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.
Pre-trained language models (PLMs) like BERT have made great progress in NLP. News articles usually contain rich textual information, and PLMs have the potentials to enhance news text modeling for various intelligent news applications like news recom
mendation and retrieval. However, most existing PLMs are in huge size with hundreds of millions of parameters. Many online news applications need to serve millions of users with low latency tolerance, which poses great challenges to incorporating PLMs in these scenarios. Knowledge distillation techniques can compress a large PLM into a much smaller one and meanwhile keeps good performance. However, existing language models are pre-trained and distilled on general corpus like Wikipedia, which has gaps with the news domain and may be suboptimal for news intelligence. In this paper, we propose NewsBERT, which can distill PLMs for efficient and effective news intelligence. In our approach, we design a teacher-student joint learning and distillation framework to collaboratively learn both teacher and student models, where the student model can learn from the learning experience of the teacher model. In addition, we propose a momentum distillation method by incorporating the gradients of teacher model into the update of student model to better transfer the knowledge learned by the teacher model. Thorough experiments on two real-world datasets with three tasks show that NewsBERT can empower various intelligent news applications with much smaller models.
Pretrained language models (PTLMs) yield state-of-the-art performance on many natural language processing tasks, including syntax, semantics and commonsense. In this paper, we focus on identifying to what extent do PTLMs capture semantic attributes a
nd their values, e.g., the correlation between rich and high net worth. We use PTLMs to predict masked tokens using patterns and lists of items from Wikidata in order to verify how likely PTLMs encode semantic attributes along with their values. Such inferences based on semantics are intuitive for humans as part of our language understanding. Since PTLMs are trained on large amount of Wikipedia data we would assume that they can generate similar predictions, yet our findings reveal that PTLMs are still much worse than humans on this task. We show evidence and analysis explaining how to exploit our methodology to integrate better context and semantics into PTLMs using knowledge bases.
Short text nowadays has become a more fashionable form of text data, e.g., Twitter posts, news titles, and product reviews. Extracting semantic topics from short texts plays a significant role in a wide spectrum of NLP applications, and neural topic
modeling is now a major tool to achieve it. Motivated by learning more coherent and semantic topics, in this paper we develop a novel neural topic model named Dual Word Graph Topic Model (DWGTM), which extracts topics from simultaneous word co-occurrence and semantic correlation graphs. To be specific, we learn word features from the global word co-occurrence graph, so as to ingest rich word co-occurrence information; we then generate text features with word features, and feed them into an encoder network to get topic proportions per-text; finally, we reconstruct texts and word co-occurrence graph with topical distributions and word features, respectively. Besides, to capture semantics of words, we also apply word features to reconstruct a word semantic correlation graph computed by pre-trained word embeddings. Upon those ideas, we formulate DWGTM in an auto-encoding paradigm and efficiently train it with the spirit of neural variational inference. Empirical results validate that DWGTM can generate more semantically coherent topics than baseline topic models.
Human evaluation has always been expensive while researchers struggle to trust the automatic metrics. To address this, we propose to customise traditional metrics by taking advantages of the pre-trained language models (PLMs) and the limited availabl
e human labelled scores. We first re-introduce the hLEPOR metric factors, followed by the Python version we developed (ported) which achieved the automatic tuning of the weighting parameters in hLEPOR metric. Then we present the customised hLEPOR (cushLEPOR) which uses Optuna hyper-parameter optimisation framework to fine-tune hLEPOR weighting parameters towards better agreement to pre-trained language models (using LaBSE) regarding the exact MT language pairs that cushLEPOR is deployed to. We also optimise cushLEPOR towards professional human evaluation data based on MQM and pSQM framework on English-German and Chinese-English language pairs. The experimental investigations show cushLEPOR boosts hLEPOR performances towards better agreements to PLMs like LABSE with much lower cost, and better agreements to human evaluations including MQM and pSQM scores, and yields much better performances than BLEU. Official results show that our submissions win three language pairs including English-German and Chinese-English on News domain via cushLEPOR(LM) and English-Russian on TED domain via hLEPOR. (data available at https://github.com/poethan/cushLEPOR)
Can pre-trained BERT for one language and GPT for another be glued together to translate texts? Self-supervised training using only monolingual data has led to the success of pre-trained (masked) language models in many NLP tasks. However, directly c
onnecting BERT as an encoder and GPT as a decoder can be challenging in machine translation, for GPT-like models lack a cross-attention component that is needed in seq2seq decoders. In this paper, we propose Graformer to graft separately pre-trained (masked) language models for machine translation. With monolingual data for pre-training and parallel data for grafting training, we maximally take advantage of the usage of both types of data. Experiments on 60 directions show that our method achieves average improvements of 5.8 BLEU in x2en and 2.9 BLEU in en2x directions comparing with the multilingual Transformer of the same size.
In this paper, we describe our submission to the WMT 2021 Metrics Shared Task. We use the automatically-generated questions and answers to evaluate the quality of Machine Translation (MT) systems. Our submission builds upon the recently proposed MTEQ
A framework. Experiments on WMT20 evaluation datasets show that at the system-level the MTEQA metric achieves performance comparable with other state-of-the-art solutions, while considering only a certain amount of information from the whole translation.
In this work, we present our approaches on the toxic comment classification task (subtask 1) of the GermEval 2021 Shared Task. For this binary task, we propose three models: a German BERT transformer model; a multilayer perceptron, which was first tr
ained in parallel on textual input and 14 additional linguistic features and then concatenated in an additional layer; and a multilayer perceptron with both feature types as input. We enhanced our pre-trained transformer model by re-training it with over 1 million tweets and fine-tuned it on two additional German datasets of similar tasks. The embeddings of the final fine-tuned German BERT were taken as the textual input features for our neural networks. Our best models on the validation data were both neural networks, however our enhanced German BERT gained with a F1-score = 0.5895 a higher prediction on the test data.
We describe our participation in all the subtasks of the Germeval 2021 shared task on the identification of Toxic, Engaging, and Fact-Claiming Comments. Our system is an ensemble of state-of-the-art pre-trained models finetuned with carefully enginee
red features. We show that feature engineering and data augmentation can be helpful when the training data is sparse. We achieve an F1 score of 66.87, 68.93, and 73.91 in Toxic, Engaging, and Fact-Claiming comment identification subtasks.
Toxicity is pervasive in social media and poses a major threat to the health of online communities. The recent introduction of pre-trained language models, which have achieved state-of-the-art results in many NLP tasks, has transformed the way in whi
ch we approach natural language processing. However, the inherent nature of pre-training means that they are unlikely to capture task-specific statistical information or learn domain-specific knowledge. Additionally, most implementations of these models typically do not employ conditional random fields, a method for simultaneous token classification. We show that these modifications can improve model performance on the Toxic Spans Detection task at SemEval-2021 to achieve a score within 4 percentage points of the top performing team.
