Hello,
I believe there might be a problem with the Advanced Search's related topic suggestion system.
When creating a new topic heading, the system seems to prioritize semantic/content similarity between headings rather than direct heading similarity.
Example:
Current topic heading:
"ViPSrO Online 120 Cap Silkroad PvP Server"
While writing:
"Vipsro"
The suggestion system for similar topics does not suggest the current ViPSrO topic for the following reason:
*The same words appear in the subject heading.
* A standard search will find the topic quickly and accurately.
Instead, the system only suggests other PvP server topics that share similar content patterns, for example:
This behavior makes duplicate content detection unreliable in communities where the topic/project/server name is the primary identifying element.
It appears that the recommendation algorithm places too much weight on semantic/full text similarity and insufficient weight on exact or partial matches with the subject line.
In a related topic suggestion system, is it possible to prioritize topic headings that exactly match before using semantic similarity scoring?
Thank you.
I believe there might be a problem with the Advanced Search's related topic suggestion system.
When creating a new topic heading, the system seems to prioritize semantic/content similarity between headings rather than direct heading similarity.
Example:
Current topic heading:
"ViPSrO Online 120 Cap Silkroad PvP Server"
While writing:
"Vipsro"
The suggestion system for similar topics does not suggest the current ViPSrO topic for the following reason:
*The same words appear in the subject heading.
* A standard search will find the topic quickly and accurately.
Instead, the system only suggests other PvP server topics that share similar content patterns, for example:
- 110 covers
- EU/Switzerland
- progressive
- balanced
- play2win
This behavior makes duplicate content detection unreliable in communities where the topic/project/server name is the primary identifying element.
It appears that the recommendation algorithm places too much weight on semantic/full text similarity and insufficient weight on exact or partial matches with the subject line.
In a related topic suggestion system, is it possible to prioritize topic headings that exactly match before using semantic similarity scoring?
Thank you.
Last edited by a moderator: