A.Volpentesta (volpente@giuda.deis.unical.it)
A conventional electronic Bulletin Board (BBS) has the same function as a cork and pin bulletin board, as it serves as a collector of messages. Contents of a bulletin board can be read or be updated by other users. A natural extension is a MBBS , i.e. a system which adds multimedia capabilities to the traditional services of a BBS.
In this paper, we describe a feature of such a system which has been realized at theWWW page, without requiring any user knowledge about HTML. The bulletin is automatically linked to a client page index as soon as user composition session of the resulting page is finished.
Moreover, in order to allow a user to select suitable images for his WWW page, an image database has been embodied in the system. The image retrieval problem has been tackled by handling structured and dynamic image descriptions. Finally we envisage, through selected examples, some possible applications in a collaborative work environment.
When we say BBS (Bulletin Board System) we mean exactly the electronic extension of a traditional bulletin board with pins and sheets, where the system operator (shortened as SysOp) can put on line some actual bulletins for his clients who connect to BBS, via telephone, by using a PC and a modem. In fact, the functions of a BBS are much more extended, but the abore said paradigm can be easily multimedia extensible utilizing the WWW environment.
To extend a traditional model of BBS toward a Multimedia BBS a communication protocol must be defined, to integrate a video-audio-text traffic.
In our case, we use the HTTP protocol which already implements this kind of integration. The following step for a MBBS is to implement a paradigm for a collaboration work, thus giving to users the possibility of compiling individual or group bulletins. A first protoype of a multimedia collaboration bulletin could have a text inserted by a user and one image selected by a Pictorial Database (PDB in short). This last operation should obviously be done through a structured query. Furthermore, to get a dynamic management of images in these environments, it will be necessary to define a model for a distributed database concerning the encoded descriptions of the same images.
Referring to the WWW environment, usually anyone who has access to server can only compose some pages and leave them available to users-clients; in addition, anyone who has to compose these pages should be familiar with HTML, what is not certain for a final-user, especially in non-academic environments.
Therefore, the decision was made to implement a kind of interactive system with a page on our server, to which the user is able to connect through whatever WWW browser (e.g. Mosaic) . Starting this point on, he is guided into the creation of his own bulletin, without affecting the system security and having no necessity of knowing HTML. A page could be thought of as a front-end to interact with the client.
The necessary steps to compose are the following:
In order to activate the FORM option, some scripts (in C language) being in conformity with CGI (Common Gateway Interface) protocol were utilized .
The most crucial phase of the entire process is that one decribed in items 2 and 3, for it touches on a well known problem concerning image indexing and retrieval, i.e. performing queries against the indices.
In this paper, we don’t deal with picture physical representation problem; we assume that images are stored in an external videobank, such as an optical storage server, connected to the PDB, which keeps track of their location.
A logical picture representation consists of some keywords, and the organizing methodology is a variation of concept indexing. In our context, the normalized dictionary is the set of all keywords. The indexing technique is based on a knowledge base (KB) encapsulated in the PDB. The KB structure is a decision lattice which imposes a classification hierarchy on the keywords dictionary.
In such a decision lattice, terminal nodes are associated with a set of keywords used for image descriptions (i.e. in our normalized dictionary) and intermediate nodes with new abstract terms which denote different layers of classification.
The idea is to let the system drive the user in the keyword search phase. At each step of the interaction, the system proposes a set of class terms and the user is required to select the appropriate one which the keyword is supposed to belong to. Of course a low-level control structure, such as a decision lattice, is hard to build and difficult to modify and debug. In spite of this, it has been chosen essentially because its use makes the keyword search, and thus the user-system interaction, faster than other more complex methods, where matching, binding or backtracking is needed.
The prototype query let the user insert a free number of keywords selected in the KSS, linked together with logic operators. Let’s suppose that we have images of wild life scenes in our PDB, and we have associated to them some keywords that describe actors and landscape in an image. Users can retrieve an image by formulating different queries and each one of them is specified by keywords selected through some paths (root-terminal node) on the decision lattice.
Let’s assume that the user wants to find an image of a hunting feline , he/she could access to the appropriate keywords, selecting either the path crossing the “Hunting” node or the path crossing the “Feline” node.
To make the loading procedure faster and to allow a quick view of the query result, images are resized in order to fit them in a page as a fixed-size stamps list .
The physical addresses (disk path or videodisc frame number) are stored in a temporary array. After user’s image selection (step #5), the chosen pictures in their original size are inserted in the client page by the HTML issue, where their physical addresses are specified.
The actual system is nothing but an experimental prototype. Of course many technical improvements are possible. Here we list some of them that can be implemented in the future
:
Moreover, by giving the capability to remote clients to update the video-bank and the PDB, it’s inconsistent too. Indeed different users may choose different word to characterize the same image or very similar ones.
Therefore, some works has been directed towards a definition of an appropriate collaborative work model for both driving a remote user in his/her description of an image and (at least partially) automatic indexing, i.e. updating the decision lattice. Such a system could have different applications, both in commercial than in academic environments. In effect, the first utilization of the protocol HTTP was developed at CERN , with the aim of making the exchange of the research results possible. Through this system, the users who have no server or do not know HTML, can nonetheless participate in the scientific debate creating their documents.
Its application in collaborative work cases might be distributed-asynchronous conferences, group desk top publishing (Joint editing) for an electronic newspaper, and so for.
For example our prototype has been utilized during a University course of the last academic year , concerning the degree in Computer Science Engineering, to stimulate the collaborative group works; the aim was to compose an hypertextual guide on tourist resorts in Calabria. This experimentation was the natural prosecution of a previous one effected through a BBS working at our Laboratory.
Furthermore, the system is particularly fit for commercial applications, such as tourist promotion, products promotion, multimedia announcements and chat sites, car and estate sale. In conclusion, our work was born from two needs, the first one of giving to users-clients the possibility to contribute to the server growth and improvement , enabling them with an easy interactive device to accomplish this aim; the second one of realizing in the WWW environment a valid interface with the management of pictorial databases, on the basis of experiences and competencies developed in this laboratory and related to this subject.