Reinforcement learning (RL) can align language models with non-differentiable reward signals, such as human preferences. However, a major challenge arises from the sparsity of these reward signals - typically, there is only a single reward for an entire output. This sparsity of rewards can lead to inefficient and unstable learning. To address this challenge, our paper introduces an novel framework that utilizes the critique capability of Large Language Models (LLMs) to produce intermediate-step rewards during RL training. Our method involves coupling a policy model with a critic language model, which is responsible for providing comprehensive feedback of each part of the output. This feedback is then translated into token or span-level rewards that can be used to guide the RL training process. We investigate this approach under two different settings: one where the policy model is smaller and is paired with a more powerful critic model, and another where a single language model fulfills both roles. We assess our approach on three text generation tasks: sentiment control, language model detoxification, and summarization. Experimental results show that incorporating artificial intrinsic rewards significantly improve both sample efficiency and the overall performance of the policy model, supported by both automatic and human evaluation.
Evaluating natural language generation (NLG) systems automatically poses significant challenges.Recent studies have employed large language models (LLMs) as reference-free metrics for NLG evaluation, enhancing adaptability to new tasks tasks. However, these methods still show lower correspondence with human judgments compared to specialized neural evaluators.In this paper, we introduce “Fusion-Eval”, an innovative approach that leverages LLMs to integrate insights from various assistant evaluators. The LLM is given the example to evaluate along with scores from the assistant evaluators. Each of these evaluators specializes in assessing distinct aspects of responses.Fusion-Eval achieves a 0.962 system-level Kendall-Tau correlation with humans on SummEval and a 0.744 turn-level Spearman correlation on TopicalChat, which is significantly higher than baseline methods. These results highlight Fusion-Eval’s significant potential in the realm of natural language system evaluation.
Large Language Models (LLMs) have demonstrated impressive capabilities for text rewriting. However creating a smaller yet potent language model for text rewriting presents two formidable challenges: costly data collection and absence of emergent capabilities.In this paper we present solutions to address the above challenges.We propose an new instruction tuning method to develop a mo-bile text rewriting model that leverages LLM-generated data and heuristic reinforcement learning, eliminating the need for human data collection. Moreover, to bridge the performance gap from the constraint size, we pro-pose a cascading approach based on the confidence levels which are distilled from the large server model’s critiques. To evaluate the text rewriting tasks for mobile scenarios, we introduce MessageRewriteEval, a human-labeled benchmark that focuses on text rewriting of messages through natural language instructions. Through empirical experiments, we demonstrate that our on-device model surpasses the current state-of-the-art LLMs in text rewriting while maintaining a significantly reduced model size using public benchmark EditEval and our new benchmark. We also demonstrate that our proposed cascading approach improves model performance further.
Online conversation is a ubiquitous way to share information and connect everyone but repetitive idiomatic text typing takes users a lot of time. This paper demonstrates a simple yet effective cloud based smart compose system to improve human-to-human conversation efficiency. Heuristics from different perspectives are designed to achieve the best trade-off between quality and latency. From the modeling side, the decoder-only model exploited the previous turns of conversational history in a computation lightweight manner. Besides, a novel phrase tokenizer is proposed to reduce latency without losing the composing quality further. Additionally, the caching mechanism is applied to the serving framework. The demo video of the system is available at https://youtu.be/U1KXkaqr60g.We open-sourced our phrase tokenizer in https://github.com/tensorflow/text.
Pre-trained language models have been recently shown to benefit task-oriented dialogue (TOD) systems. Despite their success, existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub-tasks and greater data annotation overhead. In this study, we present PPTOD, a unified plug-and-play model for task-oriented dialogue. In addition, we introduce a new dialogue multi-task pre-training strategy that allows the model to learn the primary TOD task completion skills from heterogeneous dialog corpora. We extensively test our model on three benchmark TOD tasks, including end-to-end dialogue modelling, dialogue state tracking, and intent classification. Experimental results show that PPTOD achieves new state of the art on all evaluated tasks in both high-resource and low-resource scenarios. Furthermore, comparisons against previous SOTA methods show that the responses generated by PPTOD are more factually correct and semantically coherent as judged by human annotators.
Domain-adaptive pre-training (or DA-training for short), also known as post-training, aimsto train a pre-trained general-purpose language model (LM) using an unlabeled corpus of aparticular domain to adapt the LM so that end-tasks in the domain can give improved performances. However, existing DA-training methods are in some sense blind as they do not explicitly identify what knowledge in the LM should be preserved and what should be changed by the domain corpus. This paper shows that the existing methods are suboptimal and proposes a novel method to perform a more informed adaptation of the knowledge in the LM by (1) soft-masking the attention heads based on their importance to best preserve the general knowledge in the LM and (2) contrasting the representations of the general and the full (both general and domain knowledge) to learn an integrated representation with both general and domain-specific knowledge. Experimental results will demonstrate the effectiveness of the proposed approach.
Recent work on applying large language models (LMs) achieves impressive performance in many NLP applications. Adapting or posttraining an LM using an unlabeled domain corpus can produce even better performance for end-tasks in the domain. This paper proposes the problem of continually extending an LM by incrementally post-train the LM with a sequence of unlabeled domain corpora to expand its knowledge without forgetting its previous skills. The goal is to improve the few-shot end-task learning in these domains. The resulting system is called CPT (Continual PostTraining), which to our knowledge, is the first continual post-training system. Experimental results verify its effectiveness.
Masked language models (MLMs) such as BERT have revolutionized the field of Natural Language Understanding in the past few years. However, existing pre-trained MLMs often output an anisotropic distribution of token representations that occupies a narrow subset of the entire representation space. Such token representations are not ideal, especially for tasks that demand discriminative semantic meanings of distinct tokens. In this work, we propose TaCL (Token-aware Contrastive Learning), a novel continual pre-training approach that encourages BERT to learn an isotropic and discriminative distribution of token representations. TaCL is fully unsupervised and requires no additional data. We extensively test our approach on a wide range of English and Chinese benchmarks. The results show that TaCL brings consistent and notable improvements over the original BERT model. Furthermore, we conduct detailed analysis to reveal the merits and inner-workings of our approach.
In this work, we address the open-world classification problem with a method called ODIST, open world classification via distributionally shifted instances. This novel and straightforward method can create out-of-domain instances from the in-domain training instances with the help of a pre-trained generative language model. Experimental results show that ODIST performs better than state-of-the-art decision boundary finding method.
This paper studies continual learning (CL) of a sequence of aspect sentiment classification (ASC) tasks in a particular CL setting called domain incremental learning (DIL). Each task is from a different domain or product. The DIL setting is particularly suited to ASC because in testing the system needs not know the task/domain to which the test data belongs. To our knowledge, this setting has not been studied before for ASC. This paper proposes a novel model called CLASSIC. The key novelty is a contrastive continual learning method that enables both knowledge transfer across tasks and knowledge distillation from old tasks to the new task, which eliminates the need for task ids in testing. Experimental results show the high effectiveness of CLASSIC.
This paper analyzes the pre-trained hidden representations learned from reviews on BERT for tasks in aspect-based sentiment analysis (ABSA). Our work is motivated by the recent progress in BERT-based language models for ABSA. However, it is not clear how the general proxy task of (masked) language model trained on unlabeled corpus without annotations of aspects or opinions can provide important features for downstream tasks in ABSA. By leveraging the annotated datasets in ABSA, we investigate both the attentions and the learned representations of BERT pre-trained on reviews. We found that BERT uses very few self-attention heads to encode context words (such as prepositions or pronouns that indicating an aspect) and opinion words for an aspect. Most features in the representation of an aspect are dedicated to the fine-grained semantics of the domain (or product category) and the aspect itself, instead of carrying summarized opinions from its context. We hope this investigation can help future research in improving self-supervised learning, unsupervised learning and fine-tuning for ABSA. The pre-trained model and code can be found at https://github.com/howardhsu/BERT-for-RRC-ABSA.
This paper focuses on learning domain-oriented language models driven by end tasks, which aims to combine the worlds of both general-purpose language models (such as ELMo and BERT) and domain-specific language understanding. We propose DomBERT, an extension of BERT to learn from both in-domain corpus and relevant domain corpora. This helps in learning domain language models with low-resources. Experiments are conducted on an assortment of tasks in aspect-based sentiment analysis (ABSA), demonstrating promising results.
This work introduces Focused-Variation Network (FVN), a novel model to control language generation. The main problems in previous controlled language generation models range from the difficulty of generating text according to the given attributes, to the lack of diversity of the generated texts. FVN addresses these issues by learning disjoint discrete latent spaces for each attribute inside codebooks, which allows for both controllability and diversity, while at the same time generating fluent text. We evaluate FVN on two text generation datasets with annotated content and style, and show state-of-the-art performance as assessed by automatic and human evaluations.
Question-answering plays an important role in e-commerce as it allows potential customers to actively seek crucial information about products or services to help their purchase decision making. Inspired by the recent success of machine reading comprehension (MRC) on formal documents, this paper explores the potential of turning customer reviews into a large source of knowledge that can be exploited to answer user questions. We call this problem Review Reading Comprehension (RRC). To the best of our knowledge, no existing work has been done on RRC. In this work, we first build an RRC dataset called ReviewRC based on a popular benchmark for aspect-based sentiment analysis. Since ReviewRC has limited training examples for RRC (and also for aspect-based sentiment analysis), we then explore a novel post-training approach on the popular language model BERT to enhance the performance of fine-tuning of BERT for RRC. To show the generality of the approach, the proposed post-training is also applied to some other review-based tasks such as aspect extraction and aspect sentiment classification in aspect-based sentiment analysis. Experimental results demonstrate that the proposed post-training is highly effective.
Dialogue management (DM) plays a key role in the quality of the interaction with the user in a task-oriented dialogue system. In most existing approaches, the agent predicts only one DM policy action per turn. This significantly limits the expressive power of the conversational agent and introduces unwanted turns of interactions that may challenge users’ patience. Longer conversations also lead to more errors and the system needs to be more robust to handle them. In this paper, we compare the performance of several models on the task of predicting multiple acts for each turn. A novel policy model is proposed based on a recurrent cell called gated Continue-Act-Slots (gCAS) that overcomes the limitations of the existing models. Experimental results show that gCAS outperforms other approaches. The datasets and code are available at https://leishu02.github.io/.
This paper proposes a novel end-to-end architecture for task-oriented dialogue systems. It is based on a simple and practical yet very effective sequence-to-sequence approach, where language understanding and state tracking tasks are modeled jointly with a structured copy-augmented sequential decoder and a multi-label decoder for each slot. The policy engine and language generation tasks are modeled jointly following that. The copy-augmented sequential decoder deals with new or unknown values in the conversation, while the multi-label decoder combined with the sequential decoder ensures the explicit assignment of values to slots. On the generation part, slot binary classifiers are used to improve performance. This architecture is scalable to real-world scenarios and is shown through an empirical evaluation to achieve state-of-the-art performance on both the Cambridge Restaurant dataset and the Stanford in-car assistant dataset.
One key task of fine-grained sentiment analysis of product reviews is to extract product aspects or features that users have expressed opinions on. This paper focuses on supervised aspect extraction using deep learning. Unlike other highly sophisticated supervised deep learning models, this paper proposes a novel and yet simple CNN model employing two types of pre-trained embeddings for aspect extraction: general-purpose embeddings and domain-specific embeddings. Without using any additional supervision, this model achieves surprisingly good results, outperforming state-of-the-art sophisticated existing methods. To our knowledge, this paper is the first to report such double embeddings based CNN model for aspect extraction and achieve very good results.
Traditional supervised learning makes the closed-world assumption that the classes appeared in the test data must have appeared in training. This also applies to text learning or text classification. As learning is used increasingly in dynamic open environments where some new/test documents may not belong to any of the training classes, identifying these novel documents during classification presents an important problem. This problem is called open-world classification or open classification. This paper proposes a novel deep learning based approach. It outperforms existing state-of-the-art techniques dramatically.
This paper makes a focused contribution to supervised aspect extraction. It shows that if the system has performed aspect extraction from many past domains and retained their results as knowledge, Conditional Random Fields (CRF) can leverage this knowledge in a lifelong learning manner to extract in a new domain markedly better than the traditional CRF without using this prior knowledge. The key innovation is that even after CRF training, the model can still improve its extraction with experiences in its applications.