User Profile

The implementation of an information filtering system requires the creation of a user model, and subsequently a user profile.

User modeling is defined as follows:

User modeling is a discipline that deals with both how information about the user can be acquired and used by an automated system

The description of what information is of interest to a user is commonly referred to as a user profile. The representation space of a user profile is necessarily abstract because of storage limitations and because the information about a user is limited. The representation also has to be reasonably compact in terms of memory and complexity to allow a system to apply a user profile effectively. This means that the user profile can only be an approximation of the real user’s interests. When used in the context of information filtering a user profile contains a representation of the user’s interest in the filtering domain. Because the usage of a user profile is dependent on the type of information filtering system it will not be described here but in later sections.

When considering how information can be acquired from the user a distinction can be made between an explicit model and an implicit model. An explicit model is constructed by the user who provides the system with his interests directly. An implicit model is constructed by the system itself on the basis of feedback it receives from the user. These observations provide evidence about the user’s interests in specific items. Machine learning techniques can be used to combine this evidence with information about the items (acquired either from content or by annotations from users) to create an implicit user model.

User Feedback

The information need of a person on different topics normally changes over time. In addition, the interests of a user are not always known beforehand such as in the case of implicit modeling. The user profile must therefore be able to adapt to changes in the user’s actual interests. Adaptation can be based on explicit or implicit feedback. Explicit feedback requires the user to evaluate examined items on a scale. In every day life explicit ratings are used frequently. A book review for example provides a rating in a textual form. In order to compare ratings a discrete scale is often used, such as star ratings for restaurants or marks out of ten for movies.

Information filtering systems that incorporate explicit ratings have some disadvantages. First, it requires more involvement from the user which increases the cognitive load on the user. Furthermore, if a user does not perceive some benefit from rating he might cease evaluating items or even leave the system. Second, a numerical value may not be a good representation for the user’s perception of an item. People often differ in rating items and are not always consistent in their use of criteria to rate items.

In implicit feedback the user’s interests are inferred by observing the user’s actions, which is more convenient for the user but more difficult to implement. Amazon.com for example uses purchase information from users to make recommendations. When the entry of a book is displayed other book titles are shown which were purchased by customers who also bought the selected book. The time that a user spends on reading an article could also be considered as implicit feedback. The problem with this approach is that it is not possible to determine whether the user is actually reading the article or has taken a break.

Potential types of implicit feedback

Follow