Leveraging Linguistic Linked Data for Cross-Lingual Model Transfer in the Pharmaceutical Domain

Abstract

We describe the use of linguistic linked data to support a cross-lingual transfer framework for sentiment analysis in the pharmaceutical domain. The proposed system dynamically gathers translations from the Linked Open Data (LOD) cloud, particularly from Apertium RDF, in order to project a deep learning-based sentiment classifier from one language to another, thus enabling scalability and avoiding the need of model re-training when transferred across languages. We describe the whole pipeline traversed by the multilingual data, from their conversion into RDF based on a new dynamic and flexible transformation framework, through their linking and publication as linked data, and finally their exploitation in the particular use case. Based on experiments on projecting a sentiment classifier from English to Spanish, we demonstrate how linked data techniques are able to enhance the multilingual capabilities of a deep learning-based approach in a dynamic and scalable way, in a real application scenario from the pharmaceutical domain.

LLOD-Driven Bilingual Word Embeddings Rivaling Cross-Lingual Transformers in Quality of Life Concept Detection from French Online Health Communities

“…parallel or aligned corpora as in early work on cross-lingual transfer [6]. In particular, we adopt workflows for using LLOD in cross-lingual transfer learning based on task-informed, bilingual word embeddings (adopted from bilingual sentiment embeddings [7]) presented in [8] and apply them to a different target language (Spanish vs. French), a much more varied task (HRQoL aspect detection vs. sentiment analysis) and different text genre (online health community posts vs. medical experts' interview transcripts).…”

“…Our approach to language-and task-informed transfer learning (LTTL) relies on the framework described in our previous work [8]. Using this architecture based on bilingual word embeddings [7], task-informed bilingual embedding spaces can be learned for any task which can be framed as text classification.…”

“…In this section we describe the relevant lexical resources used in LTTL. A detailed description of the LLOD pipeline used to generate these resources and the individual processing steps involved is presented in [8]. While our focus is on Apertium RDF as a bilingual lexicon in this paper, these workflows have the potential of growing the LLOD cloud over time both in terms of data volume and richness of available resources.…”

“…https://github.com/acoli-repo/acoli-dicts/tree/master/stable/apertium/ apertium-rdf-2020-03-183 https://doi.org/10.5281/zenodo.22056904 This procedure relies on the PoS information that is integrated into Apertium 2.0 via mapping lexical entries to the LexInfo ontology[8].K. Allgaier et al / LLOD-Driven Bilingual Word Embeddings Rivaling Cross-Lingual Transformers 93…”

From Pivots to Graphs: Augmented CycleDensity as a Generalization to One Time InverseConsultation

“…We retain the Fig. 1: Apertium RDF graph (figure taken from [8]), which covers 44 languages and 53 language pairs. The nodes represent monolingual lexicons and edges the translation sets among them.…”