17/01/2020
The Human Resources Strategy for Researchers

A Study of transfer learning approaches for entity linking on new domain and generated synthetic data

This job offer has expired


  • ORGANISATION/COMPANY
    CNRS
  • RESEARCH FIELD
    Engineering
    Physics
    Technology
  • RESEARCHER PROFILE
    First Stage Researcher (R1)
  • APPLICATION DEADLINE
    07/02/2020 23:59 - Europe/Brussels
  • LOCATION
    France › ORSAY
  • TYPE OF CONTRACT
    Temporary
  • JOB STATUS
    Full-time
  • HOURS PER WEEK
    35
  • OFFER STARTING DATE
    01/03/2020

LIMSI, Orsay
Groupe ILES
Projet PSPC Aîda

The thesis topic is about information extraction form text, it consists on several steps :
— locate in the texts the elements of interest, and to make the link with the ontologies,
check if they are appropriate, or if others are needed,
— extract the entities
— identify the links between the entities (coreference resolution)
— link these entities and their relations to the properties of the knowledge base in relation
to the type of analyzed texts (entity linking)
Entity Linking (EL) is the task of recognizing (cf. Named Entity Recognition (NER)) and
disambiguating (Named Entity Disambiguation (ED)) named entities to a knowledge base (i.g.
Wikidata, DBpedia, or YAGO). It is sometimes also simply known as Named Entity Recogni-
tion and Disambiguation. Different approaches for EL have been proposed, that can be fell into
three categories : pipeline approach [BP06, HYB + 11, CLO + 13] in which the two sub-problems
(NER and ED) are trained independently, however the dependency between the two steps is
ignored and errors made by the NER system will propagate and have bad impact on ED sys-
tem. Joint approach [NTW16, LHLN15] that consists on jointly model NER and linking tasks
and capture the mutual dependency between them. Last, end-to-end approach [KGH18], that
consists on processing a piece of text to extract the entities (i.e. NER) and then disambiguate
these extracted entities to the correct entry in a given knowledge base (i.g. Wikidata, DBpedia,
YAGO). The EL can be seen as a disambiguation-only task, that directly takes gold standard
named entities as input and only disambiguates them to the correct entry in a given knowledge
base (disambiguation-only) [RR18, RTV18, LT18, HYB + 11].
The state of the art approaches for EL, are based on neural architectures trained with
supervised manner. This kind of approaches needs a lot of annotated data to get good perfor-
mance. However, these labeled data are often expensive, and sometimes impossible to collect.
Transfer learning is one of the approaches we can explore, that addresses the problem of how
to utilize plenty of labeled data in a source domain to solve related but different problems in
a target domain, even when the training and testing problems have different distributions or
features [LZWJ17, PKY + 08, PY09, EOM18]. This thesis topic proposes to answer the following
research questions :
— Is the use of transfer learning with EL beneficial in the new domain, especially when
using generated synthetic data ?
— What kind of existing approach should be used for transfer learning in EL problem ?

Required Research Experiences

  • RESEARCH FIELD
    Engineering
  • YEARS OF RESEARCH EXPERIENCE
    None
  • RESEARCH FIELD
    Physics
  • YEARS OF RESEARCH EXPERIENCE
    None
  • RESEARCH FIELD
    Technology
  • YEARS OF RESEARCH EXPERIENCE
    None

Offer Requirements

  • REQUIRED EDUCATION LEVEL
    Engineering: Master Degree or equivalent
    Physics: Master Degree or equivalent
    Technology: Master Degree or equivalent
  • REQUIRED LANGUAGES
    FRENCH: Basic
Work location(s)
1 position(s) available at
Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur
France
ORSAY

EURAXESS offer ID: 482169
Posting organisation offer ID: 13867

Disclaimer:

The responsibility for the jobs published on this website, including the job description, lies entirely with the publishing institutions. The application is handled uniquely by the employer, who is also fully responsible for the recruitment and selection processes.

 

Please contact support@euraxess.org if you wish to download all jobs in XML.