COSPLAY: Concept Set Guided Personalized Dialogue Generation Across Both Party Personas

Maintaining a consistent persona is essential for building a human-like conversational model. However, the lack of attention to the partner makes the model more egocentric: they tend to show their persona by all means such as twisting the topic stiffly, pulling the conversation to their own interests regardless, and rambling their persona with little curiosity to the partner. In this work, we propose COSPLAY(COncept Set guided PersonaLized dialogue generation Across both partY personas) that considers both parties as a "team": expressing self-persona while keeping curiosity toward the partner, leading responses around mutual personas, and finding the common ground. Specifically, we first represent self-persona, partner persona and mutual dialogue all in the concept sets. Then, we propose the Concept Set framework with a suite of knowledge-enhanced operations to process them such as set algebras, set expansion, and set distance. Based on these operations as medium, we train the model by utilizing 1) concepts of both party personas, 2) concept relationship between them, and 3) their relationship to the future dialogue. Extensive experiments on a large public dataset, Persona-Chat, demonstrate that our model outperforms state-of-the-art baselines for generating less egocentric, more human-like, and higher quality responses in both automatic and human evaluations.

Event Transition Planning for Open-ended Text Generation

Open-ended text generation tasks, such as dialogue generation and story completion, require models to generate a coherent continuation given limited preceding context. The open-ended nature of these tasks brings new challenges to the neural auto-regressive text generators nowadays. Despite these neural models are good at producing human-like text, it is difficult for them to arrange causalities and relations between given facts and possible ensuing events. To bridge this gap, we propose a novel two-stage method which explicitly arranges the ensuing events in open-ended text generation. Our approach can be understood as a specially-trained coarse-to-fine algorithm, where an event transition planner provides a "coarse" plot skeleton and a text generator in the second stage refines the skeleton. Experiments on two open-ended text generation tasks demonstrate that our proposed method effectively improves the quality of the generated text, especially in coherence and diversity. The code is available at: \url{https://github.com/qtli/EventPlanforTextGen}.

Parameter-Efficient Tuning by Manipulating Hidden States of Pretrained Language Models For Classification Tasks

Parameter-efficient tuning aims to distill knowledge for downstream tasks by optimizing a few introduced parameters while freezing the pretrained language models (PLMs). Continuous prompt tuning which prepends a few trainable vectors to the embeddings of input is one of these methods and has drawn much attention due to its effectiveness and efficiency. This family of methods can be illustrated as exerting nonlinear transformations of hidden states inside PLMs. However, a natural question is ignored: can the hidden states be directly used for classification without changing them? In this paper, we aim to answer this question by proposing a simple tuning method which only introduces three trainable vectors. Firstly, we integrate all layers hidden states using the introduced vectors. And then, we input the integrated hidden state(s) to a task-specific linear classifier to predict categories. This scheme is similar to the way ELMo utilises hidden states except that they feed the hidden states to LSTM-based models. Although our proposed tuning scheme is simple, it achieves comparable performance with prompt tuning methods like P-tuning and P-tuning v2, verifying that original hidden states do contain useful information for classification tasks. Moreover, our method has an advantage over prompt tuning in terms of time and the number of parameters.

"Is Whole Word Masking Always Better for Chinese BERT?": Probing on Chinese Grammatical Error Correction

Whole word masking (WWM), which masks all subwords corresponding to a word at once, makes a better English BERT model. For the Chinese language, however, there is no subword because each token is an atomic character. The meaning of a word in Chinese is different in that a word is a compositional unit consisting of multiple characters. Such difference motivates us to investigate whether WWM leads to better context understanding ability for Chinese BERT. To achieve this, we introduce two probing tasks related to grammatical error correction and ask pretrained models to revise or insert tokens in a masked language modeling manner. We construct a dataset including labels for 19,075 tokens in 10,448 sentences. We train three Chinese BERT models with standard character-level masking (CLM), WWM, and a combination of CLM and WWM, respectively. Our major findings are as follows: First, when one character needs to be inserted or replaced, the model trained with CLM performs the best. Second, when more than one character needs to be handled, WWM is the key to better performance. Finally, when being fine-tuned on sentence-level downstream tasks, models trained with different masking strategies perform comparably.

Contrastive Representation Learning for Exemplar-Guided Paraphrase Generation

Exemplar-Guided Paraphrase Generation (EGPG) aims to generate a target sentence which conforms to the style of the given exemplar while encapsulating the content information of the source sentence. In this paper, we propose a new method with the goal of learning a better representation of the style andthe content. This method is mainly motivated by the recent success of contrastive learning which has demonstrated its power in unsupervised feature extraction tasks. The idea is to design two contrastive losses with respect to the content and the style by considering two problem characteristics during training. One characteristic is that the target sentence shares the same content with the source sentence, and the second characteristic is that the target sentence shares the same style with the exemplar. These two contrastive losses are incorporated into the general encoder-decoder paradigm. Experiments on two datasets, namely QQP-Pos and ParaNMT, demonstrate the effectiveness of our proposed constrastive losses.

Sentence Semantic Regression for Text Generation

Recall the classical text generation works, the generation framework can be briefly divided into two phases: \textbf{idea reasoning} and \textbf{surface realization}. The target of idea reasoning is to figure out the main idea which will be presented in the following talking/writing periods. Surface realization aims to arrange the most appropriate sentence to depict and convey the information distilled from the main idea. However, the current popular token-by-token text generation methods ignore this crucial process and suffer from many serious issues, such as idea/topic drift. To tackle the problems and realize this two-phase paradigm, we propose a new framework named Sentence Semantic Regression (\textbf{SSR}) based on sentence-level language modeling. For idea reasoning, two architectures \textbf{SSR-AR} and \textbf{SSR-NonAR} are designed to conduct sentence semantic regression autoregressively (like GPT2/3) and bidirectionally (like BERT). In the phase of surface realization, a mixed-granularity sentence decoder is designed to generate text with better consistency by jointly incorporating the predicted sentence-level main idea as well as the preceding contextual token-level information. We conduct experiments on four tasks of story ending prediction, story ending generation, dialogue generation, and sentence infilling. The results show that SSR can obtain better performance in terms of automatic metrics and human evaluation.

Dialogue Summarization with Supporting Utterance Flow Modeling and Fact Regularization

Dialogue summarization aims to generate a summary that indicates the key points of a given dialogue. In this work, we propose an end-to-end neural model for dialogue summarization with two novel modules, namely, the \emph{supporting utterance flow modeling module} and the \emph{fact regularization module}. The supporting utterance flow modeling helps to generate a coherent summary by smoothly shifting the focus from the former utterances to the later ones. The fact regularization encourages the generated summary to be factually consistent with the ground-truth summary during model training, which helps to improve the factual correctness of the generated summary in inference time. Furthermore, we also introduce a new benchmark dataset for dialogue summarization. Extensive experiments on both existing and newly-introduced datasets demonstrate the effectiveness of our model.

CLINE: Contrastive Learning with Semantic Negative Examples for Natural Language Understanding

Despite pre-trained language models have proven useful for learning high-quality semantic representations, these models are still vulnerable to simple perturbations. Recent works aimed to improve the robustness of pre-trained models mainly focus on adversarial training from perturbed examples with similar semantics, neglecting the utilization of different or even opposite semantics. Different from the image processing field, the text is discrete and few word substitutions can cause significant semantic changes. To study the impact of semantics caused by small perturbations, we conduct a series of pilot experiments and surprisingly find that adversarial training is useless or even harmful for the model to detect these semantic changes. To address this problem, we propose Contrastive Learning with semantIc Negative Examples (CLINE), which constructs semantic negative examples unsupervised to improve the robustness under semantically adversarial attacking. By comparing with similar and opposite semantic examples, the model can effectively perceive the semantic changes caused by small perturbations. Empirical results show that our approach yields substantial improvements on a range of sentiment analysis, reasoning, and reading comprehension tasks. And CLINE also ensures the compactness within the same semantics and separability across different semantics in sentence-level.

A Training-free and Reference-free Summarization Evaluation Metric via Centrality-weighted Relevance and Self-referenced Redundancy

In recent years, reference-based and supervised summarization evaluation metrics have been widely explored. However, collecting human-annotated references and ratings are costly and time-consuming. To avoid these limitations, we propose a training-free and reference-free summarization evaluation metric. Our metric consists of a centrality-weighted relevance score and a self-referenced redundancy score. The relevance score is computed between the pseudo reference built from the source document and the given summary, where the pseudo reference content is weighted by the sentence centrality to provide importance guidance. Besides an $F_1$-based relevance score, we also design an $F_\beta$-based variant that pays more attention to the recall score. As for the redundancy score of the summary, we compute a self-masked similarity score with the summary itself to evaluate the redundant information in the summary. Finally, we combine the relevance and redundancy scores to produce the final evaluation score of the given summary. Extensive experiments show that our methods can significantly outperform existing methods on both multi-document and single-document summarization evaluation.