Two-model recommender

摘要

A method for recommending choices to a user comprises receiving data. A first set of recommendation choices based on the data is derived using a first model that models a user's subconscious reactions (i.e., intuition) to the data, and a second set of recommendation choices based on the data is derived using a second model that models the user's consciousness. The first and second sets of recommendation choices are blended using a blending function based on characteristics of a domain of the data to create a third set of recommendation choices that are arranged in an order, which is presented to the user. A conscious decision of the user and subconscious reactions of the user are monitored, and data representing the subconscious reactions of the user and data representing the conscious decisions of the user are appended to a first and second memory space, respectively.

主权项

1. A computer-implemented method of information filtering, the method comprising:
receiving data by a recommender system, wherein the recommender system comprises a processor that executes computer-readable instructions to interact with a first memory space that stores data associated with reactions, and a second memory space that stores data associated with decisions, the processor further communicatively coupled to subject-monitoring equipment selected from a group consisting of a sensor, a camera, a microphone, and any combination thereof; deriving, using a first model, a first set of recommendation choices based at least in part on the data; deriving, using a second model, a second set of recommendation choices based at least in part on the data; blending, by the processor of the recommender system, using a blending function, the first set of recommendation choices and the second set of recommendation choices to create a third set of recommendation choices that are arranged in an order, wherein the blending function performs the blending based on:
a number of cues, wherein the blending function places more weight on the first set of recommendation choices when there is a larger number of cues;subjectivity of the cues, wherein the blending function places more weight on the first set of recommendation choices when there is a larger subjectivity of the cues;a redundancy of the cues, wherein the blending function places more weight on the first set of recommendation choices when there is a high redundancy among the cues;a degree of task certainty, wherein the blending function places more weight on the first set of recommendation choices when there is a low certainty of the tasks;a manner in which the cues are displayed, wherein the blending function places more weight on the first set of recommendation choices if the cues are displayed simultaneously; anda predicted time period for a user to respond, wherein blending function places more weight on the first set of recommendation choices if there is a short predicted time period for the user to respond; presenting the third set of recommendation choices to the user; monitoring a conscious decision made by the user by:
receiving a selection from the third set of recommendation choices performed by the user; andmonitoring what was not selected from the third set of recommendation choices by the user; monitoring subconscious reactions of the user with the subject-monitoring equipment, where the subject-monitoring equipment captures a measurement comprising at least one of a physiological measurement, an affective observable measurement, a paralinguistic measurement, and a time measurement, by:
monitoring a subconscious reaction of the user while the user is performing the selection; andmonitoring a subconscious reaction of the user while the user is not performing the selection; appending data representing the subconscious reactions of the user to the first memory space as previous subconscious reactions, wherein the data representing the subconscious reactions is based upon the captured measurement; appending data representing the conscious decisions of the user to the second memory space as previous conscious decisions; and updating at least one of the first model and the second model based upon information from at least one of the first memory space and the second memory space.