Multilingual neural machine translation models support fine-tuning hundreds of languages simultaneously. However, fine-tuning on full parameters solely is inefficient potentially leading to negative interactions among languages. In this work, we demonstrate that the fine-tuning for a language occurs in its intrinsic language-specific subspace with a tiny fraction of entire parameters. Thus, we propose language-specific LoRA to isolate intrinsic language-specific subspaces. Furthermore, we propose architecture learning techniques and introduce a gradual pruning schedule during fine-tuning to exhaustively explore the optimal setting and the minimal intrinsic subspaces for each language, resulting in a lightweight yet effective fine-tuning procedure. The experimental results on a 12-language subset and a 30-language subset of FLORES-101 show that our methods not only outperform full-parameter fine-tuning up to 2.25 spBLEU scores but also reduce trainable parameters to 0.4% for high and medium-resource languages and 1.6% for low-resource ones.

Phrase-level dense retrieval has shown many appealing characteristics in downstream NLP tasks by leveraging the fine-grained information that phrases offer. In our work, we propose a new task formulation of dense retrieval, cross-lingual contextualized phrase retrieval, which aims to augment cross-lingual applications by addressing polysemy using context information. However, the lack of specific training data and models are the primary challenges to achieve our goal. As a result, we extract pairs of cross-lingual phrases using word alignment information automatically induced from parallel sentences. Subsequently, we train our Cross-lingual Contextualized Phrase Retriever (CCPR) using contrastive learning, which encourages the hidden representations of phrases with similar contexts and semantics to align closely. Comprehensive experiments on both the cross-lingual phrase retrieval task and a downstream task, i.e, machine translation, demonstrate the effectiveness of CCPR. On the phrase retrieval task, CCPR surpasses baselines by a significant margin, achieving a top-1 accuracy that is at least 13 points higher. When utilizing CCPR to augment the large-language-model-based translator, it achieves average gains of 0.7 and 1.5 in BERTScore for translations from X=>En and vice versa, respectively, on WMT16 dataset. We will release our code and data.

Multilingual neural machine translation aims to encapsulate multiple languages into a single model. However, it requires an enormous dataset, leaving the low-resource language (LRL) underdeveloped. As LRLs may benefit from shared knowledge of multilingual representation, we aspire to find effective ways to integrate unseen languages in a pre-trained model. Nevertheless, the intricacy of shared representation among languages hinders its full utilisation. To resolve this problem, we employed target language prediction and a central language-aware layer to improve representation in integrating LRLs. Focusing on improving LRLs in the linguistically diverse country of Indonesia, we evaluated five languages using a parallel corpus of 1,000 instances each, with experimental results measured by BLEU showing zero-shot improvement of 7.4 from the baseline score of 7.1 to a score of 15.5 at best. Further analysis showed that the gains in performance are attributed more to the disentanglement of multilingual representation in the encoder with the shift of the target language-specific representation in the decoder.

A Knowledge Graph (KG) is the directed graphical representation of entities and relations in the real world. KG can be applied in diverse Natural Language Processing (NLP) tasks where knowledge is required. The need to scale up and complete KG automatically yields Knowledge Graph Embedding (KGE), a shallow machine learning model that is suffering from memory and training time consumption issues. To mitigate the computational load, we propose a parameter-sharing method, i.e., using conjugate parameters for complex numbers employed in KGE models. Our method improves memory efficiency by 2x in relation embedding while achieving comparable performance to the state-of-the-art non-conjugate models, with faster, or at least comparable, training time. We demonstrated the generalizability of our method on two best-performing KGE models 5^★E (CITATION) and ComplEx (CITATION) on five benchmark datasets.

Multilingual neural machine translation can translate unseen language pairs during training, i.e. zero-shot translation. However, the zero-shot translation is always unstable. Although prior works attributed the instability to the domination of central language, e.g. English, we supplement this viewpoint with the strict dependence of non-centered languages. In this work, we propose a simple, lightweight yet effective language-specific modeling method by adapting to non-centered languages and combining the shared information and the language-specific information to counteract the instability of zero-shot translation. Experiments with Transformer on IWSLT17, Europarl, TED talks, and OPUS-100 datasets show that our method not only performs better than strong baselines in centered data conditions but also can easily fit non-centered data conditions. By further investigating the layer attribution, we show that our proposed method can disentangle the coupled representation in the correct direction.