Commonsense question answering has primarily been tackled through supervised transfer learning, where a language model pre-trained on large amounts of data is used as the starting point. While successful, the approach requires large amounts of labeled question-answer pairs, with increasingly larger amounts of data required as the complexity of scenarios or tasks such as commonsense QA increases. In this paper, we hypothesize that large-scale pre-training of language models encodes the necessary commonsense knowledge to answer common questions in context without labeled data. We propose a novel framework called Iterative Self Distillation for QA (ISD-QA), which extracts the “dark knowledge” encoded during largescale pre-training of language models to provide supervision for commonsense question answering. We show that the approach can be used to train common neural QA models for commonsense question answering by distilling knowledge from language models in an unsupervised manner. With no bells and whistles, we achieve an average of 68% of the performance of fully supervised QA models while requiring no labeled training data. Extensive experiments on three public benchmarks (OpenBookQA, HellaSWAG, and CommonsenseQA) show the effectiveness of the proposed approach.
more »
« less
This content will become publicly available on June 20, 2024
Leveraging Symbolic Knowledge Bases for Commonsense Natural Language Inference using Pattern Theory
The commonsense natural language inference (CNLI) tasks aim to select the most likely follow-up statement to a contextual description of ordinary, everyday events and facts. Current approaches to transfer learning of CNLI models across tasks require many labeled data from the new task. This paper presents a way to reduce this need for additional annotated training data from the new task by leveraging symbolic knowledge bases, such as ConceptNet. We formulate a teacher-student framework for mixed symbolic-neural reasoning, with the large-scale symbolic knowledge base serving as the teacher and a trained CNLI model as the student. This hybrid distillation process involves two steps. The first step is a symbolic reasoning process. Given a collection of unlabeled data, we use an abductive reasoning framework based on Grenander's pattern theory to create weakly labeled data. Pattern theory is an energy-based graphical probabilistic framework for reasoning among random variables with varying dependency structures. In the second step, the weakly labeled data, along with a fraction of the labeled data, is used to transfer-learn the CNLI model into the new task. The goal is to reduce the fraction of labeled data required. We demonstrate the efficacy of our approach by using three publicly available datasets (OpenBookQA, SWAG, and HellaSWAG) and evaluating three CNLI models (BERT, LSTM, and ESIM) that represent different tasks. We show that, on average, we achieve 63% of the top performance of a fully supervised BERT model with no labeled data. With only 1000 labeled samples, we can improve this performance to 72%. Interestingly, without training, the teacher mechanism itself has significant inference power. The pattern theory framework achieves 32.7% accuracy on OpenBookQA, outperforming transformer-based models such as GPT (26.6%), GPT-2 (30.2%), and BERT (27.1%) by a significant margin. We demonstrate that the framework can be generalized to successfully train neural CNLI models using knowledge distillation under unsupervised and semi-supervised learning settings. Our results show that it outperforms all unsupervised and weakly supervised baselines and some early supervised approaches, while offering competitive performance with fully supervised baselines. Additionally, we show that the abductive learning framework can be adapted for other downstream tasks, such as unsupervised semantic textual similarity, unsupervised sentiment classification, and zero-shot text classification, without significant modification to the framework. Finally, user studies show that the generated interpretations enhance its explainability by providing key insights into its reasoning mechanism.
more »
« less
- NSF-PAR ID:
- 10433572
- Date Published:
- Journal Name:
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- ISSN:
- 0162-8828
- Page Range / eLocation ID:
- 1 to 18
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Extracting relations from text corpora is an important task with wide applications. However, it becomes particularly challenging when focusing on weakly-supervised relation extraction, that is, utilizing a few relation instances (i.e., a pair of entities and their relation) as seeds to extract from corpora more instances of the same relation. Existing distributional approaches leverage the corpuslevel co-occurrence statistics of entities to predict their relations, and require a large number of labeled instances to learn effective relation classiffers. Alternatively, pattern-based approaches perform boostrapping or apply neural networks to model the local contexts, but still rely on a large number of labeled instances to build reliable models. In this paper, we study the integration of distributional and pattern-based methods in a weakly-supervised setting such that the two kinds of methods can provide complementary supervision for each other to build an effective, uniffed model. We propose a novel co-training framework with a distributional module and a pattern module. During training, the distributional module helps the pattern module discriminate between the informative patterns and other patterns, and the pattern module generates some highly-conffdent instances to improve the distributional module. The whole framework can be effectively optimized by iterating between improving the pattern module and updating the distributional module. We conduct experiments on two tasks: knowledge base completion with text corpora and corpus-level relation extraction. Experimental results prove the effectiveness of our framework over many competitive baselines.more » « less
-
Karlapalem, Kamal ; Cheng, Hong ; Ramakrishnan, Naren ; null ; null ; Reddy, P. Krishna ; Srivastava, Jaideep ; Chakraborty, Tanmoy (Ed.)Constrained learning, a weakly supervised learning task, aims to incorporate domain constraints to learn models without requiring labels for each instance. Because weak supervision knowledge is useful and easy to obtain, constrained learning outperforms unsupervised learning in performance and is preferable than supervised learning in terms of labeling costs. To date, constrained learning, especially constrained clustering, has been extensively studied, but was primarily focused on data in the Euclidean space. In this paper, we propose a weak supervision network embedding (WSNE) for constrained learning of graphs. Because no label is available for individual nodes, we propose a new loss function to quantify the constraint-based loss, and integrate this loss in a graph convolutional neural network (GCN) and variational graph auto-encoder (VGAE) combined framework to jointly model graph structures and node attributes. The joint optimization allows WSNE to learn embedding not only preserving network topology and content, but also satisfying the constraints. Experiments show that WSNE outperforms baselines for constrained graph learning tasks, including constrained graph clustering and constrained graph classification.more » « less
-
more » « less
Abstract Background Natural language processing (NLP) tasks in the health domain often deal with limited amount of labeled data due to high annotation costs and naturally rare observations. To compensate for the lack of training data, health NLP researchers often have to leverage knowledge and resources external to a task at hand. Recently, pretrained large-scale language models such as the Bidirectional Encoder Representations from Transformers (BERT) have been proven to be a powerful way of learning rich linguistic knowledge from massive unlabeled text and transferring that knowledge to downstream tasks. However, previous downstream tasks often used training data at such a large scale that is unlikely to obtain in the health domain. In this work, we aim to study whether BERT can still benefit downstream tasks when training data are relatively small in the context of health NLP.
Method We conducted a learning curve analysis to study the behavior of BERT and baseline models as training data size increases. We observed the classification performance of these models on two disease diagnosis data sets, where some diseases are naturally rare and have very limited observations (fewer than 2 out of 10,000). The baselines included commonly used text classification models such as sparse and dense bag-of-words models, long short-term memory networks, and their variants that leveraged external knowledge. To obtain learning curves, we incremented the amount of training examples per disease from small to large, and measured the classification performance in macro-averaged
$$F_{1}$$ Results On the task of classifying all diseases, the learning curves of BERT were consistently above all baselines, significantly outperforming them across the spectrum of training data sizes. But under extreme situations where only one or two training documents per disease were available, BERT was outperformed by linear classifiers with carefully engineered bag-of-words features.
Conclusion As long as the amount of training documents is not extremely few, fine-tuning a pretrained BERT model is a highly effective approach to health NLP tasks like disease classification. However, in extreme cases where each class has only one or two training documents and no more will be available, simple linear models using bag-of-words features shall be considered.
-
We address the challenge of getting efficient yet accurate recognition systems with limited labels. While recognition models improve with model size and amount of data, many specialized applications of computer vision have severe resource constraints both during training and inference. Transfer learning is an effective solution for training with few labels, however often at the expense of a compu- tationally costly fine-tuning of large base models. We propose to mitigate this unpleasant trade-off between compute and accuracy via semi-supervised cross- domain distillation from a set of diverse source models. Initially, we show how to use task similarity metrics to select a single suitable source model to distill from, and that a good selection process is imperative for good downstream performance of a target model. We dub this approach DISTILLNEAREST. Though effective, DISTILLNEAREST assumes a single source model matches the target task, which is not always the case. To alleviate this, we propose a weighted multi-source distilla- tion method to distill multiple source models trained on different domains weighted by their relevance for the target task into a single efficient model (named DISTILL- WEIGHTED). Our methods need no access to source data, and merely need features and pseudo-labels of the source models. When the goal is accurate recognition under computational constraints, both DISTILLNEAREST and DISTILLWEIGHTED approaches outperform both transfer learning from strong ImageNet initializations as well as state-of-the-art semi-supervised techniques such as FixMatch. Averaged over 8 diverse target tasks our multi-source method outperforms the baselines by 5.6%-points and 4.5%-points, respectively.more » « less
