Embedding Projection for Targeted Cross-lingual Sentiment: Model Comparisons and a Real-World Study

Abstract

Sentiment analysis benefits from large, hand-annotated resources in order to train and test machine learning models, which are often data hungry. While some languages, e. g., English, have a vast array of these resources, most under-resourced languages do not, especially for fine-grained sentiment tasks, such as aspect-level or targeted sentiment analysis. To improve this situation, we propose a cross-lingual approach to sentiment analysis that is applicable to under-resourced languages and takes into account target-level information. This model incorporates sentiment information into bilingual distributional representations, by jointly optimizing them for semantics and sentiment, showing state-of-the-art performance at sentence-level when combined with machine translation. The adaptation to targeted sentiment analysis on multiple domains shows that our model outperforms other projection-based bilingual embedding methods on binary targeted sentiment tasks. Our analysis on ten languages demonstrates that the amount of unlabeled monolingual data has surprisingly little effect on the sentiment results. As expected, the choice of a annotated source language for projection to a target leads to better results for source-target language pairs which are similar. Therefore, our results suggest that more efforts should be spent on the creation of resources for less similar languages to those which are resourcerich already. Finally, a domain mismatch leads to a decreased performance. This suggests resources in any language should ideally cover varieties of domains.1

skweak: Weak Supervision Made Easy for NLP

“…However, except for zero-shot learning approaches (Artetxe and Schwenk, 2019;Barnes and Klinger, 2019;Pires et al, 2019), they still require some amounts of labelled data from the target domain to fine-tune the neural models to the task at hand. The skweak framework (pronounced /skwi:k/) is a new Python-based toolkit that provides solutions to this scarcity problem.…”

Evaluating morphological typology in zero-shot cross-lingual transfer

“…Cross-lingual transfer has become ubiquitous in recent years, including cross-lingual POS tagging (Täckström et al, 2013;Huck et al, 2019) and cross-lingual sentiment analysis (Mihalcea et al, 2007;Balahur and Turchi, 2014;Barnes and Klinger, 2019). While earlier research focused on annotation projection (Yarowsky et al, 2001;Banea et al, 2008) or cross-lingual embeddings (Kim et al, 2017;Artetxe et al, 2017;Barnes et al, 2018b), multi-lingual pretraining currently leads to state-of-the-art results (Devlin et al, 2019;Lample and Conneau, 2019).…”

Emotional Expression and Information Communication in English Texts Based on Artificial Intelligence Technology

“…In expressing emotion and communicating information in English text, it must be understood that translation is not only the conversion of language corpora and semantic translation but also the translation of cultural behavior, social behavior, ideological understanding and so on [6][7]. All major Internet companies have also launched their online education platforms, and students receiving education are no longer limited to learning in the traditional mode but can learn independently through online education platforms [8][9][10]. At present, the quality of online education platforms varies, and the topic of studying the quality of education on online platforms has become very meaningful, while the quality of that online education platform can be derived by analyzing the sentiment of the teaching evaluation of the education platform [11][12].…”