|
Building a Reference Base for News Metadata: Joint Entity Recognition
and Linking with Domain Adaptation| old_uid | 10449 |
|---|
| title | Building a Reference Base for News Metadata: Joint Entity Recognition
and Linking with Domain Adaptation |
|---|
| start_date | 2011/11/25 |
|---|
| schedule | 11h-13h |
|---|
| online | no |
|---|
| summary | In the applicative context of news production, structuring and
enrichment, we attempt to provide a complete chain of information
extraction, from raw text to the creation and dynamic population of a
metadata reference base. The identification and referencing process of
relevant metadata is done through the use of named entity recognition
(NER) and entity linking (EL) techniques, which are applied jointly. As
opposed to the general case, where NER doesn't access the referential
level of information, and where EL is applied on gold entity
recognition, EL must here be achieved on the basis of noisy detection of
entity mentions. Moreover, we intend to carry out EL by using domain
specific knowledge, while most EL systems use Wikipedia as their
knowledge base.
Keywords: information extraction, entity linking, knowledge base
population |
|---|
| responsibles | Candito |
|---|
| |
|