Content Recommendation via Group Approximation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information retrieval methods for recommending content, such as game software, on user terminals like cellular phones often fail to provide unexpected and entertaining options when user intentions are vague, leading to a lack of appeal and entertainment in search results.
Innovation Solution
An information retrieval method that uses approximation calculations based on user positional information and content classification information to recommend contents from groups that are not only closest to the user's preferences but also from relatively approximate groups, reducing computational load by focusing on group representatives rather than individual content calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the server mechanically matches downloadable game softwares with data relating to the preference of the user and recommends game softwares with high matching in order, then the matching precision with user preference is improved, but the entertainment value and unexpectedness of recommended game softwares deteriorates
Solution Approach 1:
The patent applies partial action by selecting only a portion of content groups (those within a predetermined rank order) rather than exhaustively calculating all groups. This allows the system to recommend content from both highly matching groups and relatively approximate groups, balancing precision with entertainment value while reducing computational burden.
Solution Approach 2:
The patent changes the parameter of selection criteria from solely based on matching order to include both order and rank. By incorporating rank as an additional parameter, the system can select content from groups that are not the closest match but still fall within an acceptable range, thereby increasing unexpectedness and entertainment value.
2Measurement precision
If the server calculates approximation between user positional information and group positional information for all content groups, then the recommendation precision is improved, but the computational load increases
Solution Approach 1:
The patent segments the content database into multiple groups and calculates approximation only for representative positions of each group rather than all individual contents. This segmentation approach maintains recommendation precision at the group level while significantly reducing computational load.
Solution Approach 2:
The patent introduces group positional information as an intermediary between user positional information and individual content items. By first calculating approximation between user and group representatives, then selecting contents from approved groups, the system reduces the number of calculations needed while maintaining recommendation quality.
3Productivity
If the server recommends contents closer to the user's intention in order, then the efficiency of information specification is improved, but the entertainment value and unexpectedness deteriorates
Solution Approach 1:
The patent makes the recommendation system dynamic by incorporating both the closest matching groups and relatively approximate groups within a predetermined rank. This dynamic approach allows the system to adapt between efficiency (closest matches) and entertainment (unexpected matches) based on the user's vague intention, rather than following a fixed ordering.
Data Source
AI summary
Each of information on user characteristic and information on game content characteristic is arranged as position information using two or more parameters on the same coordinate system. The contents are divided into a plurality of content groups. Upon reception of an information search request from a user, a distance between the position corresponding to the user characteristic and the position of the center of gravity of each content group is obtained unless the target content condition is decided. When the target content condition is decided, a distance between the segment connecting the user position with the content position selected as the target and the position of center of gravity of each content group is obtained. Not only the content contained in the group approximating the user characteristic or the target content but also the content contained in a comparatively approximating group is presented to the user as a selection candidate according to the obtained distance.