[jira] [Resolved] (STANBOL-1070) Entity Co-Mention Engine

["Rupert Westenthaler (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/STANBOL-1070?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rupert Westenthaler resolved STANBOL-1070.
------------------------------------------

       Resolution: Fixed
    Fix Version/s: 0.12.0

marking this as resolved by excluding the feature to support co-mentions for alternate labels. This feature was moved to STANBOL-1179

> Entity Co-Mention Engine
> ------------------------
>
>                 Key: STANBOL-1070
>                 URL: https://issues.apache.org/jira/browse/STANBOL-1070
>             Project: Stanbol
>          Issue Type: New Feature
>          Components: Enhancement Engines
>            Reporter: Rupert Westenthaler
>            Assignee: Rupert Westenthaler
>             Fix For: 0.12.0
>
>
> Entity Co-Mention Engine
> ======
> The goal of this engine is to extract co-mentions of Entities already detected for an document. The typical example are persons only mentioned by their family name after an initial mention with the full name e.g.
>     ... Barack Obama gave a talk to members of the Labor Union ... Obama specially mentioned ... 
> But also alternate names used to refer to Entities might be used for extracting co-mentions.
> NOTE: that this Engine does not use NLP level co-reference (e.g. linking a Pronoun with the Entity it stands for).
> Implementation
> -----
> This Engine will be implemented based on existing Entity linking functionality as implemented by the EntityLinkingEngine. The main difference is that an in-memory EntityMentionIndex will be used as controlled vocabulary to link against. This EntityMentionIndex will implement the EntitySearcher interface as used by the EntityLinkingEngine to search for Entities. 
> The EntityMentionIndex will contain both fise:TextAnnotations (such as NamedEntities) as well as fise:EntityAnnotations (entity suggested for fise:TextAnnotations).
> Writing results of the co-mention extraction will involve
> * creating new fise:TextAnnotations with suggested fise:EntityAnnotation (e.g. for additional mentions not previously detected by any other engine)
> * modification of existing Suggestions for fise:TextAnnotations (e.g. if 'Sevenson' was linked with "Svenson" (http://rdf.freebase.com/ns/m.0n5rh_s), a fictional character from the 1930 film The Silver Horde but "Peter Svenson" (http://rdf.freebase.com/ns/m.05wxvv9) an Author was already earlier mentioned in the document - the later would be added as additional suggestion to an existing TextAnnotation and also confidence values would be adapted accordingly.
> * creation of relations between enhancements to express entity co-mentions (most likely dc:relation from the co-mention to the initial mention of an Entity. 



--
This message was sent by Atlassian JIRA
(v6.1#6144)