Kiosk Memory Architecture for Context-Aware Recommendation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Unmanned kiosks face difficulties in providing effective product recommendations due to lack of user interaction and contextual consideration, leading to unsuitable recommendations for middle-aged and older users, as existing recommendation services rely heavily on extensive user history data and do not account for situational factors.
Innovation Solution
A method utilizing memory architecture that receives a current input vector representing context features such as facial identity, emotion, age, and weather, and applies a weight vector to similarity calculations between past and current attributes to recommend products, with a pre-trained model minimizing the difference between recommended and selected products, and efficiently using memory space by updating recently used vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing recommendation services are applied to unmanned kiosks, then product recommendations can be provided, but the service requires lots of training data related to users' histories which unmanned kiosks do not store
Solution Approach 1:
The patent transforms the recommendation system from relying on extensive historical data to using real-time contextual parameters (facial expressions, emotions, surrounding environment) as input features. This parameter transformation enables the system to generate recommendations without requiring large volumes of stored user history data.
Solution Approach 2:
The system pre-trains the neural network model offline with general knowledge about product recommendations. This preliminary training allows the model to make accurate recommendations based on minimal real-time input data during actual kiosk operation, eliminating the need to store extensive user history data.
2Extent of automation
If existing recommendation services are applied to unmanned kiosks, then automated service is achieved, but the recommended content may be unsuitable for users depending on situations
Solution Approach 1:
The patent adds new dimensions to the recommendation system by incorporating environmental context (surrounding situations) and emotional state (facial expressions) as additional input parameters. This multi-dimensional approach enables the automated system to adapt recommendations to specific situations and user emotions.
Solution Approach 2:
The neural network acts as an intermediary that processes multiple input parameters (user facial data, surrounding environment, current context) and transforms them into appropriate product recommendations. This intermediary layer enables the automated kiosk to understand and respond to situational factors.
3Productivity
If conventional kiosk interfaces focus on product ordering goals, then ordering efficiency is improved, but middle-aged and older people have difficulties using the kiosks
Solution Approach 1:
The patent replaces the traditional mechanical interaction interface (buttons, menus, text input) with an emotion recognition and facial expression analysis system. This substitution enables the kiosk to understand user needs automatically through emotional and facial cues, making the system accessible to users who struggle with conventional interfaces.
Solution Approach 2:
The system enables users to interact with the kiosk through natural emotional expressions and facial cues without requiring manual input operations. The kiosk automatically detects and responds to user emotions, eliminating the need for middle-aged and older users to navigate complex menus or perform precise input actions.
Data Source
AI summary
A kiosk for providing a recommendation service according to an embodiment displays an orderer's past ordered product as a recommended product on the screen of the kiosk, the past ordered product read based on a similarity calculation result between a current input attribute representing a contextual feature of a current order status and a past input attribute stored in memory.