Supplementary Training on Intermediate Labeled-data Tasks (STILT) is a widely applied technique, which first fine-tunes the pretrained language models on an intermediate task before on the target task of interest. While STILT is able to further impro
ve the performance of pretrained language models, it is still unclear why and when it works. Previous research shows that those intermediate tasks involving complex inference, such as commonsense reasoning, work especially well for RoBERTa-large. In this paper, we discover that the improvement from an intermediate task could be orthogonal to it containing reasoning or other complex skills --- a simple real-fake discrimination task synthesized by GPT2 can benefit diverse target tasks. We conduct extensive experiments to study the impact of different factors on STILT. These findings suggest rethinking the role of intermediate fine-tuning in the STILT pipeline.
Although pre-training models have achieved great success in dialogue generation, their performance drops dramatically when the input contains an entity that does not appear in pre-training and fine-tuning datasets (unseen entity). To address this iss
ue, existing methods leverage an external knowledge base to generate appropriate responses. In real-world practical, the entity may not be included by the knowledge base or suffer from the precision of knowledge retrieval. To deal with this problem, instead of introducing knowledge base as the input, we force the model to learn a better semantic representation by predicting the information in the knowledge base, only based on the input context. Specifically, with the help of a knowledge base, we introduce two auxiliary training objectives: 1) Interpret Masked Word, which conjectures the meaning of the masked entity given the context; 2) Hypernym Generation, which predicts the hypernym of the entity based on the context. Experiment results on two dialogue corpus verify the effectiveness of our methods under both knowledge available and unavailable settings.
This paper presents the DuluthNLP submission to Task 7 of the SemEval 2021 competition on Detecting and Rating Humor and Offense. In it, we explain the approach used to train the model together with the process of fine-tuning our model in getting the
results. We focus on humor detection, rating, and of-fense rating, representing three out of the four subtasks that were provided. We show that optimizing hyper-parameters for learning rate, batch size and number of epochs can increase the accuracy and F1 score for humor detection
This paper reports the Machine Translation (MT) systems submitted by the IIITT team for the English→Marathi and English⇔Irish language pairs LoResMT 2021 shared task. The task focuses on getting exceptional translations for rather low-resourced langu
ages like Irish and Marathi. We fine-tune IndicTrans, a pretrained multilingual NMT model for English→Marathi, using external parallel corpus as input for additional training. We have used a pretrained Helsinki-NLP Opus MT English⇔Irish model for the latter language pair. Our approaches yield relatively promising results on the BLEU metrics. Under the team name IIITT, our systems ranked 1, 1, and 2 in English→Marathi, Irish→English, and English→Irish respectively. The codes for our systems are published1 .
In this paper we describe our submission to the multilingual Indic language translation wtask MultiIndicMT'' under the team name NICT-5''. This task involves translation from 10 Indic languages into English and vice-versa. The objective of the task w
as to explore the utility of multilingual approaches using a variety of in-domain and out-of-domain parallel and monolingual corpora. Given the recent success of multilingual NMT pre-training we decided to explore pre-training an MBART model on a large monolingual corpus collection covering all languages in this task followed by multilingual fine-tuning on small in-domain corpora. Firstly, we observed that a small amount of pre-training followed by fine-tuning on small bilingual corpora can yield large gains over when pre-training is not used. Furthermore, multilingual fine-tuning leads to further gains in translation quality which significantly outperforms a very strong multilingual baseline that does not rely on any pre-training.
Fine-tuning pre-trained language models suchas BERT has become a common practice dom-inating leaderboards across various NLP tasks.Despite its recent success and wide adoption,this process is unstable when there are onlya small number of training sam
ples available.The brittleness of this process is often reflectedby the sensitivity to random seeds. In this pa-per, we propose to tackle this problem basedon the noise stability property of deep nets,which is investigated in recent literature (Aroraet al., 2018; Sanyal et al., 2020). Specifically,we introduce a novel and effective regulariza-tion method to improve fine-tuning on NLPtasks, referred to asLayer-wiseNoiseStabilityRegularization (LNSR). We extend the theo-ries about adding noise to the input and provethat our method gives a stabler regularizationeffect. We provide supportive evidence by ex-perimentally confirming that well-performingmodels show a low sensitivity to noise andfine-tuning with LNSR exhibits clearly bet-ter generalizability and stability. Furthermore,our method also demonstrates advantages overother state-of-the-art algorithms including L2-SP (Li et al., 2018), Mixout (Lee et al., 2020)and SMART (Jiang et al., 20)
Models pretrained with self-supervised objectives on large text corpora achieve state-of-the-art performance on English text summarization tasks. However, these models are typically fine-tuned on hundreds of thousands of data points, an infeasible re
quirement when applying summarization to new, niche domains. In this work, we introduce a novel and generalizable method, called WikiTransfer, for fine-tuning pretrained models for summarization in an unsupervised, dataset-specific manner. WikiTransfer fine-tunes pretrained models on pseudo-summaries, produced from generic Wikipedia data, which contain characteristics of the target dataset, such as the length and level of abstraction of the desired summaries. WikiTransfer models achieve state-of-the-art, zero-shot abstractive summarization performance on the CNN-DailyMail dataset and demonstrate the effectiveness of our approach on three additional diverse datasets. These models are more robust to noisy data and also achieve better or comparable few-shot performance using 10 and 100 training examples when compared to few-shot transfer from other summarization datasets. To further boost performance, we employ data augmentation via round-trip translation as well as introduce a regularization term for improved few-shot transfer. To understand the role of dataset aspects in transfer performance and the quality of the resulting output summaries, we further study the effect of the components of our unsupervised fine-tuning data and analyze few-shot performance using both automatic and human evaluation.
In this paper we compare the performance of three models: SGNS (skip-gram negative sampling) and augmented versions of SVD (singular value decomposition) and PPMI (Positive Pointwise Mutual Information) on a word similarity task. We particularly focu
s on the role of hyperparameter tuning for Hindi based on recommendations made in previous work (on English). Our results show that there are language specific preferences for these hyperparameters. We extend the best settings for Hindi to a set of related languages: Punjabi, Gujarati and Marathi with favourable results. We also find that a suitably tuned SVD model outperforms SGNS for most of our languages and is also more robust in a low-resource setting.
This paper mainly introduces the relevant content of the task Hope Speech Detection for Equality, Diversity, and Inclusion at LT-EDI 2021-EACL 2021''. A total of three language datasets were provided, and we chose the English dataset to complete this
task. The specific task objective is to classify the given speech into Hope speech', Not Hope speech', and Not in intended language'. In terms of method, we use fine-tuned ALBERT and K fold cross-validation to accomplish this task. In the end, we achieved a good result in the rank list of the task result, and the final F1 score was 0.93, tying for first place. However, we will continue to try to improve methods to get better results in future work.
