Keywords

1 Introduction

E-commerce (electronic commerce), a transaction of buying or selling online, has grown over years. These business transactions occur in four business models: business-to-business, business-to-consumer, consumer-to-consumer or consumer-to-businessFootnote 1. Online shopping, also named as electronic retail, is a form of E-commerce which allows consumers to directly buy goods or services from a seller over the Internet using a web browser. Because of its convenience, it has been reshaping the attitude and behavior of customers for purchasing products.

A success online shopping procedure consists of four steps, including (1) product searching, (2) products comparison and selection, (3) confirmation and payment for the product, and (4) products delivery. In order to complete an online shopping loop, customers must access to the Internet and complete the first three steps. Although it attracts the younger generation who can quickly reach such requirements, the senior population shows less interests in shopping online. Most of the current E-commerce websites do not fully cover the requirements of the elderly. Some of these websites are not clear and intuitive enough, inevitably bringing difficulties for the elderly who suffers from age-related impairments or unfamiliar with computer. It is observed that the elderly are more reluctant to use information technologies [13]. Moreover, a study indicated that the users aged over 65 years old were over 40% slower than the younger generation in using the Internet, and were more likely to give up their trials [14].

In the past decades, the world population is aging at an unprecedented rate. According to the report of An “Aging World: 2015” [16], the number of older people has increased more than 60% in just 15 years. The aging population will continue growing in the following 35 years. By 2050, the number of persons aged 65 years and above will reach 1.6 billion, which is over double size of this population in 2015. Hence, there will be a huge potential E-commerce market for elderly, in turn how to adapt the E-commerce system for the elderly to improve their living quality has become an important research topic.

1.1 Related Works

There are several barriers for the elderly using E-commerce. One of the most important reasons is the declining of physical and cognitive functions of the elderly [1]. A number of previous studies are focus on the evaluation and development of age-friendly browser design [1, 4, 5, 9, 10]. In [9], the age-friendly design principles in terms of the usability were introduced. In [8], 36 websites were evaluated by older adults in items of 25 “senior-friendly” guidelines recommended by the National Institute of Aging. In [5], the authors investigated the web search and navigation system for the elderly. The study [10] evaluated the usability of E-commerce website and highlighted the importance of clear and smart navigation design for senior customers. In [4], the author proposed a Non Browser design for older novice users.

Unfortunately, the barriers of elderly using online shopping not only come from the developer side. There are also some difficulties caused by the inexperience of old users. A well-designed training program is one of the ways to reduce such problems. However, it hardly cover all the possible conditions and is time consuming. Another way is to develop age-friendly functions to improve the usability of an E-commerce website for the elderly [3]. In [2, 15], speech technology was adopted for products search. The study [7] presented a study on using voice demands for the elderly when browsing websites. In [12], we also designed three age-friendly functions, crowd-improved speech recognition, multimodal search and personalized speech feedback, to improve the elderly’s online shopping experience.

1.2 Contributions of This Paper

To improve the usability of the age-friendly website, in this work, we improve and integrate our previous designed functions into our E-commerce system. Different to previous works focusing either on the UI design or on functional module development, we consider both aspects to improve the usability of the E-commerce system.

We extend our previous work in [12] with following aspects: (1) An age-friendly UI following the human factor design for the elderly, such as simplicity and intuitive, is designed for our E-commerce system; (2) we integrate the previous developed functions, namely multimodal search and personalized speech feedback, into our proposed system; (3) we develop the product reputation module to further improve the online shopping experience of elderly. With these integrated function modules, our age-friendly E-commerce system provides users with the benefits of flexible multimodel search, personalized speech feedback and high-quality search results.

The rest of the paper are organized as follows. Section 2 discuss the problems in existing age-friendly E-commerce system. We then present our proposed age-friendly E-commerce system in Sect. 3. Section 4 is the conclusion and future directions.

2 Problem Statement

The decline of physical and cognitive functions may cause inconvenience for older people when using the Web and shopping online. Particularly, contrast discriminations and color perception decline with age increase [6]. Hence, past studies usually focus on the problems of visual crowding and the text font choice in terms of size, style, and color [1, 4, 5, 9].

In general, it is commonly agreed to use bigger font size and list fewer items in a single page to increase the age-friendliness of a website. Although such UI designs can make an E-commerce website age-friendly to some extent, how to improve the usability is still an open problem. In this work, we mainly focus on the following aspects of an age-friendly E-commerce system.

  • Product Searching: Search accuracy of an age-friendly website is crucial for the user experience. Most elderly, however, are relatively poor at searching. They use search queries that are either too broad or too narrow. This could easily make them frustrated if irrelevant search results occupy the first several places.

  • Feedback: Due to the visual impairments and cognition changes, it may be difficult for the old user to access the information of search results they care.

  • Products Selection: The products comparison and selection is also a problem for elderly, who are less experience of filtering out the irrelevant or low-quality results.

3 Proposed Age-Friendly E-commerce System

3.1 System Overview

Our previous work [12] presents three standalone techniques, namely the crowd-improved speech recognition, the multimodal product search, and the personalized voice feedback. In this work, we organically integrate them into two engines, i.e. the multimodal search engine and the personalized speech engine. Beyond them, we further incorporate a product reputation engine to help re-rank and filter out the low-quality search results. To effectively integrate these functional modules into a practical system, we implement them using Python language, which can be linked together through Django platform.

Fig. 1.
figure 1

Architecture of the proposed system for age-friendly E-commerce.

Full size image

As shown in Fig. 1, our age-friendly E-commerce system contains two parts: (1) UI design, and (2) integrated E-commerce engine. In UI design part, similar to previous studies, our system adapt the UI of E-commerce website to assistant old users having visual and cognitive impairments. In the integrated E-commerce engine part, to improve the usability, three E-commerce engines, namely the multimodal search engine the personalized speech engine and the product reputation engine, are integrated into our system.

With age-friendly UI design and functional modules, our website provides users with the benefits of following properties,

  • Our integrated system improves the flexibility of searching. User can search products using any combination of image, text, and voice as inputs;

  • The informative feedbacks are returned to users in a personalized voice for them better understanding the results they are searching for;

  • The product reputation engine re-ranks the results based on the user feedbacks of the products to help user to filter out the irrelevant products. This will help the users to select their desired products;

  • Due to the unified implementation of all engines, our system become more robust and efficient.

In general, our integrated age-friendly E-commerce system provides a variety of adaptations, therefore meeting the needs of elderly.

In the following subsections, we will present the UI design and the integrated functional engines of our system.

3.2 UI Design

In this section, we will briefly introduce the age-friendly UI design of our system.

In order to adapt our UI to the old user, we incorporate of human factors for the elderly into the UI design of our system. As shown in Fig. 2, big font sizes with contrastive color (i.e. white text on dark backgrounds) are chosen to improve the visibility of our web browser, and fewer items are listed in a single page for simplified layouts. Additionally, to make the UI of our system more intuitive use, we chose icons to indicate the search functions.

Fig. 2.
figure 2

Illustration of our age-friendly E-commerce UI design.

Full size image

3.3 Integrated E-commerce Engine

In this section, we will present the three integrated functional engines of our system.

Multimodal Search Engine. The multimodal search engine provides multiple ways to users for searching the products. This engine is built upon an online multimodal co-indexing adaptive resonance theory (OMC-ART) [11] and crowd-improved speech recognition (CISR) function [12]. The OMC-ART is in charge of multimodal co-indexing and retrieval of weakly labeled web images. Thus, the OMC-ART enables both image- and/or text-based search. And the CISR function translates the users’ voice into text, then realize the voice search function by feeding the recognized text to the search engine. Both the efficiency and accuracy of OMC-ART and CISR function have been proved in [12].

Fig. 3.
figure 3

Illustration of the difference between text search and multimodel search. (a) The search results use text as input only. (b) The search results use both text and image as input.

Full size image

In our system, user can search by a part of speech, an image from camera or internet. As such, user is able to find the desired products more accurately with comparable time of previous search function. The example shown in Fig. 3 demonstrates the search results of our system with only text and the combination of text and image as input. Assuming the elderly found a picture of desired coat via internet, traditionally, they need to search via a text query. The searching accuracy highly depend on how well the elderly can describe the item. As shown in Fig. 3(a), the first several results may not hint the spot. While by adding the image for searching, it will be easier for them to find the desired item. As shown in Fig. 3(b), the coat, similar to the input picture, appears at the first place of the results. This example demonstrates the effectiveness our multimodel search function. The details of multimodal search function can be found in [12].

Personalized Speech Engine. For inexperience users, it may not be easy to notice all the information of the search results. In order to access the important information of search results for the old users, a personalized speech feedback engine is developed. When the product search is finished, the summary of the search results will be presented to the user with voice. A personalized voice can be generated by the engine for the old user better understanding the speech feedback. Personalized speech engine is built on voice conversion technology, which can transform one speaker’s speech as it was uttered by another speaker with limited training data. Although text-to-speech (TTS) can also realize the similar function with better speech quality, long-time recording of the target speaker is required to build such a system. Hence, voice conversion is a more cost-efficient way to achieve this goal.

Fig. 4.
figure 4

Illustration of the personalized voice feedback function.

Full size image

As shown in Fig. 4, after the product searching, a summary of search result is showed in the web page. Our speech feedback is used to help the elderly to access this information easily. Moreover, the elderly can also choose or create the voice they are prefer for better understanding. They can also replay the speech by click the audio button. The technique details can be found in [12].

Product Reputation Engine. Products comparison and selection of the searching results also might be a problem for elderly. To improve the searching accuracy and filter out the irrelevant products, a product reputation engine is developed in our system. Product reputation engine is responsible for calculating the reputation and re-ranking the products searching results based on the products quality. Three aspects of user’s rating and feedbacks, namely temporal, similarity and quantity, are considered into quality calculation [11]. The reputation is calculated using the fuzzy-logic reputation engine to give a final score of the product.

Fig. 5.
figure 5

Illustration of the product reputation function.

Full size image

Figure 5 demonstrates the product reputation function in our age-friendly E-commerce system. After searching, there is a reputation indicator below the products, which is used to re-rank the searching results. Specifically, the products with higher reputation score will be given higher rank. In this way, product reputation function ensures a robust performance in product search and benefits users for products comparison and selection.

4 Conclusion and Future Work

In this paper, we present an age-friendly E-commerce system with novel assistive functional technologies. Besides the age-friendly website design, our system integrates three functional engines, namely the multimodal search engine, the personalized speech engine and the product reputation engine, into Django platform. The multimodal search function integrate both the speech recognition and image searching techniques and enable the elderly to search for desired products using either speech, images, texts, or any combination of them. The personalized speech feedback presents the summary of search results with personalized voices, which helps the elderly better access and understand the searching results. The product reputation re-ranks the searching results based on the products quality, which improves the searching accuracy and filter out the irrelevant and low-quality products. We have shown the usability of our proposed age-friendly E-commerce system with the real-world E-commerce transaction data from REC-TMALL datasetFootnote 2.

While, it is noted that this is a preliminary study, the design for usability requires testing and iteration. Hence, in future, we will conduct a series of user study to evaluate the system. Additionally, more assistive techniques, such as personalized recommendation function, will be developed to further improve the usability of the system.