ACM SIGIR'99 Workshop on Customised Information Delivery

August 19, 1999 - Berkeley, California

The SwissCast information push service

A multidisciplinary research, a multifaceted experience

[summary]

Foreword1. A few words on push technologies  - 2. A brief presentation of the SwissCast Project -3. Architecture of the SwissCast push service - 4. Specific conditions under which push technologies can work well - 5. A short conclusion - 6. Essential bibliography

Authors: Maurizio Decina, Lorenzo Cantoni, Benedetto Lepori, Riccardo Mazza, Paolo Jannuzzi, Facoltŕ di Scienze della Comunicazione dell’Universitŕ della Svizzera italiana and Dipartimento di Arti Applicate della Scuola Universitaria Professionale della Svizzera Italiana, Lugano.
(Physical location: SwissCast Project, via Ospedale 13, CH-6900 Lugano, Switzerland, tel. +41-91-9124722, fax +41-91-9124647, e-mail: lorenzo.cantoni@lu.unisi.ch, http://swisscast.ti-edu.ch).
 
 

Foreword

This paper has three aims:

1. to briefly present the SwissCast Project, a project concerned with push technologies and services;

2. to discuss some outcomes of the project activities, among those a few conditions under which a push service seems to work;

3. showing how multidisciplinary research is needed when meeting a complex object like that of computer mediated communication.


to the top of the page

1. A few words on push technologies

The most popular Internet communication metaphor for information access is that of browsing for information on the World Wide Web. Here, users actively search for information they need by browsing websites. To do so, they use most often search engines (e.g.: Altavista) or directory services (e.g.: Yahoo).

During the last few years, with the spread of the web, Internet moved from a communication medium for a specific social community and very specialised usages to a generic public-domain medium, accessed by a great variety of users and increasingly used in order to supply generic information, ranging from scientific information to commercial, news and sports.

Thus, the shortcomings of the concept of browsing are becoming more and more clear. According to a study conducted by Dr. Steve Lawrence and Dr. C. Lee Giles at the NEC Research Institute, search engines can't keep up with the Web's rapid growth, because the coverage of the engines is increasing more slowly than the size of the Web, and the percentage of dead links returned by the engines has increased - a problem that frustrates users and leads to bottlenecks in the network. Secondly, if the user needs updated information items, s/he is forced to check periodically the same Information Provider, even when information is unchanged.

Information casting

In Internet, the paradigm of information casting or "push" (also known as WebCasting, PushCasting, Channel Broadcasting, etc.) could offer relevant answers to these issues.

Although a proper definition of "push" raises quite a few issues, a good one seems to be the following:

Push is the automatic delivery of content to users’ computer desktop; content is organised by topic defined by a publisher and users receive information according to their own pre-defined profile.

Three elements thus integrate a would-be complete definition: (1) automatic delivery, (2) content organisation, (3) user profile.

A push service can be carried out by an Information Provider itself, or by a third party, with the role of information broker. The broker receives data from various publishing sources, then selects and prepares data to suit as precisely as possible its users’ needs. To do this it has to select appropriate content from a repository, and to push it to the users according to a personal profile, provided by each of them.

In this way, the meeting of information supply and demand over the Internet could be made much easier, leading to a communication scheme that is closer to the traditional mass media (specialised TV channels, professional magazines, etc.) but with the whole set of advantages given by the digital medium. Moreover, the recipient of information has a much greater opportunity of interacting with the other active bodies in the information chain. Actually, unlike television broadcasting, where all users receive the same information according to the profile defined by the broadcaster on the basis of market analysis, information casting systems on Internet leave large space for customisation.

Push services as information brokerage services

After a first enthusiastic welcome, push technologies don’t seem to have revolutionised the web. Both Netscape and Microsoft are not maintaining their proprietary channel technology any more (nor do they make an extensive use of it), and many push technology providers have changed their interest direction.

In our opinion, this situation can be explained by a rather one-sided vision of push services, which emphasises almost only the technological side of the overall "push story".

Instead, we believe that push services are neither only nor mainly technological tools that help finding information in Internet, but true information brokerage services, which try to bridge the gap between the information contents available on Internet and the user’s needs. To do so, they perform a series of actions (see the concept of "subject gateways" developed in the European project DESIRE, http://www.desire.org) which include:

In this process, technology can be very helpful, mainly in order (1) to access information through the Internet, (2) to perform very repetitive tasks such as regularly searching web sites or matching new information items against users’ profiles and (3) to regularly monitor information flows and users’ behaviours.

Information brokerage

If we consider the activity of a push service as being that of an information broker, we have to take into account three main elements: information providers, information receivers and available information the broker has to manage.

Let us consider the brokerage function in more detail.

(1) First of all: brokers have to know their clients and their information needs, taking into account that both information providers and end users are clients. (2) Secondly: they have to find out relevant information sources according to their clients' profile and collect them. (3) Thirdly: the broker works as a translator. Actually, available information must be semantically mapped following a subject-oriented schema. This mapping is of main importance because it is a kind of shared language between information providers and clients. On one hand it helps classify information, on the other, to better define one’s own user profile. Only if keywords are well defined, and adequate for the subject area, can a good communication exchange be granted. (4) Fourthly: information brokers have to build up a platform where their clients can easily access information they are interested in. (5) Lastly: clients have to be notified – following precise guidelines concerning method and time – when something matches their profiles. This task can be completed by automatic tools.

Several commercial applications provide these services (e.g., Verity's Agent Server, NewsEdge Insight from NewsEdge Corporation, Agentware i3 from Autonomy Inc.), and many research systems (e.g. the ELFI project from Ruhr-University Bochum in Germany); nevertheless often these systems fail because of the use of automatic classification solutions, which can never reach the accuracy and the reliability of human-made classification, and because they are not customized to a specific scope of information content. Moreover, often these systems are not user-friendly for the generic user.


to the top of the page

2. A brief presentation of the SwissCast Project

Swisscast is an ongoing project, which has actually reached the second half of its course. It started in Autumn 1997 and it will end on March 2000.

Purpose

The purpose of the SwissCast Project is to study and work out a prototype communication service — Internet based — which can select and broadcast information organised to meet different user-profiles in well defined professional/interest groups: "push service".

The Project also tries to examine a number of relevant issues, among which the following should be listed: classification and choice of the most suitable hardware/software tools, analysis of information providers’ and information users’ communication needs, design of user oriented graphical interface, realisation of suitable validation procedures.

In order to implement and validate the service, two different areas in the communication market were chosen, namely (1) Research & Development in the University setting, and (2) Pharmaceutics. In the SwissCast Project activities, both information providers and end-users actively co-operate with the researchers who work at the building and validation of the service.

Who’s who and history

The SwissCast project was approved and funded by the Swiss National Science Foundation (SNSF). It is currently run by the Faculty of Communication Sciences of the University of the Italian Switzerland (USI-com, http://www.lu.unisi.ch), and the Department of Applied Arts of the University of Applied Sciences of Southern Switzerland (SUPSI-DAA, http://www.supsi.ch).

Project Leader is Professor Maurizio Decina, USI-com and Department of Electronics and Computing of the Politecnico of Milan (Italy). Other Project directors are Professor Eddo Rigotti (Dean of USI-com and Professor of General Linguistics) and Professor Fiorenzo Scaroni (R&D manager of the SUPSI and USI-com and professor of Telecommunication).

The SwissCast Project has a general accounting manager: Dr. Neviano Dal Degan, an area manager: Dr. Benedetto Lepori, a full-time researcher: Dr. Riccardo Mazza (an expert in database and computer programming for distributed systems), a part time researcher: Dr. Lorenzo Cantoni (specialist in communication and education theories) and Paolo Jannuzzi, consultant for the Graphical User Interface.

In addition, some pharmaceutical companies are involved in the SwissCast Project as Project partners, namely: Pfizer AG (http://www.pfizer.com), a leading American company with an important Switzerland branch; and three Ticino (Switzerland) based companies: IBSA, Institut Biochimique (http://www.ibsa.com); Künzle, and ActaMed Services (http://www.actamed.ch): an agency specialised in information brokering in the health area, with a deep insight and a long experience in the concerned information market.

Project activities at first where those of analysing the two research areas and the involved information market. Bibliographic research as well as interviews with professionals helped in getting the needed knowledge.

In parallel, a deep analysis of available information push research, technologies and services was carried out.

A collaboration with Eurospider (http://www.eurospider.ch) — a spin-off company of the Swiss Federal Institute of Technology in Zurich (Switzerland), led by Professor Peter Schäuble — helped the SwissCast team to better focus its research object and to understand the main differences between information retrieval services and push services.

In spring 1999, a full-functioning version of the SwissCast push service has been put on-line and it will be tested and assessed till spring 2000.


to the top of the page

3. Architecture of the SwissCast push service

In this section, we will present the architecture and the functionalities of the push service built in the framework of the SwissCast Project. This service has been designed as a system that retrieves information from various information providers and information sources, classifies information according to a pre-defined indexing system and diffuses it automatically to end-users according to their interests, as defined in their user profiles.

Information flows

From the point of view of the information flows, the system contains the following main elements:

Information retrieval & information classification. Information is in part directly inserted into the push service by the active information providers (those who have signed a special agreement with the SwissCast service) or it is retrieved by the information editor (mainly, but not exclusively, from the web) and inserted by the information editor in the information storage. All insertion activities are done via web forms. Ad hoc software modules were implemented to interface those external databases with the SwissCast document database; this is the case of active information providers whose data are stored in corporate databases and were previously classified.

All pieces of information items must be classified according to a standard keyword scheme and formatted along a pre-defined structure; this is done either by the information provider or by the information editor with the assistance of technical tools.

Information storage (document database). The document database is a key element in the system, because it contains what will be pushed to end users in a standard format.

Thus, the main function of the document database is to neatly separate between the information retrieval & classification module and the push module, so that the two modules run separately and information items to be analysed by the push module respect given formatting criteria (e.g., having a clear-defined creation date).

Moreover, many items in the document database will contain a reference to a web page that gives more information on the subject.

Push module. This module periodically matches new information items inserted in the document database against users’ profiles and attributes new relevant documents to the corresponding profiles; users can read relevant documents through a common web browser (personal web page approach). Moreover, this module contains an agent that regularly checks for new documents and sends notification to each user via e-mail.

After four weeks of displaying in a (at least once visited) profile, every document is removed, unless end users checked it to maintain it. This activity is performed due to the "push" nature of the service, and in order not to have too big — and thus useless — document lists.

User interface. Users can access SwissCast using a common web browser, through the modules profile manager and document list. Profile manager is the module used to manage users' profile; users have to log-in through an account (username, password) chosen during a preliminary phase of subscription to the system (see screenshot). This module allows users to create a new profile (see screenshot), to delete or to modify the existing ones, to display the list of them , and for each profile to display the total number of documents assigned to it and the documents themselves (see screenshot). Users log to the document list to look up the list of the documents assigned to their profiles (see screenshot). This module offers the possibility for the user to consult the list of the documents in order of date, to display the total number of documents, to delete more that one document at a time, to mark one or more documents as "keep" (to maintain them in the list), to display for each document an abbreviation mark which identifies the information provider.

When designing the service structure, a big effort was made to allow users to concentrate on their professional interests, avoiding irrelevant questions and offering only really useful choices. Three main customisation areas were distinguished: subject related (which subjects users are interested in), communication/format related (the kind of information or event connected with the chosen subjects), and time related (how often users want to be notified if new information items matching their profiles are available).

The user interfaces enable the user to register and log into the service, define interest profiles and check for new relevant documents. The main challenge in building a graphical user interface is to create an environment with a strong and effective visual presence, not to disturb but to enhance the functionality and usability of the service.

The structures have been defined in collaboration with those who are responsible for both the content and the technical part. The interface is composed of three fields, three different points of visual interest with different content: (1) a title bar on the top of the window, with the logotype integrated in a grid system; (2) a control bar on the left; (3) a field on the right where the result of the researches and the profiles are shown; we decided to make this part poor in graphics element, in order to speed up the process of downloading.

Technical realisation

Since the goal of the project is to study the communication flows in push services,0 rather than to develop new technologies, most of the SwissCast system has been developed using freeware software tools adapted to the Project needs and interfaced with the web using CGI programming. In more detail, the following tools were used:

Moreover, the interface of the service with the information editor, the information providers and the end-user is completely web-based and thus all the necessary tasks – including the management of the service and the input of documents – can be done via a web browser.

The SwissCast system is thus based on a database, which acts as a container both for information and user’s profiles data. Information is classified and inserted within the database manually from active providers (filling information via web forms) or automatically through dedicated gateways from other databases accessed via Internet. An account (username and password) is assigned by the information editor of the service to each active provider, in order to guarantee its identification.

A gatherer unit is being implemented in order to help the information editor to perform repetitive tasks. In detail, it constantly checks a list of web pages and automatically upload a web page into a local cache if this was recently modified; through a filter unit the information editor will be able to discard all useless information and to insert into the SwissCast documents’ database only relevant information items.

Once a new document is inserted into the SwissCast documents’ database, a match against all existing profiles is performed, and the document will be automatically and instantly assigned to the corresponding profiles. The match criteria is based on Boolean operators, working on the keywords schema through which every document is classified and every profile is created. In this version the push system is completely deterministic; it works according to a pre-defined set of keywords shared by both the document side and the user profile side.

Moreover, an agent periodically checks each profile for new documents, and sends notification to the corresponding user via e-mail if new documents were found.

Other pieces of software were developed in order to monitor and assess the service (usage statistics, keywords schema efficiency, etc.) as well as the system administration.

What’s currently on-line

Since May 1999 a running prototype has been working for the area of University R&D information (http://swisscast.ti-edu.ch:8002).

The R&D area has been chosen as the first area of application due to the possibility of a close co-operation with the Research Office of the University of Italian Switzerland, thus having from the very beginning of the Project an interested partner that knows very well the specific application field. At the same time, R&D information is a domain well suited for the development of push services because most of the internal communication and of the information exchange in the scientific community already develops via Internet, researchers are thus used to this medium. A very important push service in the research area is run by the European Community (CORDIS-Rapidus; http://www.cordis.lu); another service — in its project phase — is being developed by the University of Bochum (ELFI; http://www.elfi.ruhr-uni-bochum.de/elfi/).

The SwissCast service for R&D has been specifically targeted to the information needs of researchers in Ticino and, in general, of Italian-speaking scientists in Switzerland; it is now fully operational and can be freely accessed from everywhere.

The prototype for the medical-pharmaceutical information service is currently under in-house testing, and will be operational in a few months.

For the University Research area the Subject Classification Codes by Cordis were adopted; for the pharmaceutical area, the Medical Subject Headings by the U.S. National Library of Medicine. For this last area, other keywords were added, to map various information types. Those ones were elaborated through focus groups and interviews held with health care professionals.

Assessment activity and next steps

The R&D service test phase has started in July 1999, it involves about 50 University professors, researchers and students, and assesses the service professional relevance, as well as its functionalities and graphical user interface.

Service assessment will take into account what can be automatically recorded (service usage, number of user profiles, and so on) as well as what the users say and perceive about the service.

In parallel a national survey on the usage of Internet by health care professionals is being designed, and the medical-pharmaceutical service will start being run by ActaMed, and monitored and assessed by the SwissCast team.


to the top of the page

4. Specific conditions under which push technologies can work well

One of the results of the SwissCast project is the identification of a number of key factors which must be fulfilled in order to build a well-functioning push service.

1. Technology matters. At the beginning of the push history, there was a tendency to develop new applications, using very specific interfaces to the user-side; many tools, like PointCast, are quite intrusive and demanding at the hardware and connection level, in order to achieve real-time updating. But not every push service is so demanding at the level of technology: quite often it can use the oldest and simplest push technology — e-mail — to deliver (or, at least, to notify) news, and a common web-browser to view relevant documents. And can be afforded also by people who use slow modem connections.

2. Information relevance matters. Push technology has to meet well defined information needs, specific professional interests. In other words, if information got through the Internet is perceived and remains a second-choice information, only for a lay audience, not so many people will be interested in defining user profiles, for the simple reason that they can’t and don’t want to. Browsing without an end will remain more appealing than just getting specific, narrow-scope information. This explains why the push over intranets, where professional information needs are better defined, seems to grow better than the push over the Internet.

3. Information classification matters. The relevance issue concerns also the issue of how to define user-profiles and — at the same time — how to semantically map the content. After many experiences and experiments, one must agree with the necessity of a human intervention, although assisted by suitable automatic tools, in order to establish how to define a user profile, and to map each information item. Here the difference between search and push services becomes most clear: the first ones have to provide useful stuff, to be further analysed, the last ones must offer only relevant items.

4. Information quality matters. Closely connected with the previous areas, is that of information quality: professionals need to know that information they get is certified. Otherwise they will need nonetheless to go other ways — more traditional, better known by them and thus more secure; which means turning information push into a very uneconomical tool, at least from the point of view of time.

5. Information layout matters. Getting relevant good quality information is the strength of push services, but one more element has to be added. It consists in the way new items are presented, a way which must be easy, clear and not time demanding to be accessed. Otherwise paper technology will win the competition.

6. Graphical User Interface matters. The proper public for push services being mainly professionals, and not necessarily IT professionals, and their reason to use these services being to access updated, good quality, added value, relevant information items, the access must be as easy as possible. Non computer experts (or addicted) professionals don’t want to become IT professionals, they just want to use the computer as a tool, and could be refrained from use it if they perceive the instrument as an obstacle.


to the top of the page

5. A short conclusion

SwissCast research activities have shown how necessary it is to pull together different competencies and skills to effectively meet communicational needs: technological tools are fundamental, but can't work without a continuous collaboration with communication and visual graphics competencies.

The outcomes of the SwissCast project open a series of very interesting developments in research; in fact, SwissCast, being a functioning prototype with real users, offers an ideal platform to test more advanced methods and tools.

In the actual version of the service, technical tools are used only to manage documents and profiles, to deliver information and to execute repetitive tasks. This approach is very robust, because all complex tasks – e.g., information classifying – are left to human competence; on the other side, it can be, if the service becomes large, very time and resource consuming.

The first research line we wish to develop is thus the use of artificial intelligence methods to implement smarter functions, either to help the service manager (e.g., tools which pre-sort the information), or to interact with the user (e.g., to develop ‘dynamic’ user profiles, in order to obtain an improved information casting which could be automatically personalised according to users’ behaviour).

A second research line is that of the semantic mapping of information and of complex systems of keywords; while the actual system is based on a very simple (one-dimensional) scheme, which is the same for the information classification and for the definition of user’s profiles, more complex structures, possibly integrating user’s behaviour, could give more flexibility and could be very helpful to cope with increasing information volumes.

Finally, we wish to stress the issue of customisation of the service: tools and procedures which allow to adapt the system to the structure of each information market and to changing user needs would greatly add to the practical interest of such a service and thus open to new application fields.


to the top of the page

6. Essential bibliography

Decina Maurizio, Rigotti Eddo, Scaroni Fiorenzo, SwissCast. Scientific Description, Autumn 1997

Cantoni Lorenzo, Dal Degan Neviano, Lepori Benedetto, Mazza Riccardo, Scaroni F., SwissCast. Push on the Net: between Market and Technology. Market Analysis and User Needs (Deliverable D 2.1) - Analysis of Available Technologies (Deliverable D 3.1), June 1998

Cantoni L., Jannuzzi Paolo, Lepori B., Mazza R., Delivering Relevant Information to Interested Users. General philosophy, system architecture & technical realisation of the SwissCast push service, Scientific Report, May 1999

Brasethvik Terje, A semantic modeling approach to metadata, Internet Research: Electronic Networking Applications and Policy, Vol. 8, n. 5, 1998

Cerami Ethan, Delivering Push (Hands-On Web Development), Computing McGraw-Hill, 1998

Commission Européenne, Recommandation de la Commission, du 6 mai 1991, concernant l'harmonisation au sein de la Communauté des bases de données dans le domaine de la recherche et du développement technologique, 91/337/CEE, Journal officiel n° L 189 du 13/07/1991 p. 0001 - 0034

Dean Damon A., Web Channel Development for Dummies (For Dummies), IDG Books Worldwide, 1997

Decina M., Trecordi V., Convergence of telecommunications and computing to networking models for integrated services and applications, Proceedings of the IEEE, Vol. 85, n. 12, p. 1887-1914, December 1997

Keyes Jessica, Webcasting: How to Broadcast to Your Customers over the Net, Computing McGraw-Hill, 1997

Miles Peggy, Internet World Guide to Webcasting, John Wiley & Sons, 1998

Petrovsky Michele Jo, Implementing CDF Channels (Hands-On Web Development), Computing McGraw-Hill, 1998

Smith Bud E., Push Technology for Dummies (For Dummies), IDG Books Worldwide, 1997

Szuprowicz Bohdan O., Webcasting and Push Technology Strategies: Effective Communications for Intranets and Extranets, Computer Technology Research Corporation, 1998

Vounakis I., The Exploitation of Current Research Information Systems for Technology Transfer: How CORDIS Works, Luxembourg, CRIS98, 1998 (http://www.cordis.lu/cybercafe/src/vounakis.htm).

Worsfold E., Subject gateways – fulfilling the DESIRE for knowledge, Computer Networks and ISDN Systems, Vol. 30, N. 12-18, 30. Sept. 1998 (http://www.desire.org/html/research/publications/tnc98gateways/)

Yoon Bum Ryeol, You Jae Jeong, Kim Soo Dong, COPEN: A CORBA-Based Intelligent Push-engine, Software Engineering Conference, 1998. Proceedings. 1998 Asia Pacific, p. 330-337.


to the top of the page

Summary

The SwissCast Project aims at the design and implementation of Web Casting communication channels, comprehensive of content information and human interfaces, addressed to the needs of specific Swiss market segments.

The whole casting system is realised by using Internet Protocol (IP) technologies, in order to provide the general and interoperable platform required for end-terminals compatibility, irrespective of channel types and quality available to users.

The focus of the project is on the identification of communities of homogeneous users for whom to tailor the casting service according to their elicited interests, needs and profiles. The users constitute indeed an integral part of the project team, by assuming a direct role in specifying the service features, and in testing and validating the Project results. Two communities of interest are currently involved in SwissCast, namely R&D people (researchers, professors, industry managers, …) and people dealing with pharmaceutical information (doctors, pharmacists, drug vendors, ...); both of them are mainly located in the Ticino territory.

At this stage of the project, searching, filtering and editing of information, to be delivered to users via e-mail and pushed Web pages, is performed by using 'static' (or updated on demand) user profiles. The project target is to explore measurements on the 'performance' of this kind of information delivery (Web casting), by evaluating data 'relevance' in relation with the defined user profile. To implement a 'stable ' and controllable prototype, the use of static profiles is mandatory in the learning phase. We need to test and evaluate the system behaviour in the processes of information 'search-filter-edit-deliver', by appropriate 'measurements' in a 'live' environment.

At a later stage 'dynamic' (i.e., based on a feedback control mechanism) user profiles can be added to the system, in order to obtain an improved information casting which could be automatically personalised according to the statistical behaviour of the user in his actual handling of delivered data.

Research on human behaviour in information gathering processes is a genuine mission of the Faculty of Communication Sciences at the University of Lugano. SwissCast is and will indeed be an appropriate test bed to start building a true interdisciplinary research team, as today needed to foster development of effective networked information systems.


to the top of the page