Language Adaptation for Entity Relation Classification via Adversarial Neural Networks

Abstract

Entity relation classification aims to classify the semantic relationship between two marked entities in a given sentence, and plays a vital role in various natural language processing applications. However, existing studies focus on exploiting mono-lingual data in English, due to the lack of labeled data in other languages. How to effectively benefit from a richly-labeled language to help a poorly-labeled language is still an open problem. In this paper, we come up with a language adaptation framework for cross-lingual entity relation classification. The basic idea is to employ adversarial neural networks (AdvNN) to transfer feature representations from one language to another. Especially, such a language adaptation framework enables feature imitation via the competition between a sentence encoder and a rival language discriminator to generate effective representations. To verify the effectiveness of AdvNN, we introduce two kinds of adversarial structures, dual-channel AdvNN and single-channel AdvNN. Experimental results on the ACE 2005 multilingual training corpus show that our single-channel AdvNN achieves the best performance on both unsupervised and semi-supervised scenarios, yielding an improvement of 6.61% and 2.98% over the state-of-the-art, respectively. Compared with baselines which directly adopt a machine translation module, we find that both dual-channel and single-channel AdvNN significantly improve the performances (F 1) of cross-lingual entity relation classification. Moreover, extensive analysis and discussion demonstrate the appropriateness and effectiveness of different parameter settings in our language adaptation framework.

The Use of Wireless Network Combined with Artificial Intelligence Technology in the Reform of Music Online Teaching System

“…Hence, its processing basis is the need for massive datasets for training. Nonlinear, distributed, parallel computing, adaptive, and self-organizing are the main features of NN processing programs [15].…”

Remote Music Teaching Classroom Based on Machine Learning and 5G Network Station

“…There are numerous metaphors used to describe teacher education, including a large grassland with numerous paths to travel through it [26]. A virtual learning environment has been used in a number of studies to examine learning communities and the knowledge connotations of learning models and educational apps in great detail [27]. A methodology for creating virtual learning communities that emphasizes the growth of members' knowledge construction was laid out in the literature as a starting point [28].…”

Macedon: Minimizing Representation Coding Rate Reduction for Cross-Lingual Natural Language Understanding

“…Some works attempt to align the contextual word embedding of different languages, such as learning projection transformations (Aldarmaki and Diab, 2019; and forcing model into having similar predictions for parallel sentences (Yang et al, 2021;Gritta and Iacobacci, 2021;Pan et al, 2020;Dou and Neubig, 2021). And there are also a lot of works exploring adversarial training to align different languages (Chen et al, 2021(Chen et al, , 2018bKeung et al, 2019;Lee and Lee, 2019;Wang et al, 2021a;Zou et al, 2018). Their motivation is to learn language-agnostic representation to make model focus more on understand text semantics and generalize better on lowresource languages.…”

“…To enable the learned representation to be transferable among different languages, we hope the model can focus on understanding the language-agnostic semantics of text following (Chen et al, 2021(Chen et al, , 2018bKeung et al, 2019;Wang et al, 2021a;Zou et al, 2018). In the other word, the sentence representation should preserve semantic information while removing language-related information.…”

“…For languages with limited resources, however, parallel corpora are extremely difficult to obtain. Therefore, another line of approaches leverages adversarial training and employs Generative Adversarial Networks (GAN) (Goodfellow et al, 2014) to learn language-independent representations (Chen et al, 2021(Chen et al, , 2018bKeung et al, 2019;Lee and Lee, 2019;Wang et al, 2021a;Zou et al, 2018). It is worthwhile to note that adversarial training-based methods rely on an accurate estimation of data distributions, which may be infeasible with limited labeled examples in low-resource languages.…”