Elsevier

Expert Systems with Applications

Volume 39, Issue 3, 15 February 2012, Pages 3135-3144
Expert Systems with Applications

A framework for designing closed domain virtual assistants

https://doi.org/10.1016/j.eswa.2011.08.177Get rights and content

Abstract

Since its beginning in 1969, the Internet has grown rapidly, especially over the past few years. Companies and organizations store more and more information about themselves on the Internet. Sometimes, that information is not well organized. Other times, the huge volume of available data makes useful information difficult to be found. The result is that users have to waste their time looking for what they want to know using the traditional menu-driven navigation and keyword search that websites provide. This is a critical issue because it decreases users interest about companies. In order to avoid this problem, in this paper we propose a framework for designing virtual assistants, which are, considering first results, an ideal alternative to help users find, not only the information that they are looking for, but also some related information which could be of the highest interest.

Highlights

Virtual assistant framework that helps users find useful information in a website. ► It engages users in conversation using natural language, as real assistants. ► Users can omit some words if they are implicit in the conversation (context). ► It answers people’s questions and offer related information of the highest interest. ► It is multilingual and can be integrated into every existing website.

Introduction

With the arrival of the digital era and the development of the Internet over the last years, there are more and more companies that store all the information they own in a digital medium and publish it on the Internet, so that everybody around the world can have access to it. Traditionally, when people want to find some information about a company, they surf the Net and go to the website of that company, where they navigate through its menus and use keywords to look for what they want. However, due to the big amount of information, it gets very difficult for companies and organizations to structure their contents in a logical way so that they can be accessed using only one or two mouse clicks. Nowadays, this objective is usually a utopia because users have to waste their time looking for what they want to know using the traditional menu-driven navigation and keyword search that websites provide. This decreases users’ interest in surfing websites and hence in finding out about companies. If the information to handle is small, this problem can be resolved changing the structure of the contents and making the access easier. However, as the information grows, it is more and more difficult to provide users with an easy and fast access to all that data. Therefore, the necessity of new ways of accessing the information becomes evident.

First of all, we must stop to think why this problem is emerging, that is, why users have to spend a lot of time wandering through websites to get some useful information. There are many different reasons. The first one arises from the diverse nature of the Internet. Nowadays, there are millions and millions of websites about different topics. The main difficulty from a surfing point of view, although it is also one of the most valuable features of the Web, is that each website is different from the others. For this reason, there is a learning curve when you visit a website for the first time, and the steepness of the slope depends on how well organized the contents are and how skillful the user is. This is the second problem, there are many different kinds of user. Some of them are used to working with new technologies, and they can find the information that they are looking for in a website very quickly, even though they have never visited it before. However, other users find it very difficult to navigate through a website using menus to discover new information, and they need somebody to support them. Another pitfall is the fact that companies store more and more information. Perhaps at the moment they do not have any problem, but in the future, as the company builds up, it is very likely that they do not know how to organize such an amount of information so that users have a fast and effective access to it. Eventually, there is another complicated problem that we should try to resolve. Sometimes people are not really sure about what they are looking for, so we cannot want them to choose a category from a menu because they do not know it.

All these problems that we have just mentioned make it evident the necessity of an efficient and effective mechanism for organizing and accessing the information of a company when it becomes very big. Therefore, that new system must have certain features. First, it must be real time. Users’ time is highly valuable, so if they had to wait every time they wanted to get some information, the system would be useless. Another important feature is effectiveness, that is, the system must achieve the goals and objectives that users want. In this sense, the quality of the results must be high. Moreover, the communication between the system and the users must be natural, allowing them to ask natural language questions, that is, complete sentences, rather than simple keywords. Finally, the interface of the system must keep as simple as possible so that everybody can use it, regardless of their computer skills. To sum up, the aim of the system must be to make the access to information easier for everybody.

The rest of this paper is organized as follows. We begin with presenting some related work involving the use of intelligent systems for supporting users in surfing websites. We then present our Virtual Assistant, describing its main features and making the advantages that it provides clear. We explain the architecture of our system in detail, as well as the different modules of which it is made. Then, we describe the user interface and its functionality. Afterwards, we present the results of the first six months of our virtual assistant for the University of Granada. Finally, we conclude the paper and provide some directions for future work.

Section snippets

Related work

An Embodied Conversational Agent (ECA) is an intelligent system represented by a character which is able to engage in conversation with a human being. This technology makes human–computer interaction much friendlier and can be used in any situation where there is a human-to-human interaction, with the ECA playing the role of any of them. Nowadays, ECAs are being more and more applied to many different fields. Entertainment, tourism, e-learning, e-commerce and medicine are just some examples of

System design

In this paper we present a framework for designing closed domain virtual assistants, that is, embodied conversational agents that help people find useful information about a topic in a website. These virtual assistants are real time intelligent systems because users do not want to wait for getting the information they want. They engage users in conversation using natural language, as if they were real assistants. This way of communication is much better than the one offered by traditional

The user interface

There is a famous quotation from Albert Einstein which says ‘make things as simple as possible, but no simpler’. This should always be the maxim when designing the user interface of an application. Following this line of work, we propose the interface shown in Fig. 6. This picture displays a web page with two different areas. On the one hand, the top of the page contains the specific interface of the virtual assistant. On the other hand, the remaining area is used by the system to show any web

Performance evaluation

In this section, we analyze the reliability, effectiveness and efficiency of our system. For this purpose, we consider Elvira (2010), the virtual assistant that provides information about the University of Granada. The kind of information that this assistant is able to supply is very wide: offices, services, studies, access, registration, faculties and schools, courses, activities, regulations, forms, scholarships, accommodation, projects, and so on. In all, more than two thousand different

Conclusion and future work

In this work we have studied the problem of the accessibility of information on websites. As this information grows, it is more and more difficult to organize it in an appropriate way so that users can easily find it. In addition, sometimes the information about a topic is scattered throughout the website, so it is even more difficult to be found. For this reason, we have reviewed some of the existing models for improving web navigation, making some of the disadvantages of these systems clear.

Acknowledgments

This work has been supported by the Spanish Ministry of Science and Technology under research project TIN2007-67984-C02-01, the Regional Andalusian Government under research project TIC-P06-01424, and a FPU scholarship from the Spanish Ministry of Science and Innovation.

References (25)

  • F. Abbattista et al.

    Samir: A smart 3d assistant on the web

    Psychology Journal

    (2004)
  • Alicebot. (2007)....
  • Armstrong, R., Freitag, D., Joachims, T., & Mitchell, T. (1995). Webwatcher: A learning apprentice for the world wide...
  • Artificial-Solutions. (2010)....
  • G. Ball et al.

    Lifelike computer characters: The persona project at microsoft

    (1997)
  • J. Cassell et al.

    Embodied conversational agents

    (2000)
  • Cockburn, A., Greenberg, S., McKenzie, B., Jasonsmith, M., & Kaasten, S. (November 1999). Webview: A graphical aid for...
  • Elvira. (2010). Elvira: the virtual assistant of the university of granada....
  • Flash. (2010). Adobe flash professional....
  • Gruber, T. R. (1993). A translation approach to portable ontology specifications. Knowledge Acquisition (Vol. 5, pp....
  • M. Jaczynski et al.

    Www assisted browsing by reusing past navigations of a group of users

  • Jena. (2002). Jena semantic web framework for java....
  • Cited by (0)

    View full text