For topics that run longer than 50 posts, there is a button at the bottom of the OP’s post allows you to zero-in on only the “top” posts. Presumably, this is a meaningful summary of the long thread.
Do you ever use it?
- I rely on it
- I use it sometimes
- I often forget it’s there
- Huh? There’s a “what” to do “what”?
I wondered how the Top Replies are chosen. According to this link on Discourse Meta…
…everything is in the ScoreCalculator class
In case you are curious, here is the code used by the ScoreCalculator to calculate a post’s score
And here is a summary of how that code works
The primary objective of this code is to calculate scores for posts and topics, and to rank posts within topics based on these scores. The scores and rankings are updated in the database, indicating that the system uses these metrics to surface or prioritize content for users.Here is a simplified breakdown of how posts are scored and ranked:
Initialization with Default Weights: The
ScoreCalculatorclass initializes with a set of default score weights for different engagement metrics such as replies, likes, incoming links, bookmarks, and reads. These weights are customizable and define the importance of each metric in the overall score calculation.Scoring Posts Based on User Engagement: Imagine each post on the platform as a contestant in a popularity contest. The score of each post is determined by how much attention and interaction it receives from the audience (users). The interactions include how many times the post is replied to, liked, linked to, bookmarked, and read. Each type of interaction is given a certain number of points. For example, likes might be worth more points than reads, reflecting their perceived value in measuring engagement. The final score of a post is calculated by adding up all the points from these interactions. If an interaction hasn’t occurred for a post (for example, if nobody has bookmarked it), that interaction contributes zero points to the total score. This way, each post gets a score that reflects how much it has engaged the audience, with higher scores indicating more engagement.
Batch Update Process: The update process is designed to handle potentially large numbers of posts by processing them in batches of up to 20,000 at a time. This approach helps manage system performance and resource utilization.
Ranking Posts within Topics: After scoring, posts are ranked within their respective topics based on their scores, from highest to lowest. The ranking is expressed as a percentage rank, which indicates the relative position of a post among all posts within the same topic.
Topic Scoring and Summary Flagging: In addition to scoring individual posts, the system also calculates an average score for each topic and identifies whether a topic meets certain criteria to have a summary. This is determined by comparing the topic’s like count, post count, and the maximum score of its posts against predefined thresholds.
Conditional Updates: The system updates scores, ranks, and summary flags only if they have changed, or if they haven’t been set before. This condition helps to minimize unnecessary database writes.
Topic Filtering Options: The scoring and ranking processes support optional filters that can be applied to topics, such as filtering by the age of the topic or the number of posts in the topic. This allows for more granular control over which topics are included in the scoring and ranking updates.
In essence, the
ScoreCalculatorclass is a sophisticated tool for dynamically evaluating user engagement with content and adjusting the visibility and prioritization of that content accordingly. This approach enables the platform to highlight popular, engaging posts and topics, potentially improving the user experience by ensuring that users are more likely to see content that is relevant and interesting to them.