In this paper we wish to describe a joint British/Dutch project that is funded by the European Community . The name of the project is Signbase and its main objective has been to build a sign language database which can be used to store linguistic information about a particular sign language. This repository then can be used to generate different types of signed language applications. The project commenced in March 1994 and ended in december 1996.
In this paper we will first discuss the linguistic specifiations of Signbase, subsequently we will demonstrate the prototype of an end-user application (CD-ROM British/ English) Dictionary) and a commercial application which has been developed during the Tide project with additional funding ( CD-ROM SLN/ Dutch Dictionary for parents and teachers deaf children).
The first phase of the Signbase project involved the creation of the linguistic specifications based on DSRU and NSDSK's earlier work on sign lexicography. The objective was to create a database which can be used to store both textual and visual information about a large number of signs from a particular signed language.
The starting point for the creation of the specifications involved defining what would constitute the Signbase domain which in turn would define the type of information to be stored in the repository. The domain was defined as being the signed and written languages used by members of Deaf communities.
The Signbase repository may be visualised as being composed of a number of entities. Entities are abstract database represenations of the information stored in the database. The terms ´Signs` and ´Words` were used to describe the two main conceptual entities (with words being used to refer to the words in the written language to distinguish these from signs (i.e. words of a signed language). In this way the bilingual nature of the repository was clearly established. It is possible however to use the repository as a single language repository. This means that instances for the Word entity do not have to be created if there is no need for the information that this entity contains.
In order to enable the database to store information about both the frozen and the productive lexicon of a signed language seperate entities for Sign and Morpheme can be found in the database.Traditional dictionaries tend to focus on the so called frozen lexicon of a signed language. This is the established lexicon. The productive lexicon is a kind of Do it Yourself lexicon as Brennan (1992) describes: the signer has to put together the different component parts to create a sign which is precisely appropriate for the particular context in which it is being used. The different component parts may be morphemes such as classifier handshapes, non-manual morphemes, meaningful locations and movements which can all be exploited in the creation of signs of signed lanuages such as BSL and SLN.
We defined the following entities for the Signbase database:
Entities within the Signbase repository are related to one another in a number of different ways. There are a number of different types of relations that a Sign can have with other entities. Relations are visualised by a numbered line.
In the figure below the relationships between the entities are visualised.
1.The Sign-Sign relationship
Signs may be related to other signs in a number of different ways. They may have synonym, homonym, antonym or collocation relationships with each other.
2. The Sign-Morpheme relationship
The morpheme entity include different morpheme types such as manual (one-handed, two-handed symmetrical, two-handed-a-symmetrical), non-manual and multi-channel.
The Morpheme entity identifies the relationship between the morphemes that constitute a sign, e.g. whether the relationship is sequential or simultaneous.
3. The Sign-Still relationship
Each sign is linked to at least one still image of the sign.
4. The Sign-Word relationship
Each sign is related to one or morewritten translational equivalents.
5. The Word-Word relationship
Words may have synonym, homonym and antonym relationships with each other.
6.Sign-Movie relationship
Signs may be related to a number of different movies. These are:
7. The Sign-Informant relationship
Each sign may relate to one or more informants.
Information is organised at different levels within the database:
The most detailed information is entered at the attribute level. Attributes may be grouped to form categories which in turn may be grouped to form entities.
To illustrate this we will discuss the Sign entity, its categories and some attributes.
In the figure below the Sign Entity and its categories is shown:
Let us look at the following attributes and subattributes that make up the category
The entity Sign representation is made up of the following attributes and subattributes:
As can be inferred from the type of information and the level of detail, it will not be possible to enter information into the Signbase repository without a considerable degree of linguistic expertise. This is reflected in the nature of the maintenance interface which has been designed to process the range of linguistic information described above.
A sign can be stored in the Signbase database in a number of different ways: sign administrative information (sign name, record status, sign type) resources: information about a sign (source, representation, domain, usage) resources: grammatical information: morphemes, syntactic, discourse sign references: other signs, meaning in written language (words), informants still image(s) of the sign (still(s)), moving images of the sign (movies).
The database user can store information about stills, movies, words and informants in seperate -maintable lists. The first step is to select the source file from the hard disc of the computer, or from the network. We will briefly describe the still and movie records.
Stills are digital images that are imported into the database along with their specific information. During the import of such images a preview (or thumbnail) view of the still image is automatically created, and a compressed version of the full-size image is created. Both image versions are stored in the Signbase database. Apart from the thumbnail, the still record includes the name of the sign, and a notes field providing additional information about the image. A selected image can be viewed at full size and its accompanying text information can be edited. If a picture is considered unsuitable for the database or is rejected for some other reason, it is very easy to delete the still and/or import another one, replacing the original.
Movies can be attached to signs in the database in a very similar way that is described for stills. Signs can be attached to more than one movies. Movies can be of different types, as we have described under ´signbase structure`. The movie record shows the original location of the movie, the movie title and the translation of the movie.
Signbase makes use of a computer archiving system that was developed by Bright Side of Life Computer Consultancy (BSL), called VisageTM. It has been specifically designed for image database tasks. VisageTM images are compressed in real time during storage to approximately 5% of their original size and are decompressed in real time during retrieval for viewing or printing. This not only means that a large number of images can be stored on discs that would otherwise contain a very limited number of uncompressed images, but also that images can be fetched quickly from slow media such as CD-ROM or over networks. Visage ™ supports most common file formats for import and export from most international video standards. Images are compressed and stored in accordance with internationally defined compression standards and can be output to any postscript level 1 or 2 printing device.
Considerable time and effort has been spend by the DSRU and NSDSK consortium members to produce high quality movies that would meet the demands of the users of the end-applications. During the TIDE project we conducted a small study that involved 25 people (hearing and Deaf signers and non-signers) to judge a series of test-movies.
The movies differed with respect to the following factors:
The results indicated that most subjects preferred a darkblue background for both blond and dark haired models. The clothing of the models should neither be too dark or too light enabling the viewer to see the contours of the body while hands and face remain sharp.
Subjects preferred a greyish background colour for the interface. The computerscreen can be adjusted for colour according to the personal preference of the user.
Subjects preferred signs recorded with the model standing straight in front of the camera, looking straight into the camera (i.e. the user in front of the computer) at a distance such that the model is seen from the waist up. The speed of the signing should be normal and as natural as possible.
Although Signbase can store any quicktime movie it is evident that the quality of the movies is very important for the quality of the end-user applications.
At this point in time Signbase is mainly a repository of signs and information about signs. Very few search options are available. We will however continue to improve the database and try to find addtional funding to add a number of search functions and to facilitate the proces of data-entry.
Introduction
The bilingual nature of the Signbase database is reflected in the structure of the CD-ROM dictionary application. The dictionary allows users to access information about individual signs either through the sign language itself or through written English. The CD-ROM dictionary displays information relating to each sign entry, as in traditional text dictionaries ( the 'show' function). In addition it allows the user to search for specific signs or particular sets of signs based on common characteristics selected by the user, such as handshape or region (the 'find' function).
A unique advantage of this CD-ROM dictionary application is that it provides the user with demonstrations of how the sign is produced. These demonstrations are available both with respect to individual signs (citation forms, definition movies) and signs in context (example movies, story movies). The presentation of meanings of signs directly in a sign language (in this case BSL) represents a major step forward for Deaf users.
The CD-ROM seeks to provide an authoritative dictionary application which confirms the status of of signed languages as languages in their own right. It therefore contains in-depth information on key areas of sign structure and use.
However, users may bypass information that is not directly relevant to their current needs. In the production of the application effort has been taken to simplify the interface and thereby support the non-technical user.
The application is tightly linked to the Signbase application but is simplified at some points:
In the demonstration of the proto-type of the CD-ROM BSL/English dictionary the five different groups of functions will be shown:
With additional funding the NSDSK was able to produce together with BSL a first bilingual CD-ROM for parents and teacher of Deaf children in Sign Language of the Netherlands (SLN) and written Dutch. The first CD-ROM was produced in 1996, the second in 1997 en the third has been produced in 1999.
The CD-ROM`s draws on the technology that was developed in the TIDE-project but is not linked directly to the Signbase database. A new, simpler database was developed.
Each CD-ROM contains about a 1000 sign movies (citation form, grammatical variations and stories) and the translation equivalents into written Dutch. The information can be accessed via two parameters of the sign namely handshape and location and via the written word. All signs are grouped into certain domains. The user can also search for signs within a particular domain or theme.
The interface is designed for both children and adults: a special font is used that is attractive to children and special attention has been paid to the graphical elements in the interface.
1 The consortium partners of the Signbase project (Tide project 1282) are the Deaf Studies Research Unit (Durham): David Brien (co-cordinator), Mary Brennan, Judith Collins, Maureen Reed, Ernst Thoutenhoofd, the Dutch Foundation For the Deaf and Hard of Hearing Child (NSDSK, Amsterdam), Truus van der Lem, Trude Schermer, Corline Koolhof, Bright Side of Life Computer Consultancy (BSL, Maarssen), Martyn Simpson and Pindrop Publications(Newark),Peter Llewellyn Jones (back to text)
2 Detailed information about movie production is given in the TIDE report that was written by BSL and NSDSK consortium members: CD-ROM movie production, 1995.(back to text)
Posted: 2.2.2000