Context-Aware Sticker Recommendation via Dialogue Situation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for recommending emoticons or stickers in social network services and instant messengers fail to consider dialogue context and emotional state, often resulting in inappropriate emoticon recommendations, as they rely on one-to-one keyword matching without context or situational analysis.
Innovation Solution
A method that analyzes dialogue situation information by generating metadata from user interactions, including keywords, dialogue acts, and emotional states, to recommend responsive stickers that are contextually appropriate, allowing for personalized and situational responses to the other party's utterances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If one-to-one keyword matching is used to recommend emoticons, then the recommendation process is simple and fast, but the accuracy and appropriateness of emoticon selection deteriorates due to lack of context consideration
Solution Approach 1:
The patent segments the emoticon recommendation process into multiple stages: first extracting keywords from dialogue content, then analyzing dialogue situation and context, and finally selecting appropriate emoticons based on both keywords and context. This segmentation allows the system to maintain simplicity in keyword matching while adding contextual analysis to improve accuracy without completely redesigning the entire system.
Solution Approach 2:
The patent transitions from one-dimensional keyword matching to two-dimensional recommendation by adding the dimension of dialogue situation and context analysis. The system now considers both the literal keyword meaning and the situational context in which the keyword appears, enabling more accurate emoticon selection while maintaining reasonable processing speed through efficient algorithms.
2Ease of manufacture
If keyword-based emoticon matching is implemented, then the system is easy to implement, but it fails to consider dialogue context and emotional state resulting in inappropriate recommendations
Solution Approach 1:
The patent makes the keyword extraction and matching mechanism universal by applying it to multiple types of dialogue situations and contexts. The same keyword extraction framework works across different dialogue types, and the system can adapt to various emotional states and situational contexts by extending the analysis scope rather than creating separate systems for each scenario.
Solution Approach 2:
The system dynamically adjusts its analysis depth and scope based on the dialogue situation. When context indicates a need for deeper analysis (such as emotional state or situational nuance), the system automatically increases the level of context consideration. This dynamic adaptation allows the system to maintain ease of implementation while achieving high adaptability to different dialogue situations.
3Ease of operation
If simple text input is used for emoticon selection, then user operation is convenient, but the system cannot understand emotional state and dialogue context
Solution Approach 1:
The patent implements feedback mechanisms where the system analyzes the dialogue context and emotional state inferred from the text input, then uses this analysis to provide more accurate emoticon recommendations. The system continuously refines its understanding of the dialogue situation based on the interaction flow, allowing simple text input to convey rich emotional and contextual information that the system then utilizes for intelligent recommendation.
Data Source
AI summary
Provided are a method and a computer program of recommending a responsive sticker. The method includes: generating dialog situation information by analyzing pairs of the last utterance of a second user and previous utterances and previous utterances of a first user as an utterance of the second user terminal is inputted into the server; determining a similar situation from a dialog situation information database that is already collected and constructed, using the generated dialog situation information; determining whether it is a turn for the first user terminal to input a response; selecting a responsive sticker candidate group from the determined similar situation when it is a turn for the first user terminal to input the response; and providing information on at least one responsive sticker of the responsive sticker candidate group for the first user terminal.