Previous zero-shot dialogue state tracking (DST) methods only apply transfer learning, but ignore unlabelled data in the target domain.We transform zero-shot DST into few-shot DST by utilising such unlabelled data via joint and self-training methods. Our method incorporates auxiliary tasks that generate slot types as inverse prompts for main tasks, creating slot values during joint training. Cycle consistency between these two tasks enables the generation and selection of quality samples in unknown target domains for subsequent fine-tuning. This approach also facilitates automatic label creation, thereby optimizing the training and fine-tuning of DST models. We demonstrate this method’s effectiveness on general language models in zero-shot scenarios, improving average joint goal accuracy by 8% across all domains in MultiWOZ.

The automatic evaluation of natural language generation (NLG) systems presents a long-lasting challenge. Recent studies have highlighted various neural metrics that align well with human evaluations. Yet, the robustness of these evaluators against adversarial perturbations remains largely under-explored due to the unique challenges in obtaining adversarial data for different NLG evaluation tasks. To address the problem, we introduce AdvEval, a novel black-box adversarial framework against NLG evaluators. AdvEval is specially tailored to generate data that yield strong disagreements between human and victim evaluators. Specifically, inspired by the recent success of large language models (LLMs) in text generation and evaluation, we adopt strong LLMs as both the data generator and gold evaluator. Adversarial data are automatically optimized with feedback from the gold and victim evaluator. We conduct experiments on 12 victim evaluators and 11 NLG datasets, spanning tasks including dialogue, summarization, and question evaluation. The results show that AdvEval can lead to significant performance degradation of various victim metrics, thereby validating its efficacy.

Large Language Models (LLMs) provide a possibility to make a great breakthrough in medicine. The establishment of a standardized medical benchmark becomes a fundamental cornerstone to measure progression. However, medical environments in different regions have their local characteristics, e.g., the ubiquity and significance of traditional Chinese medicine within China. Therefore, merely translating English-based medical evaluation may result in contextual incongruities to a local region. To solve the issue, we propose a localized medical benchmark called CMB, a Comprehensive Medical Benchmark in Chinese, designed and rooted entirely within the native Chinese linguistic and cultural framework. While traditional Chinese medicine is integral to this evaluation, it does not constitute its entirety. Using this benchmark, we have evaluated several prominent large-scale LLMs, including ChatGPT, GPT-4, dedicated Chinese LLMs, and LLMs specialized in the medical domain. We hope this benchmark provide first-hand experience in existing LLMs for medicine and also facilitate the widespread adoption and enhancement of medical LLMs within China. Our data and code are publicly available at https://github.com/FreedomIntelligence/CMB.

This paper is devoted to the development of a localized Large Language Model (LLM) specifically for Arabic, a language imbued with unique cultural characteristics inadequately addressed by current mainstream models. Significant concerns emerge when addressing cultural sensitivity and local values. To address this, the paper proposes a comprehensive solution that includes further pre-training with Arabic texts, Supervised Fine-Tuning (SFT) utilizing native Arabic instructions, and GPT-4 responses in Arabic, alongside Reinforcement Learning with AI Feedback (RLAIF) employing a reward model attuned to local culture and values. The goal is to cultivate culturally cognizant and value-aligned Arabic LLMs capable of accommodating the diverse, application-specific needs of Arabic-speaking communities. Comprehensive evaluations reveal that the resulting model, dubbed ‘AceGPT’, sets the state-of-the-art standard for open Arabic LLMs across various benchmarks. Codes, data, and models are in https://github.com/FreedomIntelligence/AceGPT.

Topic segmentation and outline generation strive to divide a document into coherent topic sections and generate corresponding subheadings, unveiling the discourse topic structure of a document. Compared with sentence-level topic structure, the paragraph-level topic structure can quickly grasp and understand the overall context of the document from a higher level, benefitting many downstream tasks such as summarization, discourse parsing, and information retrieval. However, the lack of large-scale, high-quality Chinese paragraph-level topic structure corpora restrained relative research and applications. To fill this gap, we build the Chinese paragraph-level topic representation, corpus, and benchmark in this paper. Firstly, we propose a hierarchical paragraph-level topic structure representation with three layers to guide the corpus construction. Then, we employ a two-stage man-machine collaborative annotation method to construct the largest Chinese Paragraph-level Topic Structure corpus (CPTS), achieving high quality. We also build several strong baselines, including ChatGPT, to validate the computability of CPTS on two fundamental tasks (topic segmentation and outline generation) and preliminarily verified its usefulness for the downstream task (discourse parsing).

Training or finetuning large-scale language models (LLMs) requires substantial computation resources, motivating recent efforts to explore parameter-efficient adaptation to downstream tasks. One approach is to treat these models as black boxes and use forward passes (Inference APIs) to interact with them. Current research focuses on adapting these black-box models to downstream tasks using gradient-free prompt optimization, but this often involves an expensive process of searching task-specific prompts. Therefore, we are motivated to study black-box language model adaptation without prompt search. Specifically, we introduce a label-enhanced cross-attention network called CrossTune, which models the semantic relatedness between the input text sequence and task-specific label descriptions. Its effectiveness is examined in the context of few-shot text classification. To improve the generalization of CrossTune, we utilize ChatGPT to generate additional training data through in-context learning. A switch mechanism is implemented to exclude low-quality ChatGPT-generated data. Through extensive experiments on seven benchmark text classification datasets, we demonstrate that our proposed approach outperforms the previous state-of-the-art gradient-free black-box tuning method by 5.7% on average. Even without using ChatGPT-augmented data, CrossTune performs better or comparably than previous black-box tuning methods, suggesting the effectiveness of our approach.

Large language models, like ChatGPT, have shown remarkable capability in many downstream tasks, yet their ability to understand discourse structures of dialogues remains less explored, where it requires higher level capabilities of understanding and reasoning. In this paper, we aim to systematically inspect ChatGPT’s performance in two discourse analysis tasks: topic segmentation and discourse parsing, focusing on its deep semantic understanding of linear and hierarchical discourse structures underlying dialogue. To instruct ChatGPT to complete these tasks, we initially craft a prompt template consisting of the task description, output format, and structured input. Then, we conduct experiments on four popular topic segmentation datasets and two discourse parsing datasets. The experimental results showcase that ChatGPT demonstrates proficiency in identifying topic structures in general-domain conversations yet struggles considerably in specific-domain conversations. We also found that ChatGPT hardly understands rhetorical structures that are more complex than topic structures. Our deeper investigation indicates that ChatGPT can give more reasonable topic structures than human annotations but only linearly parses the hierarchical rhetorical structures. In addition, we delve into the impact of in-context learning (e.g., chain-of-thought) on ChatGPT and conduct the ablation study on various prompt components, which can provide a research foundation for future work. The code is available at https://github.com/yxfanSuda/GPTforDDA.

Recent advancements in reference-free learned metrics for open-domain dialogue evaluation have been driven by the progress in pre-trained language models and the availability of dialogue data with high-quality human annotations. However, current studies predominantly concentrate on English dialogues, and the generalization of these metrics to other languages has not been fully examined. This is largely due to the absence of a multilingual dialogue evaluation benchmark. To address the issue, we introduce xDial-Eval, built on top of open-source English dialogue evaluation datasets. xDial-Eval includes 12 turn-level and 6 dialogue-level English datasets, comprising 14930 annotated turns and 8691 annotated dialogues respectively. The English dialogue data are extended to nine other languages with commercial machine translation systems. On xDial-Eval, we conduct comprehensive analyses of previous BERT-based metrics and the recently-emerged large language models. Lastly, we establish strong self-supervised and multilingual baselines. In terms of average Pearson correlations over all datasets and languages, the best baseline outperforms OpenAI’s ChatGPT by absolute improvements of 6.5% and 4.6% at the turn and dialogue levels respectively, albeit with much fewer parameters. The data and code are publicly available at https://github.com/e0397123/xDial-Eval.

Text embedding models have significantly contributed to advancements in natural language processing by adeptly capturing semantic properties of textual data. However, the ability of these models to generalize across a wide range of syntactic contexts remains under-explored. In this paper, we first develop an evaluation set, named SR, to scrutinize the capability for syntax understanding of text embedding models from two crucial syntactic aspects: Structural heuristics, and Relational understanding among concepts, as revealed by the performance gaps in previous studies. Our findings reveal that existing text embedding models have not sufficiently addressed these syntactic understanding challenges, and such ineffectiveness becomes even more apparent when evaluated against existing benchmark datasets. Furthermore, we conduct rigorous analysis to unearth factors that lead to such limitations and examine why previous evaluations fail to detect such ineffectiveness. Lastly, we propose strategies to augment the generalization ability of text embedding models in diverse syntactic scenarios. This study serves to highlight the hurdles associated with syntactic generalization and provides pragmatic guidance for boosting model performance across varied syntactic contexts.

In this paper, we present HuatuoGPT, a Large Language Model (LLM) for medical consultation. The core recipe of HuatuoGPT is to leverage both distilled data from **ChatGPT** and real-world data from **doctors** in the supervised fine-tuning stage. This is not only because purely using **ChatGPT**-distilled data might cause ‘model collapse’, but also because real-world data from **doctors** would be complementary to **ChatGPT**-distilled data. The responses from ChatGPT are usually detailed, well-presented, fluent, and instruction-followed, but it cannot perform like a doctor in many aspects, e.g. for interactive diagnosis. Therefore, the extra doctors’ data could tame a distilled language model to perform like doctors. To synergize the strengths of both data sources, we introduce RLMF (Reinforcement Learning from Mixed Feedback) where a reward model is trained to align the language model with the merits that both sources (ChatGPT and doctors) bring. Experimental results (in GPT-4 evaluation, human evaluation, and medical benchmark datasets) demonstrate that HuatuoGPT achieves state-of-the-art results in performing medical consultation among open-source LLMs. It is worth noting that by using additional real-world data and RLMF, the distilled language model (i.e., HuatuoGPT) outperforms its teacher model (i.e., ChatGPT) in most cases.

Recognizing code-switching (CS) speech often presents challenges for an automatic speech recognition system (ASR) due to limited linguistic context in short monolingual segments, resulting in language confusion. To mitigate this issue, language identity (LID) is often integrated into the speech recognition system to provide additional linguistic context. However, previous works predominately focus on extracting language identity from speech signals. We introduce a novel approach to learn language identity from pure text data via a dedicated language identity-language model. Besides, we explore two strategies: LID state fusion and language posterior biasing, to integrate the text-derived language identities into the end-to-end ASR system. By incorporating hypothesized language identities, our ASR system gains crucial contextual cues, effectively capturing language transitions and patterns within code-switched utterances. We conduct speech recognition experiments on the SEAME corpus and demonstrate the effectiveness of our proposed methods. Our results reveal significantly improved transcriptions in code-switching scenarios, underscoring the potential of text-derived LID in enhancing code-switching speech recognition.

Prior research on dialogue state tracking (DST) is mostly based on written dialogue corpora. For spoken dialogues, the DST model trained on the written text should use the results (or hypothesis) of automatic speech recognition (ASR) as input. But ASR hypothesis often includes errors, which leads to significant performance drop for spoken dialogue state tracking. We address the issue by developing the following ASR error correction modules. First, we train a model to convert ASR hypothesis to ground truth user utterance, which can fix frequent patterns of errors. The model takes ASR hypotheses of two ASR models as input and fine-tuned in two stages. The corrected hypothesis is fed into a large scale pre-trained encoder-decoder model (T5) for DST training and inference. Second, if an output slot value from the encoder-decoder model is a name, we compare it with names in a dictionary crawled from Web sites and, if feasible, replace with the crawled name of the shortest edit distance. Third, we fix errors of temporal expressions in ASR hypothesis by using hand-crafted rules. Experiment results on the DSTC 11 speech-aware dataset, which is built on the popular MultiWOZ task (version 2.1), show that our proposed method can effectively mitigate the performance drop when moving from written text to spoken conversations.

Despite much success in natural language processing (NLP), pre-trained language models typically lead to a high computational cost during inference. Multi-exit is a mainstream approach to address this issue by making a trade-off between efficiency and accuracy, where the saving of computation comes from an early exit. However, whether such saving from early-exiting is robust remains unknown. Motivated by this, we first show that directly adapting existing adversarial attack approaches targeting model accuracy cannot significantly reduce inference efficiency. To this end, we propose a simple yet effective attacking framework, SAME, a novel slowdown attack framework on multi-exit models, which is specially tailored to reduce the efficiency of the multi-exit models. By leveraging the multi-exit models’ design characteristics, we utilize all internal predictions to guide the adversarial sample generation instead of merely considering the final prediction. Experiments on the GLUE benchmark show that SAME can effectively diminish the efficiency gain of various multi-exit models by 80% on average, convincingly validating its effectiveness and generalization ability.

Recent model-based reference-free metrics for open-domain dialogue evaluation exhibit promising correlations with human judgment. However, they either perform turn-level evaluation or look at a single dialogue quality dimension. One would expect a good evaluation metric to assess multiple quality dimensions at the dialogue level. To this end, we are motivated to propose a multi-dimensional dialogue-level metric, which consists of three sub-metrics with each targeting a specific dimension. The sub-metrics are trained with novel self-supervised objectives and exhibit strong correlations with human judgment for their respective dimensions. Moreover, we explore two approaches to combine the sub-metrics: metric ensemble and multitask learning. Both approaches yield a holistic metric that significantly outperforms individual sub-metrics. Compared to the existing state-of-the-art metric, the combined metrics achieve around 16% relative improvement on average across three high-quality dialogue-level evaluation benchmarks.

Dialogue summarization is abstractive in nature, making it suffer from factual errors. The factual correctness of summaries has the highest priority before practical applications. Many efforts have been made to improve faithfulness in text summarization. However, there is a lack of systematic study on dialogue summarization systems. In this work, we first perform the fine-grained human analysis on the faithfulness of dialogue summaries and observe that over 35% of generated summaries are faithfully inconsistent respective the source dialogues. Furthermore, we present a new model-level faithfulness evaluation method. It examines generation models with multi-choice questions created by rule-based transformations. Experimental results show that our evaluation schema is a strong proxy for the factual correctness of summarization models. The human-annotated faithfulness samples and the evaluation toolkit are released to facilitate future research toward faithful dialogue summarization.

Most sentence embedding techniques heavily rely on expensive human-annotated sentence pairs as the supervised signals. Despite the use of large-scale unlabeled data, the performance of unsupervised methods typically lags far behind that of the supervised counterparts in most downstream tasks. In this work, we propose a semi-supervised sentence embedding framework, GenSE, that effectively leverages large-scale unlabeled data. Our method include three parts: 1) Generate: A generator/discriminator model is jointly trained to synthesize sentence pairs from open-domain unlabeled corpus; 2) Discriminate: Noisy sentence pairs are filtered out by the discriminator to acquire high-quality positive and negative sentence pairs; 3) Contrast: A prompt-based contrastive approach is presented for sentence representation learning with both annotated and synthesized data. Comprehensive experiments show that GenSE achieves an average correlation score of 85.19 on the STS datasets and consistent performance improvement on four domain adaptation tasks, significantly surpassing the state-of-the-art methods and convincingly corroborating its effectiveness and generalization ability.

The emotional state of a speaker can be influenced by many different factors in dialogues, such as dialogue scene, dialogue topic, and interlocutor stimulus. The currently available data resources to support such multimodal affective analysis in dialogues are however limited in scale and diversity. In this work, we propose a Multi-modal Multi-scene Multi-label Emotional Dialogue dataset, M³ED, which contains 990 dyadic emotional dialogues from 56 different TV series, a total of 9,082 turns and 24,449 utterances. M³ED is annotated with 7 emotion categories (happy, surprise, sad, disgust, anger, fear, and neutral) at utterance level, and encompasses acoustic, visual, and textual modalities. To the best of our knowledge, M³ED is the first multimodal emotional dialogue dataset in Chinese.It is valuable for cross-culture emotion analysis and recognition. We apply several state-of-the-art methods on the M³ED dataset to verify the validity and quality of the dataset. We also propose a general Multimodal Dialogue-aware Interaction framework, MDI, to model the dialogue context for emotion recognition, which achieves comparable performance to the state-of-the-art methods on the M³ED. The full dataset and codes are available.

Word and sentence embeddings are useful feature representations in natural language processing. However, intrinsic evaluation for embeddings lags far behind, and there has been no significant update since the past decade. Word and sentence similarity tasks have become the de facto evaluation method. It leads models to overfit to such evaluations, negatively impacting embedding models’ development. This paper first points out the problems using semantic similarity as the gold standard for word and sentence embedding evaluations. Further, we propose a new intrinsic evaluation method called EvalRank, which shows a much stronger correlation with downstream tasks. Extensive experiments are conducted based on 60+ models and popular datasets to certify our judgments. Finally, the practical evaluation toolkit is released for future benchmarking purposes.

As high-quality labeled data is scarce, unsupervised sentence representation learning has attracted much attention. In this paper, we propose a new framework with a two-branch Siamese Network which maximizes the similarity between two augmented views of each sentence. Specifically, given one augmented view of the input sentence, the online network branch is trained by predicting the representation yielded by the target network of the same sentence under another augmented view. Meanwhile, the target network branch is bootstrapped with a moving average of the online network. The proposed method significantly outperforms other state-of-the-art unsupervised methods on semantic textual similarity (STS) and classification tasks. It can be adopted as a post-training procedure to boost the performance of the supervised methods. We further extend our method for learning multilingual sentence representations and demonstrate its effectiveness on cross-lingual STS tasks. Our code is available at https://github.com/yanzhangnlp/BSL.

A dialogue is essentially a multi-turn interaction among interlocutors. Effective evaluation metrics should reflect the dynamics of such interaction. Existing automatic metrics are focused very much on the turn-level quality, while ignoring such dynamics. To this end, we propose DynaEval, a unified automatic evaluation framework which is not only capable of performing turn-level evaluation, but also holistically considers the quality of the entire dialogue. In DynaEval, the graph convolutional network (GCN) is adopted to model a dialogue in totality, where the graph nodes denote each individual utterance and the edges represent the dependency between pairs of utterances. A contrastive loss is then applied to distinguish well-formed dialogues from carefully constructed negative samples. Experiments show that DynaEval significantly outperforms the state-of-the-art dialogue coherence model, and correlates strongly with human judgements across multiple dialogue evaluation aspects at both turn and dialogue level.

As unlabeled data carry rich task-relevant information, they are proven useful for few-shot learning of language model. The question is how to effectively make use of such data. In this work, we revisit the self-training technique for language model fine-tuning and present a state-of-the-art prompt-based few-shot learner, SFLM. Given two views of a text sample via weak and strong augmentation techniques, SFLM generates a pseudo label on the weakly augmented version. Then, the model predicts the same pseudo label when fine-tuned with the strongly augmented version. This simple approach is shown to outperform other state-of-the-art supervised and semi-supervised counterparts on six sentence classification and six sentence-pair classification benchmarking tasks. In addition, SFLM only relies on a few in-domain unlabeled data. We conduct a comprehensive analysis to demonstrate the robustness of our proposed approach under various settings, including augmentation techniques, model scale, and few-shot knowledge transfer across tasks.

Language modeling is the technique to estimate the probability of a sequence of words. A bilingual language model is expected to model the sequential dependency for words across languages, which is difficult due to the inherent lack of suitable training data as well as diverse syntactic structure across languages. We propose a bilingual attention language model (BALM) that simultaneously performs language modeling objective with a quasi-translation objective to model both the monolingual as well as the cross-lingual sequential dependency. The attention mechanism learns the bilingual context from a parallel corpus. BALM achieves state-of-the-art performance on the SEAME code-switch database by reducing the perplexity of 20.5% over the best-reported result. We also apply BALM in bilingual lexicon induction, and language normalization tasks to validate the idea.

Customized translation need pay spe-cial attention to the target domain ter-minology especially the named-entities for the domain. Adding linguistic features to neural machine translation (NMT) has been shown to benefit translation in many studies. In this paper, we further demonstrate that adding named-entity (NE) feature with named-entity recognition (NER) into the source language produces better translation with NMT. Our experiments show that by just including the different NE classes and boundary tags, we can increase the BLEU score by around 1 to 2 points using the standard test sets from WMT2017. We also show that adding NE tags using NER and applying in-domain adaptation can be combined to further improve customized machine translation.

Transliteration is defined as phonetic translation of names across languages. Transliteration of Named Entities (NEs) is necessary in many applications, such as machine translation, corpus alignment, cross-language IR, information extraction and automatic lexicon acquisition. All such systems call for high-performance transliteration, which is the focus of shared task in the NEWS 2018 workshop. The objective of the shared task is to promote machine transliteration research by providing a common benchmarking platform for the community to evaluate the state-of-the-art technologies.

This report presents the results from the Named Entity Transliteration Shared Task conducted as part of The Seventh Named Entities Workshop (NEWS 2018) held at ACL 2018 in Melbourne, Australia. Similar to previous editions of NEWS, the Shared Task featured 19 tasks on proper name transliteration, including 13 different languages and two different Japanese scripts. A total of 6 teams from 8 different institutions participated in the evaluation, submitting 424 runs, involving different transliteration methodologies. Four performance metrics were used to report the evaluation results. The NEWS shared task on machine transliteration has successfully achieved its objectives by providing a common ground for the research community to conduct comparative evaluations of state-of-the-art technologies that will benefit the future research and development in this area.

In this paper, we describe the system and approach used by the Institute for Infocomm Research (I2R) for the IWSLT 2009 spoken language translation evaluation campaign. Two kinds of machine translation systems are applied, namely, phrase-based machine translation system and syntax-based machine translation system. To test syntax-based machine translation system on spoken language translation, variational systems are explored. On top of both phrase-based and syntax-based single systems, we further use rescoring method to improve the individual system performance and use system combination method to combine the strengths of the different individual systems. Rescoring is applied on each single system output, and system combination is applied on all rescoring outputs. Finally, our system combination framework shows better performance in Chinese-English BTEC task.

In this paper, we describe the system and approach used by the Institute for Infocomm Research (I2R) for the IWSLT 2008 spoken language translation evaluation campaign. In the system, we integrate various decoding algorithms into a multi-pass translation framework. The multi-pass approach enables us to utilize various decoding algorithm and to explore much more hypotheses. This paper reports our design philosophy, overall architecture, each individual system and various system combination methods that we have explored. The performance on development and test sets are reported in detail in the paper. The system has shown competitive performance with respect to the BLEU and METEOR measures in Chinese-English Challenge and BTEC tasks.

This paper gives a description of the statistical machine translation (SMT) systems developed at the TALP Research Center of the UPC (Universitat Polite`cnica de Catalunya) for our participation in the IWSLT’08 evaluation campaign. We present Ngram-based (TALPtuples) and phrase-based (TALPphrases) SMT systems. The paper explains the 2008 systems’ architecture and outlines translation schemes we have used, mainly focusing on the new techniques that are challenged to improve speech-to-speech translation quality. The novelties we have introduced are: improved reordering method, linear combination of translation and reordering models and new technique dealing with punctuation marks insertion for a phrase-based SMT system. This year we focus on the Arabic-English, Chinese-Spanish and pivot Chinese-(English)-Spanish translation tasks.

We propose a simplified Level Of Detail (LOD) algorithm to learn phrase translation for statistical machine translation. In particular, LOD learns unknown phrase translations from parallel texts without linguistic knowledge. LOD uses an agglomerative method to attack the combinatorial explosion that results when generating candidate phrase translations. Although LOD was previously proposed by (Setiawan et al., 2005), we improve the original algorithm in two ways: simplifying the algorithm and using a simpler translation model. Experimental results show that our algorithm provides comparable performance while demonstrating a significant reduction in computation time.