Existing transfer learning methods for neural machine translation typically use a well-trained translation model (i.e., a parent model) of a high-resource language pair to directly initialize a translation model (i.e., a child model) of a low-resource language pair, and the child model is then fine-tuned with corresponding datasets. In this paper, we propose a novel two-step fine-tuning (TSFT) framework for transfer learning in low-resource neural machine translation. In the first step, we adjust the parameters of the parent model to fit the child language by using the child source data. In the second step, we transfer the adjusted parameters to the child model and fine-tune it with a proposed distillation loss for efficient optimization. Our experimental results on five low-resource translations demonstrate that our framework yields significant improvements over various strong transfer learning baselines. Further analysis demonstrated the effectiveness of different components in our framework.

Open Domain Multi-Hop Question Answering (ODMHQA) plays a crucial role in Natural Language Processing (NLP) by aiming to answer complex questions through multi-step reasoning over retrieved information from external knowledge sources. Recently, Large Language Models (LLMs) have demonstrated remarkable performance in solving ODMHQA owing to their capabilities including planning, reasoning, and utilizing tools. However, LLMs may generate off-topic answers when attempting to solve ODMHQA, namely the generated answers are irrelevant to the original questions. This issue of off-topic answers accounts for approximately one-third of incorrect answers, yet remains underexplored despite its significance. To alleviate this issue, we propose the Discriminate→Re-Compose→Re- Solve→Re-Decompose (Dr3) mechanism. Specifically, the Discriminator leverages the intrinsic capabilities of LLMs to judge whether the generated answers are off-topic. In cases where an off-topic answer is detected, the Corrector performs step-wise revisions along the reversed reasoning chain (Re-Compose→Re-Solve→Re-Decompose) until the final answer becomes on-topic. Experimental results on the HotpotQA and 2WikiMultiHopQA datasets demonstrate that our Dr3 mechanism considerably reduces the occurrence of off-topic answers in ODMHQA by nearly 13%, improving the performance in Exact Match (EM) by nearly 3% compared to the baseline method without the Dr3 mechanism.

Out-of-distribution (OOD) detection, a fundamental task vexing real-world applications, has attracted growing attention in the NLP community. Recently fine-tuning based methods have made promising progress. However, it could be costly to store fine-tuned models for each scenario. In this paper, we depart from the classic fine-tuning based OOD detection toward a parameter-efficient alternative, and propose an unsupervised prefix-tuning based OOD detection framework termed PTO. Additionally, to take advantage of optional training data labels and targeted OOD data, two practical extensions of PTO are further proposed. Overall, PTO and its extensions offer several key advantages of being lightweight, easy-to-reproduce, and theoretically justified. Experimental results show that our methods perform comparably to, even better than, existing fine-tuning based OOD detection approaches under a wide range of metrics, detection settings, and OOD types.

Real-world text applications often involve composing a wide range of text control operations, such as editing the text w.r.t. an attribute, manipulating keywords and structure, and generating new text of desired properties. Prior work typically learns/finetunes a language model (LM) to perform individual or specific subsets of operations. Recent research has studied combining operations in a plug-and-play manner, often with costly search or optimization in the complex sequence space. This paper proposes a new efficient approach for composable text operations in the compact latent space of text. The low-dimensionality and differentiability of the text latent vector allow us to develop an efficient sampler based on ordinary differential equations (ODEs) given arbitrary plug-in operators (e.g., attribute classifiers). By connecting pretrained LMs (e.g., GPT2) to the latent space through efficient adaption, we then decode the sampled vectors into desired text sequences. The flexible approach permits diverse control operators (sentiment, tense, formality, keywords, etc.) acquired using any relevant data from different domains. Experiments show that composing those operators within our approach manages to generate or edit high-quality text, substantially improving over previous methods in terms of generation quality and efficiency.

Data augmentation is an effective way to enhance the performance of neural machine translation models, especially for low-resource languages. Existing data augmentation methods are either at a token level or a sentence level. The data augmented using token level methods lack syntactic diversity and may alter original meanings. Sentence level methods usually generate low-quality source sentences that are not semantically paired with the original target sentences. In this paper, we propose a novel data augmentation method to generate diverse, high-quality and meaning-preserved new instances. Our method leverages high-quality translation models trained with high-resource languages to rephrase an original sentence by translating it into an intermediate language and then back to the original language. Through this process, the high-performing translation models guarantee the quality of the rephrased sentences, and the syntactic knowledge from the intermediate language can bring syntactic diversity to the rephrased sentences. Experimental results show our method can enhance the performance in various low-resource machine translation tasks. Moreover, by combining our method with other techniques that facilitate NMT, we can yield even better results.

We participate in the WMT 2020 shared newstranslation task on Chinese→English. Our system is based on the Transformer (Vaswaniet al., 2017a) with effective variants and the DTMT (Meng and Zhang, 2019) architecture. In our experiments, we employ data selection, several synthetic data generation approaches (i.e., back-translation, knowledge distillation, and iterative in-domain knowledge transfer), advanced finetuning approaches and self-bleu based model ensemble. Our constrained Chinese→English system achieves 36.9 case-sensitive BLEU score, which is thehighest among all submissions.