply to the strong hand in ‘1’ and ‘2a’ signs and to both hands for ‘2s’
signs.
There is a limited number of options, that apply to the initial handshape and describe
the transition to the end shape:
●opening is used when the aperture relation between the
selected fingers and thumb changes from close to open (e.g.
GOOIEN-A/THROW-A)
●closing is used when the aperture relation between the selected
fingers and thumb changes from open to close (e.g.
ACCEPTEREN-A/ACCEPT-A)
●curvingcan be used when the selected fingers flex at all (distal) joints
(e.g. WC-A/TOILET-A)
●bending the selected fingers flex at the proximal joints only (e.g.
EZEL-B/DONKEY-B). NOTE: This value is deprecated, and will be
14
deleted in future versions. Bending is analysed as a (very short) path
movement (esp. on signs on the body), or a pivot orientation change.
●spreading the space between the selected fingers increases (e.g.
VRIEZEN/FREEZE). NOTE: This value is deprecated, and will be deleted
in future versions. Spreading is analysed as a side-effect of opening
movement from a closed handshape.
●unspreading the selected fingers (excluding the thumb) approach or
contact each other at the lateral side (e.g. KNIPPEN/CUT-A,
STOREN/DISTURB)
●wiggling alternated bending of the selected fingers (e.g.
REKENEN-A/CALCULATE-A)
●rubbing thumb and selected fingers rub (e.g. GELD/MONEY)
Other, exceptional values (such as some finger movements used in the Groningen counting
system) can either be entered in the field “phonology other”, or added as a specific
transition from handshape A to B by adding the value ‘A>B’ to the list of handshapes.
The Handshape strong/weak and the Handshape change field should be interpreted
in tandem: a handshape that opens from S (as in GOOIEN-A/THROW-A) will be coded as
Handshape strong “S” and Handshape change “opening”, thus not pinning down the exact
end handshape.
7.6. Relation between articulators
In this field, the position of the articulators in two-handed signs can be specified if they are
not in a default state and cannot be predicted from values for Handedness and Location.
●In most symmetric signs the two hands are next to each other in neutral space
without touching each other. The default relation for symmetric signs is ‘next to each
other’ and is not encoded.
●In asymmetric signs, the precise location value (such as ‘back of weak hand’ in
ZWEDEN/SWEDEN) typically suffices to encode where exactly the strong hand is with
respect to the weak hand; only if this is not apparent (as in
SCHILDPAD-A/TORTOISE-A), the relation can be specified in this field. Alternatively,
the value for Relative Orientation: Location can be used to make this explicit.
Some examples:
●the two hands are in contact and move as a unit, and none of the more specific
values apply (BIJVOORBEELD-B/EXAMPLE-B, BEGELEIDEN-A/GUIDE-A). Note that
contact is not further specified here.
●the fingers are interlocked (CONTACT-MAKEN-A/MAKE-CONTACT-A)
●the fingers are interwoven (AMERIKA/USA)
●the hands cross (GEVANGENIS/PRISON)
●the hands are in a front/back relation (DIRECTEUR-A/DIRECTOR-A)
●the hands are in an above/below relation (INTEGREREN/INTEGRATE)
●one hand is inside the other (this includes “through”) (TWEETALIG/BILINGUAL)
15
The values front/back and above/below can be used in pairs if it is not clear which hand is
the dominant hand (e.g. DIRECTEUR-A/DIRECTOR-A and INTEGREREN/INTEGRATE) or
separately to characterise the position of the dominant hand with respect to the
non-dominant hand (e.g. above in EDE/EDE and below in SCHILDPAD/TURTLE).
7.7. Location
Locations are categorised in four groups: head, body, extremities, neutral space. In the
‘neutral space’ category, we distinguish two planes (horizontal and parallel) as well as ‘R-loc’
(referential locus). R-loc is used for signs that can be spatially inflected. For signs that can be
inflected for one referential locus, this is R-loc, for those that can be inflected for two
referential loci, this is R-loc>R-loc. In addition, there is a value ‘variable’, to be used for signs
that can be made at different locations on the body, resulting in a predictable meaning. E.g.
ONTSTEKING-A,-B/INFECTION-A,-B can be made at different body locations (ear, lung,
throat, etc.) to express inflammations of those body parts.
A list of locations for each category is presented below:
Head
Body
Extremities
Space
Back of head
(CI-A)
Back
(RUG/BACK)
Arm
(LANGE-MOUW/
LONG-SLEEVE)
Neutral space
(PROCENT-A/
PERCENT-A)
Cheek
(CHOCOLADE/
CHOCOLATE)
Belly
(ZWANGER/
PREGNANT)
Knee
(KNIE/KNEE)
Horizontal plane
(CENTRUM/CENTRE)
Cheekbone
(MAAND-B/
MONTH-B)
Trunk
(TRUI/SWEATER)
Leg
(HOND-C/DOG-C)
Parallel plane
(BEL/BELL)
Chin
(WIE/WHO)
Chest
(COLLEGA/
COLLEAGUE)
Weak hand: thumb
(AANPASSEN/
ADAPT)
R-loc
(AFPAKKEN/TAKE-A
WAY, GEVEN/GIVE,
VINDEN-B/FIND-B)
Ear
(DOOF-A/DEAF-A)
Flank
(TROUWEN-A/
MARRY-A,
EEND/DUCK)
Weak hand: index
finger
(PT:wijs/PT:index)
Variable
(ONTSTEKING-A,-B/I
NFLAMMATION-A/-B
)
Eye
(OPLETTEN-A/
WATCH-A)
Shoulder
(AAP/MONKEY)
Weak hand: middle
finger
(PT:mid)
Face (front of head)
(VERWARD-A/
CONFUSED-A)
Hip
(HANDEN-IN-ZIJ/AR
MS-AKIMBO)
Weak hand: ring
finger
(TROUWEN-B/MARR
Y-B)
Forehead
(ZOMER-A/
SUMMER-A)
Weak hand: pinkie
(PT:pink/PT:pinky)
16
Head
Body
Extremities
Space
Head (top or side of
head)
(TANTE-A/
AUNT-A)
Upper arm
(KAPITEIN/CAPTAIN)
Mouth
(ETEN-A/EAT-A)
Elbow
(GEMEEN/NASTY)
Neck
(DOOD-A/DEAD-A)
Lower arm
(COMPUTER-A)
Nose
(ZUUR/SOUR)
Wrist
(ZIEK-B/SICK-B)
Weak hand: palm
(POLITIEK-A/
POLITICS-A)
Tongue+
(TONG/TONGUE)
Weak hand: back
(NODIG-D/
NECESSARY-D)
Upper lip
(JAM-B/
MARMELADE-B,
SNOR-B/MOUSTACH
E-B)
Weak hand: thumb
side
(BROOD-B/BREAD-B)
Weak hand: pinkie
side
(TUIN-A/GARDEN-A)
Weak hand: tips
(AFRONDEN/
ROUND-UP)
Weak hand: front
(KLEUR-D/COLOUR-D
)
Weak hand: web
space
(VLEES-C/MEAT-C)
If the location of the asymmetric sign is (on) the weak hand, the location of the weak hand
itself in space is not specified. The location on the weak hand must be specified in terms of
the side of the hand, or the ‘web’ between the index finger and thumb.
The planes in neutral space represent virtual realities and their orientation needs to
be inferred from the hand orientation (often ‘palm’) in combination with a ‘contacting’
movement towards the plane, or a circular movement parallel to the plane. The horizontal
plane is suggested for instance by a horizontal circular movement (e.g.
OMGEVING-A/ENVIRONMENT-A), or by a downward movement followed by a hold (e.g.
CENTRUM/CENTRE), similar to an end contact with a body location.
17
Note that description of a location does not necessarily imply that the articulator is
in contact with the location (e.g. ‘cheek’ in BRUIN/BROWN). Location ‘face’ is also used for
‘in front of the face’.
Phonological descriptions in Global Signbank are currently restricted to monosyllabic
forms. For the relatively few disyllabic forms that have two different locations, the values are
represented in one field, separated by the character “>”. The ones attested for NGT are:
●Cheek>chin
●Chest>trunk
●Chin>chest
●Chin>neutral space
●Chin>weak hand: index finger
●Chin>weak hand: palm
●Chin>weak hand: thumb side
●Ear>cheek
●Ear>chest
●Eye>neutral space
●Face>head
●Forehead>chest
●Forehead>chin
●Forehead>neutral space
●Forehead>weak hand: palm
●Head>chest
●Head>neutral space
●Head>shoulder
●Head>weak hand: palm
●Mouth>cheek
●Mouth>chest
●Mouth>chin
●Mouth>weak hand: palm
●Neck>chest
●Neutral space>head
●Neutral space>nose
●Nose>chin
●Nose>neutral space
●Shoulder>shoulder
●Shoulder>weak hand: palm
●Temple>chest
●Weak hand>arm
●Weak hand thumb side>arm
7.8. Virtual object
In this field the object or action that is implied by the shape of the articulator and/or its
movement can be specified. The ‘contacting’ movement of the hand palm (of flat hands, but
also of C hands for instance) or the fingertip(s) (e.g. of 1 hands) can trace the outline of an
entity.
18
e.g. ‘bergen/mountains’ in the sign HEUVEL.PL/HILL.PL
‘bol/globe’ in the sign WERELD-A/WORLD-A
Also, it can indicate the typical location of an entity (e.g. that of a door bell in
AANBELLEN/RING: ‘deurbel/doorbell’).
The location for signs depicting a virtual object can vary. They may be articulated at
body locations, horizontal or vertical planes, and in neutral sign space (see section 7.7 on
Location above).
Signs for interpunction characters typically drawn on a vertical plane (‘vertical sheet
of paper’): AANHALINGSTEKENS/QUOTATION-MARKS, KOMMA/COMMA, PUNT/POINT.
7.9. Contact type
When there is no contact, this field is left empty. The articulator can contact the location in
the following ways:
●Initial contact at the start of the movement (SCHAAP-A/SHEEP-A)
●Final contact at the end of the movement
(INTERNAAT-A/BOARDING-SCHOOL-A)
●Double contact at the start and end of the movement
(JARIG-B/BIRTHDAY-B); this repetition of the movement is not also seen
as an instance of repeated movement (see section 7.14 below).
●Continuous continuous contact during the movement
(LANG-A/LONG-A)
●Brush the contacting hand brushes the location of articulation while
moving (LASTIG-A/TRICKY-A)
Final contact is specified for signs articulated in space that have a final hold, as if ending in
contact with a virtual plane. The location feature ‘horizontal plane’ is specified in these signs
(e.g. CENTRUM/CENTER).
Note: When both hands are in continuous contact and move as a unit (e.g.
BIJVOORBEELD-B/EXAMPLE-B), this is annotated in the field Relation between articulators
(section 7.6 above).
7.10. Movement direction
This field concerns the direction of the movement in space with respect to the signer, or in
asymmetrical signs with respect to an articulator.
Movement direction is predictable in case of initial or final contact with the body; in
those cases, the movement direction is not explicitly described. Movement direction is also
not specified for circular movements (see Movement shape).
As movements are rarely exactly along the three orthogonal axes, some abstraction
from the details of the direction is allowed. Only clearly diagonal movements are
represented by a sequence of values separated by “+”, e.g. contralateral + down. (See Error:
Reference source not found in section Error: Reference source not found above.) Movements
between the signer and a (real or virtual) location, or between two (real or virtual) locations,
are assigned the value ‘To and fro’.
Values for this field include:
●Upwards (RADBOUD/RADBOUD)
19
●Downwards (WACHTEN-A/WAIT-A)
●Forwards (DUWEN/PUSH)
●Backwards (TERUGTREKKEN/RETREAT)
●Contralateral (TERUG/ RETURN, MEE/COME-WITH)
●Ipsilateral (NAAST/ NEXT-TO)
●To and fro (DISCUSSIEREN/ DISPUTE)
●Up and down (AFWEGEN/CONSIDER)
●Ipsilateral and contralateral (AANVALLEN-ELKAAR/ATTACK-EACH OTHER)
●Proximal: for movements on the hands and arms (RING-B/RING)
●Distal (SOEP/SOUP)
●From location (AFPAKKEN/TAKE-AWAY)
●Towards location (RUG/BACK)
7.11. Movement shape
Most of the time shape of the movement path does not need to be described, as it is often
predictable from other phonological values. Straight movements are the consequence of a
movement from a to b. In a similar vein, many arc-shaped paths are the result of the natural
movement of the articulators moving from start to end (OCHTEND-A/MORNING-A), or from
initial contact to final contact (JARIG-B, BIRTHDAY-B or PROFESSOR/PROFESSOR). Arc is only
specified for signs in neutral space. If some movement shape is however needed, it can be
specified using the following values:
●Circle (ROLLEN/ROLL)
●Arc (REGENBOOG/RAINBOW)
●Zigzag (ZOEKEN/SEARCH; DENEMARKEN/DENMARK)
●Straight (Note: only if the movement is indeed straight, as in WEG-A/ROAD)
●Spiral (TIJD-VLIEGT/TIME-FLIES)
●Motivated shape (e.g. in MCDONALDS-B)
●Cross (KATHOLIEK-B/CATHOLIC-B; MAART-C/MARCH-C)
Exceptional values (e.g. the motivated ones, like the M-shape in MCDONALDS-B) can be
further specified in the “phonology other” field.
7.12. Relative orientation
Relative orientation specifies the part of the selected fingers of a handshape that is oriented
towards the end of a path movement, towards the location of the sign, or both. Values can
be ‘radial’ (thumb side), ‘ulnar’ (pinkie side), ‘palm’, ‘back’, ‘finger tips’, ‘base’ (the wrist side
of the hand), and combinations of those; and ‘variable’.
A. Relative orientation: movement is the default field and concerns the side of the selected
finger(s) that is oriented towards the end of the path movement. The value for this side of
the hand fixes one of the degrees of freedom of the 3D object that the fingers are. In the
typical case, the other degree of freedom of orientation is in part predictable from ease of
articulation, and in part variable based on context (coarticulation). E.g. Relative orientation:
movement ‘thumb’ in BRUS-A/SIBLING-A (location: chest) does not need further
specification for the orientation of the finger tips or palm.
20
If a sign has a movement towards a location, only relative orientation: movement is
specified.
Note that the orientation of the extended thumb in the absence of other selected
fingers is similar to that of selected fingers (although the values ‘ulnar’ and ‘radial’ may
seem a bit strange). Thus, in GISTEREN/YESTERDAY, Relative orientation: movement is ‘tips’,
but in RIJLES/DRIVING-LESSON it is ‘radial’.
B. Relative orientation: location is used in the following cases:
●When the orientation cannot be described with respect to the movement direction,
for instance because the sign has a hand-internal movement only (e.g.
KIP-A/CHICKEN-A; location: mouth, relative orientation: location = back)
●When relative orientation: movement only does not provide a sufficient description
of the orientation, e.g. in OUDERS-C/PARENTS-C: location = cheek, relative
orientation: location = palm, relative orientation: location = radial). In case no
location is specified in symmetric two-handed signs, relative orientation: location is
described with respect to the other hand. For example, BESTAAN-A/EXIST-A has
relative orientation: movement = tips and relative orientation: location = palm.
●Signs with circular movements are coded by convention as having a relative
orientation for location with respect to the location plane (e.g. ‘sagittal plane’ in
FIETSEN/BICYCLE, where the orientation value is ‘thumb side’).
●For two-handed signs, the second relative orientation tends to be predictable by the
easiest articulation, but it may be described in terms of the side of the selected
fingers that is oriented towards the other hand. The interpretation of this value is
then partly dependent on the value for ‘Relation between articulators’. For example,
in ZITTEN-A/SIT-A, the articulators are in the default position (side by side), there is a
downward movement, Relative orientation: movement is the palm side, and the
relative orientation with respect to the other hand can be redundantly specified for
thumb side (as there are no minimal pairs in NGT for which the relative orientation is
finger tips in such a configuration).
●Typically, signs related to (one-handed) fingerspelling and counting may have no
movement and location. To describe these, the field relative orientation: location has
the special values ‘Counting’ or ‘Fingerspelling’.
●In a few cases the common set of relative orientation features would predict an
incorrect pronunciation, and/or cause confusion with a sign that has the exact same
features for relative orientation (e.g. GOOIEN-A/THROW-A and BOWLEN/BOWLING).
In such cases, a set of absolute orientation (AO) values in the field relative
orientation: location can be used (e.g. AO: palm-up in BOWLEN/BOWLING).
For signs that do not contain a linear path nor a location, there is no relative orientation (e.g.
SPELEN-C/PLAY-C).
The value ‘front’ is used when the fingers are curved or closed and the relative
orientation would have been ‘Finger tips’ if the fingers had been extended. Values include:
●AO: palm-up
●AO: palm-down
21
●Base
●Finger tips
●Front
●Radial
●Ulnar
●Palm
●Back
●Counting
●Fingerspelling
Relative orientation: front
The relative orientation: location value ‘palm’ in combination with the shape information
from the handshape is interpreted as the tracing of the outline of the virtual object or
delimiting the object when a specific object is specified in ‘Virtual object’.
7.13. Orientation change
A rotation of the selected fingers about any axis is described here by use of the following
values, which are derived from typical articulations of the wrist and forearm illustrated
below. The actual articulation may vary. Similarly, many path movements are often
articulated with an accompanying phonetic orientation change (mostly flexion or extension),
without this being the phonological essence of the movement. Whether or not to specify a
value for this field will thus depend on the phonological analysis of the sign in question. It
may be useful to inspect the existing values from time to time, comparing all signs that have
‘Extension’ specified, for instance; that will help identify signs where extension is a mere
22
by-effect of the hand moving in the direction of the back of the hand (as in NGT
AARDBEI/STRAWBERRY or APART-C/SEPARATE-C).2
●Rotation (repeated supination-pronation; NOG-NIET/NOT-YET)
●Supination (1.ORD/1.ORD)
●Pronation (ERKENNEN-A/ ACKNOWLEDGE-A)
●Extension (OPEN-A/OPEN-A)
●Flexion (JA-B/ YES-B)
●Ulnar flexion (APPARTEMENT-B/APARTMENT-B)
●Radial flexion (KOKEN-B/COOK-B)
The value ‘Rotation’ refers specifically to repeated rotation of the selected fingers about
their length axis (i.e., supination and pronation in turn), not any rotation of the forearm or
rotation about any axis.
In (exceptional) alternating non-repeated signs like OVERLIJDEN/DECEASE, the value
for the dominant hand is specified (supination, in this example).
7.14. Repeated movement
If there is a repetition of the movement in the sign, “yes” is entered in this field. Repetition
may apply to (i) a path movement, (ii) a hand-internal movement or an orientation change,
or a combination of (i) and (ii).
2See Crasborn 2001 for discussion of the relation between orientation changes and path movements.
23
7.15. Alternating movement
If there is an alternating movement in the sign, “yes” is entered in this field. Alternation
concerns:
●out-of-phase movements (e.g. FIETSEN/CYCLE), including alternating hand
dominance (e.g. JEZUS-A/JESUS-A)
●opposed values for hand configuration, orientation, or movement direction (e.g.
BRAND/FIRE)
7.16. Phonology other
Information about the manual or non-manual form that is not covered by one of the other
fields can be entered here.
1.7.16. Fixed combinations (also called ‘compounds’): no phonological
description
NGT has forms in which two separate signs can be recognized in sequence and that have a
specialized meaning. E.g. the combination of OUDER-A/PARENT-A and PARENT-B/PARENT-B
together means “parents”, and the combination of BOEK (BOOK) and BESTEMPELEN-B
(MARK-B) means “passport”. For such combinations, a single ID-gloss is entered in Signbank,
together with a video of the sign combination and reference to the two signs of which it
consists in the Morphology panel. No phonological description of the signs is provided here;
the descriptions can be found in the entries for each sign that take part in the combination.
Any specific deviations from this sequence can be noted in the ‘Phonology Other’ field. See
also under section (Morphology).
7.17. Mouth
Only mouth gestures of any kind (e.g. “blow” with BLAASTEST/ALCOHOL-TEST) are described
here, not mouthings (e.g. not “vader” with VADER-A/FATHER-A).
7.18. Phonetic variation
Recurrent phonetic variations in e.g. handshape, movement, location, or handedness can be
noted in the ‘phonetic variation’ or ‘other phonology’ field. All phonetic variations can be
annotated in the corpus on dedicated tiers. Specific phonetic variation that has been
observed in the corpus can be entered here. Some examples:
●B-bent Articulations with a B-bent handshape have also been observed
(note that this is not a spear
●pinky extension Articulation of a sign with (non-phonologically
specified) extended pinky finger
●non-repeated Non-repeated articulations have also been observed
●repeated Repeated articulations have also been observed
Note: content of this field needs to be revised for the NGT dataset and partly still needs to be
translated to English.
For fixed combinations such as compounds, it is usually not possible to enter the
phonological description of all composing parts. These entries have links to the composing
24
signs (see section 6.2 above), where the phonological descriptions of the component parts
can be found. However, in case of signs that have combined into a single-syllable form, the
phonological description is presented with the entry itself, and the constituting signs are
referred to in the Blend Morphology section.
Some default values are in principle not specified:
●Shape ‘straight’ for path movements
●Movement direction for initial and final contact signs (‘away from’ and ‘towards’,
respectively)
●Relation between the hands in asymmetric two-handed signs if this corresponds to the
easiest articulation of the other parameters (location, relative orientation, movement
direction)
7.19. Combinations of values in a single field
●A right arrow (>): separates values for different sequentially expressed values. E.g.
ipsilateral > downwards (movement directions in SYSTEEM/SYSTEM). It is not used
for repeated movements, which simply receive a checkmark in the field ‘repeated
movement’. For signs with a sequence of movements, like OUDERSB/PARENTS-B (first
downward on the cheek, then sideward on the chin), only the changing aspects of
the whole sign are specified as a complex value separated by an arrow, not the field
values that remain constant (like handshape in this example).
●A plus character (+): separates values for two simultaneously expressed values. E.g.
upwards + ipsilateral (movement direction in AFWIJKEN/DEVIATE). The same
convention is used for glosses that refer to signs with multiple simultaneous meaning
components, such as number incorporation (UUR-LANG+2/HOUR-LONG+2), and for
the annotation ID glosses for classifier constructions (e.g. AT+B).
●A slash (/): separates two different values describing features of each hand. The first
value pertains to the dominant hand, the other to the non-dominant hand. E.g.
radial/ulnar (relative orientation: movement, in KEEPER/GOALIE). There are three
types of combinations of values that may occur in various fields, that are indicated
with special characters:
8. Semantics
8.1. Iconic image
If the sign represents one or more aspects of an entity or event, a description of the most
salient visual image(s) of the sign as a whole (in the sense of Taub’s (2001) image building
model) can be provided here. Some examples:
●HOND-A: opzitten (DOG-A: sit up)
●HOND-B: krant in bek (DOG-B: newspaper in mouth)
●HOND-C: op been tikken (DOG-C: call by patting leg)
●FIETSEN-A/CYCLING-A: pedalling
25
8.2. Named entity
A classification of all kinds of names can be made here, in terms of one of the following
categories:
●Brand
●Company
●Continent
●Country
●Device
●Drink
●Event
●Magazine
●Newspaper
●Organisation
●Person
●Place
●Place nickname
●Product
●Project
●Public figure
●Region
●Story character
●Team
Name signs for people who have (had) a public function (e.g. Balkenende, Adolf Hitler) get
the tag ‘public figure’ in the field named entity (Semantics);other people’s name signs are
tagged as ‘person’.
8.3. Semantic field
A simple classification of the concept in terms of the principle semantic field can be made.
The list of semantic fields has been taken over from the Auslan Signbank, and may well need
to be extended in the future. Also, the selection of multiple semantic fields is desirable; this
is technically not possible at the moment.
9. Minimal pairs
Minimal pairs are automatically generated (overnight) on the basis of the phonological
specification. Any value that’s different between two signs (incl. absence versus presence of
a feature) leads to a minimal pair listing, with exclusion of the following fields:
●Phonology other
●Mouthing
●Phonetic variation
26
10. Relations to other signs
Some first notes:
• A semantically neutral type: ‘variant’ is automatically generated for all A/B/C variants if no
more specific (semantic) relation is already present. This is not necessarily based on a
linguistic analysis of the two signs, but simply facilitates navigation between A/B/C variants.
The Variant relation does therefore not imply that they are synonyms in NGT.
• Specific relations override the generic ‘variant’.
• A relation that is manually added will be symmetrical: a (reverse) specification for that
relation will be automatically added for the related sign.
11. Relations to foreign signs
Add description & when and how to use
12. Notes
Add, incl. discussion of note types
13. Importing data
Entering new data manually through the user interface is not always the most efficient. Both
for creating new signs and for modifying the description of existing signs, batch options are
available through importing CSV tables.
13.1. Batch import of videos and images
Users can not batch import video data or image data, but they can ask the administrator of
the system to do so for them. File names should correspond to the Annotation ID Gloss of
the sign to which the file is to be linked, in any of the description languages used for the
dataset. Make sure to systematically use the same language per batch, and to keep data for
different datasets in separate folders.
Note to administrators: subfolders for videos/images need to be made manually before
media can be uploaded into import_videos resp. import_images (e.g. NGT/ned/), and
likewise under glossvideo resp. glossimage before the import can be executed (e.g. NGT/);
organization within the latter is handled by the system.
13.2. Batch creation of new lemma and sign entries
New signs can be generated by importing a CSV file (menu Signs > CSV Create). The CSV file
should contain a column ‘Dataset’ and Lemma ID Glosses and Annotation ID Glosses for each
of the description languages of a dataset. For the NGT dataset, which has the obligatory
English and the additional Dutch as description languages, the table will look something like
this:
27
Uploading the CSV file will display a list of signs to be added, so the user can check again
whether all is as intended, or list any signs that already appear to be present, so the user can
adapt the file before trying again. Reminder: Annotation ID Glosses need to be unique for
each language, but not Lemma ID Glosses (these are intended to allow grouping several
Annotation ID Glosses under one lemma entry).
Additional information about the signs can only be added later through the menu item Signs
> CSV Update menu item.
13.3. Batch changes to existing signs
Through menu Signs > CSV Update, existing glosses can be modified. The easiest workflow:
●First search for signs to be updated in the interface;
●then export the search results to a CSV file;
●then remove the columns that do not require any changes (to prevent accidental
changes and speed up processing of the import file);
●add, delete or change values for one or more columns (case-sensitive!);
●save the file to CSV again (should it have been stored as an Excel or ODF file in
between), and upload.
●Signbank will display a list of fields to be changed for each gloss in the table, to allow
for a last check, after which changes can be applied.
An example file:
Signbank will alert the user when incorrect fields or values are listed (such as ‘bC’ for Strong
Hand handshape, which should have been Baby_C).
Note:
●The Signbank ID and Dataset columns/fields are mandatory (but not the glosses); at
least one additional column is needed.
●The column headers need to be exactly as in the CSV Export.
28
●Signbank is a little critical when it comes to the exact format of the CSV file. On mac,
the standard CSV files that MS Excel saves don’t work, while LibreOffice never gives
any problem. It may pay off to try different options if Signbank gives errors.
13.4. Batch changes to existing lemmas
Lemmas can be batch edited via the menu item ‘CSV Import Update Lemmas’, and one by
one via the list displayed by ‘Show all lemmas’.
14. Users, Permissions and Access Rights
Users who do not have an account or who are not logged in, can only see signs in data sets
1. that have been published (something that can only be done in the admin by changing
the dataset from ‘private’ to ‘public’ at the top of the page with dataset properties;
currently only NGT, Kata Kolok, and EmblemsNL);
2. and then only those signs that have been given a checkmark for ‘In Dictionary’ in the
panel Publication Status. The interface options are limited in this case, and users see
only the video, glosses, and translation equivalents for a sign.
On the login page, new users can create an account and ask for access to one or more
datasets.
29
30
After registration, users are entered added to the system in the group Guest, and see more
options in the interface and all signs in the dataset to which they have access. Owners of the
dataset get an automatic log mail mentioning who has gotten access, and what the new
user’s motivation for wanting access was.
Only admin users can change the group a user is in, in the admin interface.
●Group Researcher gives the user access to CSV export of search results (for the
datasets to which they have access);
●Group Editor gives the user write rights to the datasets for which they have been
allowed to modify information;
●Group Data Manager gives the user the additional right to assign and withdraw View
and Change permissions for any registered user, and to manage some of the
dataset’s data, through the Dataset menu (see screen shot below);
●Group Publisher gives the user the additional right to publish signs.
Functions available to dataset managers:
●Set the default language for their dataset [not sure what that does, OC]
●Download a CSV file with hyperlinks to all the media for a gloss
●Upload corpus data from which sign frequencies can be extracted (see section 15.3
below)
Users can see to which datasets they have view and change permissions in their User Profile,
accessible from the menu bar (top right, Profile).
15. Using Global Signbank in Combination with ELAN
Data from a Global Signbank dataset can be made visible in ELAN in two ways: by means of
an External Controlled Vocabulary (ECV) and by means of a Lexicon Service. These two
complement each other, and serve different functions. They are also explained and
illustrated in the following screencast for an earlier version of ELAN:
https://vimeo.com/channels/radboudngt/225093547 (‘Innovations in ELAN: Linking to a
Sign Language Lexicon’)
Further, see the ELAN manual at https://www.mpi.nl/corpus/manuals/manual-elan.pdf.
31
15.1. ECV
If the Type ofaTier in ELAN is linked to an External Controlled Vocabulary, this vocabulary
functions like other Controlled Vocabularies in ELAN. It offers the user a list of options in
creating new annotations or changing existing ones (these are the values in the vocabulary).
The description of each value in the vocabulary presents some more information from
Global Signbank: the list of translation equivalents and a few phonology fields that help
identification of the right ID glosses. The exact display of these values and descriptions in the
timeline in ELAN can be configured in the ELAN preferences.
The ECV file also enables the use of multilingual glosses in ELAN: if multiple
languages are specified for a dataset in Global Signbank, these can be selected for display in
ELAN (menu Options > Language for Multilingual Content). See the following screencast for
further instruction: https://vimeo.com/channels/radboudngt/126786931 (‘Innovations in
ELAN: Multilingual Annotations’).
The ECV file is stored on the Gobal Signbank web server, and the link to this file can
be added when configuring a Controlled Vocabulary in ELAN (they all like like this:
‘https://signbank.cls.ru.nl/static/ecv/ngt.ecv’). When you open an EAF file with a link to an
ECV, ELAN will check whether there is an updated version on the server, and load it into its
cache (but not into the EAF file itself). If it cannot find a network connection, after some
time you’ll receive a message saying ELAN will use the previously cached version of the file.
This ECV cache is invisible to the user. The absence of a network connection can slow down
the starting up of ELAN files, should you often be working offline.
Global Signbank updates the ECV files for all datasets on the server overnight
(Europe’s night, that is).
15.2. Lexicon service
While the ECV allows you to see text data from Global Signbank in ELAN, the Lexicon Service
allows you to also display movies for Annotation ID Glosses. You need to configure it to refer
to the same dataset as the ECV, of course, but the two are independent of each other.
Movies can be displayed along with the ECV list of glosses when selecting a new gloss (which
you’ll need to configure in ELAN’s preferences), or they can be displayed for existing glosses
in the timeline (by holding the control key and hovering the mouse pointer over the
annotation).
A Lexicon Service likewise is a short link in the EAF file, that points to the Signbank
server. And likewise, it downloads information into a cache, but this time one where you can
see it. As a small version of all the movies is downloaded, this folder can get very big (the
NGT datasets accounts for over 2.5GB for 4,000 signs). Especially the first time a Lexicon
Service for a dataset is created on any computer, it may take a long time to download if you
have many signs in your dataset. However, you will not have to do this for every EAF file you
link it to, and updating goes much faster. You can specify in the ELAN settings where on your
hard drive you’d like this cache folder to be. (Planning of May 2023: in 2024, the Lexicon
Service will be made more robust and more flexible.)
32
15.3. Extracting corpus frequencies for Signbank glosses
The frequencies of glosses in a corpus can be displayed for signs in the Frequencies tab
for a sign. To make this work, dataset managers need to upload all the EAF files for a
corpus, as well as a CSV file with the metadata for each corpus session (= a single EAF
file) and for each signer.
Data can be uploaded via the Dataset Management page (menu Datasets >
Manage Datasets).The ‘Manage’ buttons allow for uploading new EAF files and a CSV
file. Any existing file can be monitored by displaying the Corpus Overview (see the
button with that name on the dataset’s detail page, e.g.
https://signbank.cls.ru.nl/datasets/NGT).
Further requirements:
●The gloss tiers in ELAN should have the Type ‘gloss’
●The metadata should refer to the ‘Participant’ tag of the gloss tiers
●The metadata currently support the following signer details: The
metadata CSV currently supports:
○Signer ID Used in Dataset EAF files
○Region (Geographical) Region
○Age Age at time of recording
○Gender Male (m), Female (f), Other (o)
○Handedness Right (r), Left (l), Ambidextrous (a), Unknown
●The header row of the CSV file should be (using tab as separator):
Participant Region Age Gender Handedness
●(This information is also displayed on the Corpus Overview page before
any data have been uploaded.)
16. References
Crasborn, O. (2001). Phonetic implementation of phonological categories in Sign Language of
the Netherlands. PhD thesis, Leiden University.
Crasborn, O., Hulst, H. van der, & Kooij, E. van der (2001). SignPhon: a phonological database
for sign languages. Sign Language and Linguistics, 4(1/2), 215–228.
Crasborn, O., Bank, R., Zwitserlood, I., van der Kooij, & Ormel, E. (2020) Annotation
Conventions for the Corpus NGT. Version 4, July 2020. See annotation conventions (pdf, 2
MB) at
https://www.ru.nl/en/departments/centre-for-language-studies/sign-language-linguistics
Crasborn, O., & Kooij, E. van der. (2023). The Emergence of the Second Hand in Sign
Language Phonology: From Underlying to Surface Representations. In J. van de Weijer
(Ed.), Syllable, Stress, and Sign (pp. 319–344). De Gruyter.
https://doi.org/10.1515/9783110730081-016
Fenlon, J., .Cormier, K., & .Schembri, A. (2015). Building BSL SignBank: The lemma dilemma
revisited. International Journal of Lexicography 28(2). 169–206.
Hulst, H. van der, & Kooij, E. van der. (2021). Sign Language Phonology. Theoretical
Perspectives. In J. Quer, R. Pfau, & A. Herrmann (Eds.), The Routledge Handbook of
Theoretical and Experimental Sign Language Research (p. 34). Routledge.
33
Johnston, T. (2008). Corpus linguistics and signed languages: no lemmata, no corpus. In O.
Crasborn, E. Eehimiou, T. Hanke, E. Thoutenhoofd, D., & I. Zwitserlood (Eds.), 5th
Workshop on the Representation and Processing of Signed Languages: Construction and
Exploitation of Sign Language Corpora (pp. 82–87). Paris: ELRA.
Konrad, R., & Langer, G. (2009). Synergies between transcription and lexical database
building: The case of German Sign Language (DGS). In M. Mahlberg, V. González-Díaz, & C.
Smith (Eds.), Proceedings of the Corpus Linguistics Conference (CL2009). University of
Liverpool, UK, 20-23 July 2009. Liverpool: [online proceedings].
Kooij, E. van der. (2002). Phonological categories in Sign Language of the Netherlands. The
role of phonetic implementation and iconicity [PhD Thesis]. Leiden University.
Kooij, E. van der, Zwitserlood, I., & Crasborn, O. (2023). Strategies for new word formation in
NGT: a case for simultaneous morphology. Sign Language & Linguistics.
König, S., Konrad, R., & Langer, G. (2008). What's in a sign? Theoretical lessons from practical
sign language typography. Paper presented at the TISLR9, Florianópolis, Brazil.
Taub, S. (2001). Language from the body: iconicity and metaphor in American Sign Language.
Cambridge: Cambridge University Press.
Zwitserlood, I., Kooij, E. van der, & Crasborn, O. (2021). Morphological complexity in sign
languages and the classifier-core sign dilemma [Manuscript]. Nijmegen: Radboud
University.
Zwitserlood, I., Kooij, E. van der, & Crasborn, O. (2023). Units of sub-sign meaning in NGT. A
toolbox for sub-sign meaning in a lexical database for a sign language. Sign Language &
Linguistics.
34
