WO2007119117A2

WO2007119117A2 - A method and system for providing intelligent customer service

Info

Publication number: WO2007119117A2
Application number: PCT/IB2006/004253
Authority: WO
Inventors: Lorraine Mcginty; Barry Smyth; John Doody
Original assignee: University College Dublin
Priority date: 2005-12-12
Filing date: 2006-12-12
Publication date: 2007-10-25
Also published as: WO2007119117A3

Abstract

A method and system for providing a recommendation to a user. The method includes uploading a user image, processing the user image, and generating at least one image of a product or service, depending on the purpose of the system, for the user based on the processing. The method also includes displaying at least one image of the product or service, accepting a feedback response, processing the feedback response, and producing an image of at least one type of the product or service based on the feedback response.

Description

A METHOD AND SYSTEM FOR PROVIDING INTELLIGENT

CUSTOMER SERVICE

BACKGROUND OF THE INVENTION

The present invention relates to providing customer assistance over the internet. More particularly, the invention relates to providing customer assistance to a user over the internet based on both the intelligence of the system of the present invention as well as feedback from the user and/or the user's advisors.

Customers are often challenged in deciding which style or type of either a product or service to purchase. For example, customers shopping for eyeglasses are challenged, and often overwhelmed, by the task of searching for options that "suit" them, both in terms of their design features and individual visual appeal. Another example of overwhelmed people are patients wanting cosmetic surgery who can't decide what feature of their face or body should be altered. Another example is people who cannot decide on which hairstyle they want. All of these are examples of situations that consumers generally find challenging.

A key problem here is that users are often unable to accurately describe what they are looking for in a way that a real-world sales assistant or a conventional retrieval system can understand. Instead, they rely on the optimistic notion that they "will recognize what they like when they see it!" The combination of this vocabulary problem and extreme reliance on positive visual feedback often results in customers taking a long time to make decisions.

For explanation purposes, the selection of a pair of eyeglasses is described as an example. The conventional in-store approach to solving this problem is to allow a customer to engage in the iterative and lengthy process of physically trying-on various frame options and looking in the mirror. Apart from being a time-consuming process, an obvious limitation of this approach is that customers are often unable to make decisions for themselves due to the fact that the prescription of the demonstration alternatives does not match their individual requirements. That is, they cannot see themselves in the mirror. In addition, customers often need a third party opinion in order to make a decision. This is not always available and possible, which can be a major inconvenience for the customer.

One way of addressing this issue is to show customers a photograph of how they look with the different options of whichever product or service the system caters to. While this approach addresses the visual feedback requirement it does not reduce the number of potentially irrelevant options, nor the time required to make ^• an informed decision.

Unsurprisingly, the processes of having to iteratively take these photographs actually increases- the time-taken for customers to arrive at decisions. Moreover, the user must be physically in the store in order to avail of this service.

For example, SpecSavers Optical Group Ltd. (United Kingdom), offers in-store customers at a number of their sites the kiosk-based opportunity to try on different frame options, while having their picture taken and displayed at the same time. Importantly, the service they provide does not allow users to interact in any other way, and still requires the customer to be physically present in the store. These kiosks do not have access to product data, nor are they capable of helping users by making useful recommendations; they simply take pictures and display these on a screen. Importantly, this photo-kiosk solution is not adequate for use with merchant online-stores, given the absence of any facial analysis and automated product-placement features.

Few online stores allow users to see how they will look wearing different frame options. The crucial point is that even when these features are in place, customers are still frustrated by the enormity of the product space and the lack of guidance. That is, the absence of a sales assistant means that no narrowing of product alternatives takes place, and the user is left to their own devices to find their "ideal" pair of glasses. SUMMARY OF THE INVENTION

Accordingly, a method for providing a recommendation to a user is provided. The method includes uploading a user image, processing the user image, and generating at least one image of a product or service for the user based on the processing. The method also includes displaying at least one image of the product or service, accepting and processing a feedback response, and producing an image of at least one type of the product or service based on the feedback response.

In another embodiment of the present invention, the method also includes a step of superimposing the image of at least one type of the product or service on the user image.

In yet another embodiment of the present invention, the feedback response is communicated from the user. In addition, the user may solicit a feedback response from an advisor. In yet another embodiment, the feedback response may be generated and communicated internally.

Another embodiment of the present invention is a system for providing a recommendation to a user. The system includes a user image, a processor to process the user image, a generator to generate an image of a product or service based on an output from the processor, a receiver to receive feedback, a recommender to recommend at least one product or service based on the feedback, and a display producer to produce an image of the at least one product or service based on the feedback.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and other objects and advantages of the present invention will be apparent upon consideration of the following detailed description, taken in conjunction with the accompanying drawings, in which like reference characters refer to like parts throughout, and in which:

FIG. 1 is a block diagram of a customer assistance system in accordance with an embodiment of the present invention;

FIG. 2 is a more detailed block diagram of the Application Layer of FIG. 1 , in accordance with an embodiment of the present invention;

FIG. 3 is a flowchart illustrating the method of an embodiment of the present invention; and

FIG. 3A is a continuation of the flowchart of FIG. 3 illustrating the method of an embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

FIG. 1 of the present invention is a block diagram illustrating the overview of the system architecture of an embodiment of the present invention. For purposes of explanation, the present invention will be described from here on out by referencing a method and system for assisting a customer in choosing a pair of eyeglasses over the internet. This example does not limit the breadth or the scope of the present invention in any way. The present invention may be used to assist customers and users in making a decision on endless types of products and services.

There are three main blocks in this diagram of the overall system 100 including a User Interface Layer 110, an Application Layer 120, and a Dataset Layer 160. Central to the system 100 is Application Layer 120 which in itself has three components, later described in FIG. 2.

Dataset layer 160 stores thousands of types of products and services that that particular system supports. In this example, thousands of types of eyeglasses will be stored in Dataset Layer 160. In one embodiment, each pair of eyeglasses stored in Dataset Layer

160 includes a list of feature values which include, for example, price, temple size, material, model type, etc.

User Interface Layer 1 10 manages messages passed between the user and Application Layer 120. User Interface Layer 110 also displays any recommendations or suggestions made to the user by the system of the present invention or by a third party. In addition, User Interface Layer 110 may display an image of the user "trying on" a pair of eyeglasses. User Interface Layer 110 has the capability to display a superimposed image of a pair of eyeglasses onto an image of the user. With this feature, the user can see an image of him/herself on a screen wearing a particular pair of eyeglasses.

Now turning to FIG. 2, a more detailed version of Application Layer 120 can be seen, in accordance with an embodiment of the present invention. The Application Layer 220 of FIG. 2 includes a Visualization Engine 230, a Recommendation Engine 240, and a Collaborative Manager 250. After a user sees an image of a pair of eyeglasses on User Interface Layer 110 due to the processing mechanisms of Visualization Engine 230, the user can provide feedback to the system so that the system can process the feedback to determine which pair of eyeglasses to show the user next. Feedback from the user or from a third party is accepted and relayed back to Recommender Engine 240. Recommender Engine 240 is able to process the user feedback to output a recommendation to the User Interface Layer 1 10 based on the feedback.

The system of the present invention operates generally by an algorithm that supports Recommender Engine 240 that is interactive with the user. The system also provides users with cyclic feedback opportunities to influence the retrieval of different products and styles.

There are three stages to each cycle. Firstly, there are k cases recommended to the user. In this example, there are A: types of eyeglasses recommended to the user. Second, the user provides preference-based feedback over visual examples of the cases, generated by Visualization Engine 230, by indicating a preference. For example, if the user is given 4 types of eyeglasses to choose from, the user will rank the order of preference from 1 to 4. Third, the system revises the query which represents the user's current needs, using the limited feedback that the user provides.

User feedback can be processed in the present invention with even the most limited of feedback. Even if the user simply states "I like this" or "I do not like this", the system has the intelligence and capability to narrow down product samples according to the user's taste.

There are two major examples of feedback communication strands in the present invention. The first strand is generally individual feedback from the user collected in recommendation cycles. The second strand is generally collective feedback from the advisors to the user. In the second strand option, information with ratings is collected from a set of advisors. The advisors may be internal to the system or they can be friends of the user. The advisors may be people that the user sought out for advice. Advisor and user feedback is collected by Collaborative Manager 250. Collaborative Manager 250 then organizes and evaluates the feedback and accordingly sends the recommended eyeglasses for the user to see images of on User Interface Layer 1 10. It is a user option whether or not to implement the use of advisors.

In addition, the Collaborative Manager 250 includes Keepcart 280, where the user stores styles and recommendations of his/her choice. If the user wants to store an image of a pair of eyeglasses because he/she likes it or thinks that it may be one to view again in the future, the user may store that image in Keepcart 280. Keepcart 280 provides the user with easy access to all of his/her favorite styles.

In one embodiment of the present invention, a user profile is created for each user of the system. The user profile contains information from the session on the system, including user preferences, cases that were rejected, and current query. Not only are preferred styles kept in Keepcart 280, but all reviews associated with those items and styles are also stored.

When the user requests information from the advisors, the advisors are asked to rate all of the items in Keepcart 280. The advisors are shown a review page, which is generated from the data stored in the user profile. For individual reviews, the following equation is used to average and normalize all of the advisor's ratings to generate a Hot-or-Not score: hot_or_not(C) = [(∑ζrating (C)/n))*(10/(Max(rating(P)))] , where rating(C) retrieves a rating for a product case C, and is normalized based on the maximum rating over all product cases (Max(rating(P))). The Hot-or-not rating is a value between 1 and 10, where 10 is the best rating that can be applied to a case. User Interface Layer 110 uses these values to display the appropriate information in the review screens.

When a user asks for a recommendation from the system, first, the product cases must be ranked in decreasing order of their similarity to the current query, Q. Accordingly, the final score is always a number between 0 and 1. The equation is:

SIm(Q₁C) = [(∑(n, i-l)(featureSim(FQi,Fci))]/n, where FQ and Fc are the values for the numeric features being compared.

When calculating similarity at the feature level nominal and numeric values are handled differently. For nominal features, an exact match comparison is carried out, returning the value 1 when the values match, and 0 otherwise. Numeric values, on the other hand, use their relative difference as a basis for similarity calculation. The equation for this is shown below where F_Q and Fc are the values for the numeric features being compared: featύreSim(F_Q,Fc) = 1 - [(\F_Q - Fc\)/max(F_Q,F_c)] .

FIGS. 3 and 3A depict a flow chart 300 illustrating steps of a method in accordance with an embodiment of the present invention.

A user logs into the system of the present invention, at step 305, and uploads an image of themselves 310. In this example, the system is implemented for selecting eyeglasses, but the system is not limited to such example. The system could be used for selecting types and processes of cosmetic surgery, for selecting from various hairstyles, etc. The user could also input a description of themselves using words, however a digital image of the user would make the process more accurate. Because this is an example of a system for selecting eyeglasses, the user would upload a digital image of the user's face. The remainder of user's body would be irrelevant. In addition, the user may provide some basic feature preferences, like price, shape, gender.

Because this example relates to selecting eyeglasses, the user has the option of adding more preferences to narrow down the results at step 315. If the user wants to add features, the user inputs basic feature preferences at step 320. Here, the user inputs their initial preferences regarding eyeglasses. For example, the user may input whether they preferred metal frames or plastic, large or small, dark or light, etc. In step 330, a resulting query is formed, which is made up of the user's dimensions. In this example, the resulting query would include the location of the user's eyes, and the distance between the pupils, for example.

The user is presented with a recommendation screen in step 340. On the recommendation screen, the user will see several options. In this example, the user will be presented with four images of four different types of eyeglasses. In step 350, the user reviews the recommended eyeglasses by either viewing the whole feature description of the eyeglasses or by using a "Try it On" feature of the present invention. If the user selects at step 355 to "Try it On", then the user moves on to step 360, otherwise, step 365. At step 360, the "Try it On" feature allows the user to see how any of the recommended items would look on the uploaded image. The user may indicate which option they prefer by clicking a "More Like This" link associated with their choice. This link will bring up a new set of recommendations, those most similar to the user's previous preference. The user can also view the associated feature descriptions for each product by clicking the "Show Description" link.

During the recommendation session, the user has the ability to place any of the recommended items in user Keepcart 280. At any stage during the session, shown at step 365, the user has the option of requesting opinions of their friends regarding the items in Keepcart 280. The user may ask their friends via an email, which may be pregenerated by the system of the present invention. This email message would ask that the friend log into the system and provide feedback over all the items in the current keepcart 280.

In one embodiment, an advisor is presented with the review screen. Here, the advisor is prompted to put in rating scores for each of the images shown. Each image would show the original user wearing a pair of eyeglasses (in this example) which is in their Keepcart 280. When the advisor has entered in the ratings, the advisor will proceed to click a SUBMIT button and log out of the system. When the user logs back into the system, the user can make a more informed decision in choosing a product at step 380, in this case, a pair of eyeglasses, because there will be a series of reviews on the user's screen. The reviews will include images of the eyeglasses as well as the Hot-or-Not rating. These Hot- or-Not ratings will give an overall summary of the reviews made. The user also has an option of viewing a more detailed review summary screen where the details of all of the reviews made on the items are displayed in a grid format, so it is simple to see which person gave a particular review for a pair of eyeglasses.

Although preferred embodiments of the present invention have been disclosed in a discussion of choosing eyeglasses for a customer, persons skilled in the art will appreciate that the system and method of the present invention may be used for many other products and services without departing from the spirit of the invention as shown. Persons skilled in the art also will appreciate that the present invention can be practiced by other than the specifically described embodiments. The described embodiments are presented for purposes of illustration and not of limitation, and the present invention is limited only by the claims which follow.

Claims

WHAT IS CLAIMED IS:

1. A method for providing a recommendation to a user, the method comprising: uploading a user image; processing said user image; generating at least one image of a product or service for said user based on said processing; displaying said at least one image of said product or service; accepting a feedback response; processing said feedback response; and producing an image of at least one type of said product or service based on said feedback response.

2. The method of claim 1, further comprising superimposing said image of at least one type of said product or service on said user image.

3. The method of claim 1, wherein said feedback response is communicated from said user.

4. The method of claim 1 wherein said user may solicit a feedback response from an advisor.

5. The method of claim 4, wherein said feedback response is communicated from said advisor.

6. The method of claim 1, wherein said feedback response is generated and communicated internally.

7. The method of claim 1 wherein said user image is a facial image of said user.

8. The method of claim 1 wherein said user image is a body image of said user.

9. The method of claim 1 wherein said processing of said user image includes a feature extraction of said user image.

10. The method of claim 1 wherein said feedback includes preference-based feedback.

11. A system for providing a recommendation to a user, the system comprising: a user image; a processor to process said user image; a generator to generate an image of a product or service based on an output from said processor; a receiver to receive a feedback response; a recommender to recommend at least one product or service based on said feedback response; and a display producer to produce an image of said at least one product or service based on said feedback. .

12. The system of claim 11 further comprising an applier to superimpose said at least one type of said product or service on said user image.

13. The system of claim 11 wherein said feedback response is communicated from said user.

14. The system of claim 1 1 wherein said user may solicit a feedback response from an advisor.

15. The system of claim 14 wherein said feedback response is communicated from said advisor.

16. The system of claim 11, wherein said feedback response is generated and communicated internally.

17. The system of claim 11 wherein said user image is a facial image of said user.

18. The system of claim 11 wherein said user image is a body image of said user.

19. The system of claim 1 1 wherein said processor extracts at least one feature of said user.

20. The system of claim 11 wherein said feedback includes preference-based feedback.