Hashtag Recommendation via Sentiment Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for recommending hashtags on social media platforms are either limited to past usage patterns or computationally intensive, failing to consider sentiment and resulting in inappropriate hashtag recommendations.
Innovation Solution
A method that analyzes posts using natural language processing to determine keywords, categorizes previous posts by sentiment, and ranks hashtags based on frequency, engagement, and sentiment to provide contextually relevant recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If hashtags are recommended based on past usage patterns, then the recommendation process is simple, but the recommendations fail to consider sentiment and result in inappropriate hashtag suggestions
Solution Approach 1:
The patent changes the parameters used for hashtag recommendation from solely historical usage patterns to include sentiment analysis dimensions. By introducing sentiment as a new parameter and categorizing posts based on sentiment polarity, the system can recommend hashtags that are not only frequently used but also contextually appropriate for the intended message tone, thereby improving reliability while maintaining operational simplicity through automated sentiment-based filtering
Solution Approach 2:
The patent introduces sentiment analysis as an intermediary layer between the post content and hashtag recommendation. By first determining the sentiment polarity of the post and then using this information to filter and rank hashtags, the system creates a mediating mechanism that ensures hashtags are selected based on both usage frequency and sentiment compatibility, resolving the contradiction between simple operation and reliable recommendations
2Reliability
If sentiment analysis is performed on all previous posts, then appropriate hashtags can be recommended, but the computational process becomes intensive
Solution Approach 1:
The patent extracts only the essential sentiment information from previous posts rather than processing all post data comprehensively. By determining sentiment polarity (positive, negative, or neutral) as a binary or ternary classification and using this extracted sentiment label for hashtag filtering, the system reduces computational complexity while still achieving reliable sentiment-based hashtag recommendations
Solution Approach 2:
The patent applies partial action by performing sentiment analysis only on posts that are relevant to the current context or by using pre-computed sentiment data from a subset of posts. This selective approach allows the system to maintain high recommendation accuracy without the computational burden of analyzing every single previous post, thereby balancing reliability with energy efficiency
3Productivity
If hashtags are selected based on frequency of appearance, then the recommendation process is efficient, but it does not account for engagement metrics or sentiment context
Solution Approach 1:
The patent segments the hashtag recommendation process into distinct phases: first filtering hashtags based on frequency and engagement metrics, then further filtering based on sentiment compatibility. By dividing the selection process into sequential stages, the system maintains efficiency in the initial frequency-based filtering while adding sentiment context in a subsequent step, thereby improving overall reliability without significantly reducing productivity
Solution Approach 2:
The patent introduces dynamic weighting of different selection criteria based on context. The system can adjust the relative importance of frequency, engagement metrics, and sentiment alignment depending on the specific post characteristics and target audience. This dynamic approach allows the system to optimize between efficiency and relevance by emphasizing different factors based on real-time context rather than using a fixed ranking scheme
Data Source
AI summary
A method for recommending hashtags includes determining keywords from a post planned for publishing by a publisher. An input criteria comprising at least one of age group, geographical location, date range, or a keyword is received. Previous posts associated with the keywords and satisfying the input criteria are obtained. The previous posts are categorized into one or more categories based on sentiment of each post and for each category hashtags used in the obtained previous posts in that category are determined. The hashtags are ranked based on predefined criteria comprising at least one of frequency of appearance of respective hashtag in posts, number of likes or shares or retweets of post comprising respective hashtag, number of followers of person who used respective hashtag, or sentiment of post comprising respective hashtag. The hashtags are then recommended, based on ranking, to the publisher for use with the post planned for publishing.