Existing pre-trained language models (PLMs) have demonstrated the effectiveness of self-supervised learning for a broad range of natural language processing (NLP) tasks. However, most of them are not explicitly aware of domain-specific knowledge, whi
ch is essential for downstream tasks in many domains, such as tasks in e-commerce scenarios. In this paper, we propose K-PLUG, a knowledge-injected pre-trained language model based on the encoder-decoder transformer that can be transferred to both natural language understanding and generation tasks. Specifically, we propose five knowledge-aware self-supervised pre-training objectives to formulate the learning of domain-specific knowledge, including e-commerce domain-specific knowledge-bases, aspects of product entities, categories of product entities, and unique selling propositions of product entities. We verify our method in a diverse range of e-commerce scenarios that require domain-specific knowledge, including product knowledge base completion, abstractive product summarization, and multi-turn dialogue. K-PLUG significantly outperforms baselines across the board, which demonstrates that the proposed method effectively learns a diverse set of domain-specific knowledge for both language understanding and generation tasks. Our code is available.
Generative language models trained on large, diverse corpora can answer questions about a passage by generating the most likely continuation of the passage followed by a question/answer pair. However, accuracy rates vary depending on the type of ques
tion asked. In this paper we keep the passage fixed, and test with a wide variety of question types, exploring the strengths and weaknesses of the GPT-3 language model. We provide the passage and test questions as a challenge set for other language models.
Knowledge graphs are essential for numerous downstream natural language processing applications, but are typically incomplete with many facts missing. This results in research efforts on multi-hop reasoning task, which can be formulated as a search p
rocess and current models typically perform short distance reasoning. However, the long-distance reasoning is also vital with the ability to connect the superficially unrelated entities. To the best of our knowledge, there lacks a general framework that approaches multi-hop reasoning in mixed long-short distance reasoning scenarios. We argue that there are two key issues for a general multi-hop reasoning model: i) where to go, and ii) when to stop. Therefore, we propose a general model which resolves the issues with three modules: 1) the local-global knowledge module to estimate the possible paths, 2) the differentiated action dropout module to explore a diverse set of paths, and 3) the adaptive stopping search module to avoid over searching. The comprehensive results on three datasets demonstrate the superiority of our model with significant improvements against baselines in both short and long distance reasoning scenarios.
When a model attribution technique highlights a particular part of the input, a user might understand this highlight as making a statement about counterfactuals (Miller, 2019): if that part of the input were to change, the model's prediction might ch
ange as well. This paper investigates how well different attribution techniques align with this assumption on realistic counterfactuals in the case of reading comprehension (RC). RC is a particularly challenging test case, as token-level attributions that have been extensively studied in other NLP tasks such as sentiment analysis are less suitable to represent the reasoning that RC models perform. We construct counterfactual sets for three different RC settings, and through heuristics that can connect attribution methods' outputs to high-level model behavior, we can evaluate how useful different attribution methods and even different formats are for understanding counterfactuals. We find that pairwise attributions are better suited to RC than token-level attributions across these different RC settings, with our best performance coming from a modification that we propose to an existing pairwise attribution method.
Knowledge Distillation (KD) offers a natural way to reduce the latency and memory/energy usage of massive pretrained models that have come to dominate Natural Language Processing (NLP) in recent years. While numerous sophisticated variants of KD algo
rithms have been proposed for NLP applications, the key factors underpinning the optimal distillation performance are often confounded and remain unclear. We aim to identify how different components in the KD pipeline affect the resulting performance and how much the optimal KD pipeline varies across different datasets/tasks, such as the data augmentation policy, the loss function, and the intermediate representation for transferring the knowledge between teacher and student. To tease apart their effects, we propose Distiller, a meta KD framework that systematically combines a broad range of techniques across different stages of the KD pipeline, which enables us to quantify each component's contribution. Within Distiller, we unify commonly used objectives for distillation of intermediate representations under a universal mutual information (MI) objective and propose a class of MI-objective functions with better bias/variance trade-off for estimating the MI between the teacher and the student. On a diverse set of NLP datasets, the best Distiller configurations are identified via large-scale hyper-parameter optimization. Our experiments reveal the following: 1) the approach used to distill the intermediate representations is the most important factor in KD performance, 2) among different objectives for intermediate distillation, MI-performs the best, and 3) data augmentation provides a large boost for small training datasets or small student networks. Moreover, we find that different datasets/tasks prefer different KD algorithms, and thus propose a simple AutoDistiller algorithm that can recommend a good KD pipeline for a new dataset.
Research on open-domain dialogue systems that allow free topics is challenging in the field of natural language processing (NLP). The performance of the dialogue system has been improved recently by the method utilizing dialogue-related knowledge; ho
wever, non-English dialogue systems suffer from reproducing the performance of English dialogue systems because securing knowledge in the same language with the dialogue system is relatively difficult. Through experiments with a Korean dialogue system, this paper proves that the performance of a non-English dialogue system can be improved by utilizing English knowledge, highlighting the system uses cross-lingual knowledge. For the experiments, we 1) constructed a Korean version of the Wizard of Wikipedia dataset, 2) built Korean-English T5 (KE-T5), a language model pre-trained with Korean and English corpus, and 3) developed a knowledge-grounded Korean dialogue model based on KE-T5. We observed the performance improvement in the open-domain Korean dialogue model even only English knowledge was given. The experimental results showed that the knowledge inherent in cross-lingual language models can be helpful for generating responses in open dialogue systems.
Transformer-based models have become the de facto standard in the field of Natural Language Processing (NLP). By leveraging large unlabeled text corpora, they enable efficient transfer learning leading to state-of-the-art results on numerous NLP task
s. Nevertheless, for low resource languages and highly specialized tasks, transformer models tend to lag behind more classical approaches (e.g. SVM, LSTM) due to the lack of aforementioned corpora. In this paper we focus on the legal domain and we introduce a Romanian BERT model pre-trained on a large specialized corpus. Our model outperforms several strong baselines for legal judgement prediction on two different corpora consisting of cases from trials involving banks in Romania.
Despite the success of neural dialogue systems in achieving high performance on the leader-board, they cannot meet users' requirements in practice, due to their poor reasoning skills. The underlying reason is that most neural dialogue models only cap
ture the syntactic and semantic information, but fail to model the logical consistency between the dialogue history and the generated response. Recently, a new multi-turn dialogue reasoning task has been proposed, to facilitate dialogue reasoning research. However, this task is challenging, because there are only slight differences between the illogical response and the dialogue history. How to effectively solve this challenge is still worth exploring. This paper proposes a Fine-grained Comparison Model (FCM) to tackle this problem. Inspired by human's behavior in reading comprehension, a comparison mechanism is proposed to focus on the fine-grained differences in the representation of each response candidate. Specifically, each candidate representation is compared with the whole history to obtain a history consistency representation. Furthermore, the consistency signals between each candidate and the speaker's own history are considered to drive a model prefer a candidate that is logically consistent with the speaker's history logic. Finally, the above consistency representations are employed to output a ranking list of the candidate responses for multi-turn dialogue reasoning. Experimental results on two public dialogue datasets show that our method obtains higher ranking scores than the baseline models.
In this study, we propose a self-supervised learning method that distils representations of word meaning in context from a pre-trained masked language model. Word representations are the basis for context-aware lexical semantics and unsupervised sema
ntic textual similarity (STS) estimation. A previous study transforms contextualised representations employing static word embeddings to weaken excessive effects of contextual information. In contrast, the proposed method derives representations of word meaning in context while preserving useful context information intact. Specifically, our method learns to combine outputs of different hidden layers using self-attention through self-supervised learning with an automatically generated training corpus. To evaluate the performance of the proposed approach, we performed comparative experiments using a range of benchmark tasks. The results confirm that our representations exhibited a competitive performance compared to that of the state-of-the-art method transforming contextualised representations for the context-aware lexical semantic tasks and outperformed it for STS estimation.
A computationally expensive and memory intensive neural network lies behind the recent success of language representation learning. Knowledge distillation, a major technique for deploying such a vast language model in resource-scarce environments, tr
ansfers the knowledge on individual word representations learned without restrictions. In this paper, inspired by the recent observations that language representations are relatively positioned and have more semantic knowledge as a whole, we present a new knowledge distillation objective for language representation learning that transfers the contextual knowledge via two types of relationships across representations: Word Relation and Layer Transforming Relation. Unlike other recent distillation techniques for the language models, our contextual distillation does not have any restrictions on architectural changes between teacher and student. We validate the effectiveness of our method on challenging benchmarks of language understanding tasks, not only in architectures of various sizes but also in combination with DynaBERT, the recently proposed adaptive size pruning method.
