Bilingual Sentiment Embeddings: Joint Projection of Sentiment Across Languages

Abstract

Sentiment analysis in low-resource languages suffers from a lack of annotated corpora to estimate high-performing models. Machine translation and bilingual word embeddings provide some relief through cross-lingual sentiment approaches. However, they either require large amounts of parallel data or do not sufficiently capture sentiment information. We introduce Bilingual Sentiment Embeddings (BLSE), which jointly represent sentiment information in a source and target language. This model only requires a small bilingual lexicon, a source-language corpus annotated for sentiment, and monolingual word embeddings for each language. We perform experiments on three language combinations (Spanish, Catalan, Basque) for sentencelevel cross-lingual sentiment classification and find that our model significantly outperforms state-of-the-art methods on four out of six experimental setups, as well as capturing complementary information to machine translation. Our analysis of the resulting embedding space provides evidence that it represents sentiment information in the resource-poor target language without any annotated data in that language.

Learning Bilingual Sentiment-Specific Word Embeddings without Cross-lingual Supervision

“…There are also approaches to learn sentimental embeddings in the bilingual space without any sentiment resources in the target language. Barnes et al (2018) jointly minimized an alignment objective based on a seed dictionary, and a classification objective based on the sentiment corpus. Its performance is compared to our method in Section 4.…”

“…We use the multilingual sentiment dataset provided by (Barnes et al, 2018). It contains annotated hotel reviews in English (EN), Spanish (ES), Catalan (CA) and Basque (EU).…”

A Computational Linguistic Study of Personal Recovery in Bipolar Disorder

“…Machine learning methods were found to usually outperform rule-based approaches based on look-ups in dictionaries such as LIWC. Again, most annotated resources are English, but state of the art approaches based on multilingual em-beddings allow transferring models between languages (Barnes et al, 2018a).…”

Neural and Linear Pipeline Approaches to Cross-lingual Morphological Analysis

“…Another approach for obtaining morphological analyses for languages without a morphological analyzer is based on transfer learning, which has become a widespread approach in NLP and related disciplines rather recently (Yarowsky et al, 2001;Faruqui and Kumar, 2015;Johnson et al, 2017;Barnes et al, 2018). The general idea is to train a supervised machine learning model that predicts analyses of word forms in a target language using gold-standard analyses that exist in other related languages.…”