Deep learning methods have shown to be particularly effective in inferring the sentiment polarity of a text snippet. However, in cross-domain and cross-lingual scenarios there is often a lack of training data. To tackle this issue, propagation algorithms can be used to yield sentiment information for various languages and domains by transferring knowledge from a source language(usually English). To propagate polarity scores to the target language, these algorithms take as input an initial vocabulary and a bilingual lexicon. In this paper we propose to enrich lexicon in-formation for cross-lingual propagation by inferring the bilingual semantic relationships from an aligned bilingual vector space.This allows us to exploit the underlying text similarities that are not made explicit by the lexicon. The experiments show that our approach outperforms the state-of-the-art propagation method on multilingual datasets.

Cross-Lingual Propagation of Sentiment Information Based on Bilingual Vector Space Alignment / Giobergia, Flavio; Cagliero, Luca; Garza, Paolo; Baralis, Elena. - ELETTRONICO. - (2020). (Intervento presentato al convegno Data Analytics solutions for Real-LIfe APplications (DARLI-AP). 2020 Workshops of the EDBT/ICDT Joint Conference, EDBT/ICDT-WS 2020).

Cross-Lingual Propagation of Sentiment Information Based on Bilingual Vector Space Alignment

giobergia,flavio;cagliero,luca;garza,paolo;baralis,elena
2020

Abstract

Deep learning methods have shown to be particularly effective in inferring the sentiment polarity of a text snippet. However, in cross-domain and cross-lingual scenarios there is often a lack of training data. To tackle this issue, propagation algorithms can be used to yield sentiment information for various languages and domains by transferring knowledge from a source language(usually English). To propagate polarity scores to the target language, these algorithms take as input an initial vocabulary and a bilingual lexicon. In this paper we propose to enrich lexicon in-formation for cross-lingual propagation by inferring the bilingual semantic relationships from an aligned bilingual vector space.This allows us to exploit the underlying text similarities that are not made explicit by the lexicon. The experiments show that our approach outperforms the state-of-the-art propagation method on multilingual datasets.
File in questo prodotto:
File Dimensione Formato  
DARLIAP1.pdf

accesso aperto

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Creative commons
Dimensione 1.08 MB
Formato Adobe PDF
1.08 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2846234