Local Word Embeddings for Query Expansion based on Co-Authorship and Citations

André  Rattinger; Jean-Marie Le Goff; Christian Gütl

Local Word Embeddings for Query Expansion based on Co-Authorship and Citations

André Rattinger, Jean-Marie Le Goff, Christian Gütl

Institute of Interactive Systems and Data Science (7060)

Research output: Contribution to conference › Paper › peer-review

Abstract

Word embedding techniques have gained a lot of interest
from natural language processing researchers recently and they are valuable resource in identifying a list of semantically related terms for a
search query. These related terms build a natural addition for query expansion, but might mismatch when the application domains use different
jargon. Using the Skip-Gram algorithm of Word2Vec, terms are selected
only from a specific subset of the corpus, which is extended by documents
from co-authorship and citations. We demonstrate that locally-trained
word embeddings with this extension provides a valuable augmentation
and can improve retrieval performance. First result suggest that query
expansion and word embeddings could also benefit from other related
information.

Original language	English
Pages	46
Number of pages	53
Publication status	Published - 26 Mar 2018
Event	Bibliometric-enhanced Information Retrieval - Frenoble, France Duration: 26 Mar 2018 → 26 Mar 2018 Conference number: 7 https://www.gesis.org/en/services/events/events-archive/conferences/ecir-workshops/ecir-workshop-2018/

Workshop

Workshop	Bibliometric-enhanced Information Retrieval
Abbreviated title	BIR 2018
Country/Territory	France
City	Frenoble
Period	26/03/18 → 26/03/18
Internet address	https://www.gesis.org/en/services/events/events-archive/conferences/ecir-workshops/ecir-workshop-2018/

Fields of Expertise

Information, Communication & Computing

Access to Document

http://ceur-ws.org/Vol-2080/paper5.pdfLicence: Unspecified

Cite this

@conference{0d7a9e99aec54ff8b6d8a0b9e7a8ec91,

title = "Local Word Embeddings for Query Expansion based on Co-Authorship and Citations",

abstract = "Word embedding techniques have gained a lot of interestfrom natural language processing researchers recently and they are valuable resource in identifying a list of semantically related terms for asearch query. These related terms build a natural addition for query expansion, but might mismatch when the application domains use differentjargon. Using the Skip-Gram algorithm of Word2Vec, terms are selectedonly from a specific subset of the corpus, which is extended by documentsfrom co-authorship and citations. We demonstrate that locally-trainedword embeddings with this extension provides a valuable augmentationand can improve retrieval performance. First result suggest that queryexpansion and word embeddings could also benefit from other relatedinformation.",

author = "Andr{\'e} Rattinger and {Le Goff}, Jean-Marie and Christian G{\"u}tl",

year = "2018",

month = mar,

day = "26",

language = "English",

pages = "46",

note = "Bibliometric-enhanced Information Retrieval, BIR 2018 ; Conference date: 26-03-2018 Through 26-03-2018",

url = "https://www.gesis.org/en/services/events/events-archive/conferences/ecir-workshops/ecir-workshop-2018/",

}

TY - CONF

T1 - Local Word Embeddings for Query Expansion based on Co-Authorship and Citations

AU - Rattinger, André

AU - Le Goff, Jean-Marie

AU - Gütl, Christian

N1 - Conference code: 7

PY - 2018/3/26

Y1 - 2018/3/26

N2 - Word embedding techniques have gained a lot of interestfrom natural language processing researchers recently and they are valuable resource in identifying a list of semantically related terms for asearch query. These related terms build a natural addition for query expansion, but might mismatch when the application domains use differentjargon. Using the Skip-Gram algorithm of Word2Vec, terms are selectedonly from a specific subset of the corpus, which is extended by documentsfrom co-authorship and citations. We demonstrate that locally-trainedword embeddings with this extension provides a valuable augmentationand can improve retrieval performance. First result suggest that queryexpansion and word embeddings could also benefit from other relatedinformation.

AB - Word embedding techniques have gained a lot of interestfrom natural language processing researchers recently and they are valuable resource in identifying a list of semantically related terms for asearch query. These related terms build a natural addition for query expansion, but might mismatch when the application domains use differentjargon. Using the Skip-Gram algorithm of Word2Vec, terms are selectedonly from a specific subset of the corpus, which is extended by documentsfrom co-authorship and citations. We demonstrate that locally-trainedword embeddings with this extension provides a valuable augmentationand can improve retrieval performance. First result suggest that queryexpansion and word embeddings could also benefit from other relatedinformation.

M3 - Paper

SP - 46

T2 - Bibliometric-enhanced Information Retrieval

Y2 - 26 March 2018 through 26 March 2018

ER -

Local Word Embeddings for Query Expansion based on Co-Authorship and Citations

Abstract

Workshop

Fields of Expertise

Access to Document

Fingerprint

Cite this