[STANBOL-1219] Add option the Entity Co-Mention engine to adapt the confidence of existing Suggestions

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.12.0
• Assignee: Rupert Westenthaler
• Reporter: Rupert Westenthaler
• Created: 22/Nov/13 13:42
• Updated: 30/Nov/13 09:32
• Resolved: 30/Nov/13 09:32

Description

The Entity Co-Mention engine detect repeating mentions of already detected Entities. The typical example is

Barack Obama stated [..] Obama also [..]

But as there exists a City called Obama [1] in Japan an EntityLinking engine linking against a vocabulary containing both Barack Obama AND Obama will link the mention of Obama with the City.

The Co-Mention engine will however detect that the mention of Obama is most likely a co-mention of Barack Obama. Therefore it should also be able to adapt the confidence values of existing suggestion for Obama.

This issue will introduce a new configuration property for the Entity Co-Mention Engine

enhancer.engines.comention.adjustExistingConfidence

This property will take values in the range [0..1). Confidence values of existing suggestions will be multiplied with '1-

{value}

'. Meaning that configuring '0.0' will not change existing confidence values (deactivate this features).

The default value will be set to '0.33'.

This change will be applied to both the trunk and the 0.12 releasing branch

[1] http://en.wikipedia.org/wiki/Obama,_Fukui