Prompts are the interface for eliciting the capabilities of large language models (LLMs). Understanding their structure and components is critical for analyzing LLM behavior and optimizing performance. However, the field lacks a comprehensive framework for systematic prompt analysis and understanding. We introduce PromptPrism, a linguistically-inspired taxonomy that enables prompt analysis across three hierarchical levels: functional structure, semantic component, and syntactic pattern. By applying linguistic concepts to prompt analysis, PromptPrism bridges traditional language understanding and modern LLM research, offering insights that purely empirical approaches might miss. We show the practical utility of PromptPrism by applying it to three applications: (1) a taxonomy-guided prompt refinement approach that automatically improves prompt quality and enhances model performance across a range of tasks; (2) a multi-dimensional dataset profiling method that extracts and aggregates structural, semantic, and syntactic characteristics from prompt datasets, enabling comprehensive analysis of prompt distributions and patterns; (3) a controlled experimental framework for prompt sensitivity analysis by quantifying the impact of semantic reordering and delimiter modifications on LLM performance. Our experimental results validate the effectiveness of our taxonomy across these applications, demonstrating that PromptPrism provides a foundation for refining, profiling, and analyzing prompts.

Training large foundation models for agentic tasks is increasingly impractical due to the high computational costs, long iteration cycles, and rapid obsolescence as new models are continuously released. Instead of post-training massive models for every new task or domain, we propose Supplement Generation Training (SGT), a more efficient and sustainable strategy. SGT trains a smaller LLM to generate useful supplemental text that, when appended to the original input, helps the larger LLM solve the task more effectively. These lightweight models can dynamically adapt supplements to task requirements, improving performance without modifying the underlying large models. This approach decouples task-specific optimization from large foundation models and enables more flexible, cost-effective deployment of LLM-powered agents in real-world applications.

LLM-based multi-agent systems (MAS) show promise on complex tasks but remain prone to coordination failures such as goal drift, error cascades, and misaligned behaviors. We propose Explicit Trait Inference (ETI), a psychologically grounded method for improving coordination. ETI enables agents to infer and track partner characteristics along two established psychological dimensions—warmth (e.g., trust) and competence (e.g., skill)—from interaction histories to guide decisions. We evaluate ETI in controlled settings (economic games), where it reduces payoff loss by 45–77%, and in more realistic, complex multi-agent settings (MultiAgentBench), where it improves performance by 3–29% depending on the scenario and model, relative to a CoT baseline. Additional analysis shows that gains are closely linked to trait inference: ETI profiles predict agents’ actions, and informative profiles drive improvements. These results highlight ETI as a lightweight and robust mechanism for improving coordination in diverse multi-agent settings, and provide the first systematic evidence that LLM agents can (i) reliably infer others’ traits from interaction histories and (ii) leverage structured awareness of others’ traits for coordination.

Developing language model-based dialogue agents requires effective data to train models that can follow specific task logic. However, most existing data simulation methods focus on increasing diversity in language, topics, or dialogue acts at the utterance level, largely neglecting a critical aspect of task logic diversity at the dialogue level. This paper proposes a novel data simulation method designed to enhance the diversity of synthetic dialogues by focusing on task execution logic. Our method uses LLMs to generate decision tree-structured task plans, which enables the derivation of diverse dialogue trajectories for a given task. Each trajectory, referred to as a “dialog flow”, guides the generation of a multi-turn dialogue that follows a unique trajectory. We apply this method to generate a task-oriented dialogue dataset comprising 3,886 dialogue flows across 15 different domains. We validate the effectiveness of this dataset using the next action prediction task, where models fine-tuned on our dataset outperform strong baselines, including GPT-4. Upon acceptance of this paper, we plan to release the code and data publicly.

Recent works have demonstrated that incorporating search during inference can significantly improve reasoning capabilities of language agents. Some approaches may make use of the ground truth or rely on model’s own generated feedback. The search algorithm uses this feedback to then produce values that will update its criterion for exploring and exploiting various reasoning paths. In this study, we investigate how search and model’s self-feedback can be leveraged for reasoning tasks. First, we explore differences in ground-truth feedback and self-feedback during search for math reasoning. Second, we observe limitations in applying search techniques to more complex tasks like tool-calling and design domain-specific approaches to address these gaps. Our experiments reveal challenges related to generalization when solely relying on self-feedback during search. For search to work effectively, either access to the ground-truth is needed or feedback mechanisms need to be carefully designed for the specific task.

Open domain question answering systems frequently rely on information retrieved from large collections of text (such as the Web) to answer questions. However, such collections of text often contain conflicting information, and indiscriminately depending on this information may result in untruthful and inaccurate answers. To understand the gravity of this problem, we collect a human-annotated dataset, Question Answering with Conflicting Contexts (QACC), and find that as much as 25% of unambiguous, open domain questions can lead to conflicting contexts when retrieved using Google Search. We evaluate and benchmark three powerful Large Language Models (LLMs) with our dataset QACC and demonstrate their limitations in effectively addressing questions with conflicting information. To explore how humans reason through conflicting contexts, we request our annotators to provide explanations for their selections of correct answers. We demonstrate that by finetuning LLMs to explain their answers, we can introduce richer information into their training that guide them through the process of reasoning with conflicting contexts. We publicly release our dataset and code to promote research along this line.

Temporal reasoning in multi-session dialogues presents a significant challenge which has been under-studied in previous temporal reasoning benchmarks. To bridge this gap, we propose a new evaluation task for temporal reasoning in multi-session dialogues and introduce an approach to construct a new benchmark by augmenting dialogues from LoCoMo and creating multi-choice QAs. Furthermore, we present TReMu, a new framework aimed at enhancing the temporal reasoning capabilities of LLM-agents in this context. Specifically, the framework employs time-aware memorization through timeline summarization, generating retrievable memory by summarizing events in each dialogue session with their inferred dates. Additionally, we integrate neuro-symbolic temporal reasoning, where LLMs generate Python code to perform temporal calculations and select answers. Experimental evaluations on popular LLMs demonstrate that our benchmark is challenging, and the proposed framework significantly improves temporal reasoning performance compared to baseline methods, raising from 29.83 on GPT-4o via standard prompting to 77.67 via our approach and highlighting its effectiveness in addressing temporal reasoning in multi-session dialogues.

Despite the rapid advancements in LLM agents, they still face the challenge of generating meaningful reflections due to inadequate error analysis and a reliance on rare successful trajectories, especially in complex tasks. In this work, we propose SAMULE, a new framework for self-learning agents powered by a retrospective language model that is trained based on Multi-Level Reflection Synthesis. It first synthesizes high-quality reflections across three complementary levels: Single-Trajectory Learning (micro-level) for detailed error correction; Intra-Task Learning (meso-level) to build error taxonomies across multiple trials of the same task, and Inter-Task Learning (macro-level) to extract transferable insights based on same typed errors from diverse task failures. Then we fine-tune a language model serving as the retrospective model to generate reflections during inference. We further extend our framework to interactive settings through a foresight-based reflection mechanism, enabling agents to proactively reflect and adapt during user interactions by comparing predicted and actual responses. Extensive experiments on three challenging benchmarks—TravelPlanner, NATURAL PLAN, and Tau-bench—demonstrate that our approach significantly outperforms reflection-based baselines. Our results highlight the critical role of well-designed reflection synthesis and failure-centric learning in building self-improving LLM agents.

Large language model (LLM) agents have evolved to intelligently process information, make decisions, and interact with users or tools. A key capability is the integration of long-term memory capabilities, enabling these agents to draw upon historical interactions and knowledge. However, the growing memory size and need for semantic structuring pose significant challenges. In this work, we propose an autonomous memory augmentation approach, MemInsight, to enhance semantic data representation and retrieval mechanisms. By leveraging autonomous augmentation to historical interactions, LLM agents are shown to deliver more accurate and contextualized responses. We empirically validate the efficacy of our proposed approach in three task scenarios; conversational recommendation, question answering and event summarization. On the LLM-REDIAL dataset, MemInsight boosts persuasiveness of recommendations by up to 14%. Moreover, it outperforms a RAG baseline by 34% in recall for LoCoMo retrieval. Our empirical results show the potential of MemInsight to enhance the contextual performance of LLM agents across multiple tasks.

Routing incoming queries to the most cost-effective LLM while maintaining response quality poses a fundamental challenge in optimizing performance-cost trade-offs for large-scale commercial systems.We present IPR—a quality-constrained Intelligent Prompt Routing framework that dynamically selects optimal models based on predicted response quality and user-specified tolerance levels.IPR introduces three key innovations: (1) a modular architecture with lightweight quality estimators trained on 1.5M prompts annotated with calibrated quality scores, enabling fine-grained quality prediction across model families; (2) a user-controlled routing mechanism with tolerance parameter 𝜏 ∈ [0,1] that provides explicit control over quality-cost trade-offs; and (3) an extensible design using frozen encoders with model-specific adapters, reducing new model integration from days to hours. To rigorously train and evaluate IPR, we curate an industrial-level IPR dataset, a comprehensive benchmark containing 1.5 million examples with response quality annotations across 11 LLM candidates.Deployed on a major cloud platform, IPR achieves 43.9% cost reduction while maintaining quality parity with the strongest model in the Claude family and processes requests with sub-150ms latency.

Document-grounded dialogue systems aim to answer user queries by leveraging external information. Previous studies have mainly focused on handling free-form documents, often overlooking structured data such as lists, which can represent a range of nuanced semantic relations. Motivated by the observation that even advanced language models like GPT-3.5 often miss semantic cues from lists, this paper aims to enhance question answering (QA) systems for better interpretation and use of structured lists. To this end, we introduce the LIST2QA dataset, a novel benchmark to evaluate the ability of QA systems to respond effectively using list information. This dataset is created from unlabeled customer service documents using language models and model-based filtering processes to enhance data quality, and can be used to fine-tune and evaluate QA models. Apart from directly generating responses through fine-tuned models, we further explore the explicit use of Intermediate Steps for Lists (ISL), aligning list items with user backgrounds to better reflect how humans interpret list items before generating responses. Our experimental results demonstrate that models trained on LIST2QA with our ISL approach outperform baselines across various metrics. Specifically, our fine-tuned Flan-T5-XL model shows increases of 3.1% in ROUGE-L, 4.6% in correctness, 4.5% in faithfulness, and 20.6% in completeness compared to models without applying filtering and the proposed ISL method.

We introduce Conversational Function-Calling Evaluation Through Turn-Level Interactions (CONFETTI), a conversational benchmark designed to evaluate the function-calling capabilities and response quality of large language models (LLMs). Current benchmarks lack comprehensive assessment of LLMs in complex conversational scenarios. CONFETTI addresses this gap through 109 human-simulated conversations, comprising 313 user turns and covering 86 APIs. These conversations explicitly target various conversational complexities, such as follow-ups, goal correction and switching, ambiguous and implicit goals. We perform off-policy turn-level evaluation using this benchmark targeting function-calling. Our benchmark also incorporates dialog act annotations to assess agent responses. We evaluate a series of state-of-the-art LLMs and analyze their performance with respect to the number of available APIs, conversation lengths, and chained function calling. Our results reveal that while some models are able to handle long conversations, and leverage more than 20+ APIs successfully, other models struggle with longer context or when increasing the number of APIs. We also report that the performance on chained function-calls is severely limited across the models. Overall, the top performing models onCONFETTI are Nova Pro (40.01%), Claude Sonnet v3.5 (35.46%) and Llama 3.1 405B (33.19%) followed by command-r-plus (31.18%) and Mistral-Large-2407 (30.07%).

Single document news summarization has seen substantial progress on faithfulness in recent years, driven by research on the evaluation of factual consistency, or hallucinations. We ask whether these advances carry over to other text summarization domains. We propose a new evaluation benchmark on topic-focused dialogue summarization, generated by LLMs of varying sizes. We provide binary sentence- level human annotations of the factual consistency of these summaries along with detailed explanations of factually inconsistent sentences. Our analysis shows that existing LLMs hallucinate significant amounts of factual errors in the dialogue domain, regardless of the model’s size. On the other hand, when LLMs, including GPT-4, serve as binary factual evaluators, they perform poorly and can be outperformed by prevailing state-of-the-art specialized factuality evaluation metrics. Finally, we conducted an analysis of hallucination types with a curated error taxonomy. We find that there are diverse errors and error distributions in model-generated summaries and that non-LLM based metrics can capture all error types better than LLM-based evaluators.

Large language models (LLMs) are powerful dialogue agents, but specializing them towards fulfilling a specific function can be challenging. Instructing tuning, i.e. tuning models on instruction and sample responses generated by humans (Ouyang et al., 2022), has proven as an effective method to do so, yet requires a number of data samples that a) might not be available or b) costly to generate. Furthermore, this cost increases when the goal is to make the LLM follow a specific workflow within a dialogue instead of single instructions. Inspired by the self-play technique in reinforcement learning and the use of LLMs to simulate human agents, we propose a more effective method for data collection through LLMs engaging in a conversation in various roles. This approach generates a training data via “self-talk” of LLMs that can be refined and utilized for supervised fine-tuning. We introduce an automated way to measure the (partial) success of a dialogue. This metric is used to filter the generated conversational data that is fed back in LLM for training. Based on our automated and human evaluations of conversation quality, we demonstrate that such self-talk data improves results. In addition, we examine the various characteristics that showcase the quality of generated dialogues and how they can be connected to their potential utility as training data.

Large language model advancements have enabled the development of multi-agent frameworks to tackle complex, real-world problems such as to automate workflows that require interactions with diverse tools, reasoning, and human collaboration. We present MARCO, a Multi-Agent Real-time Chat Orchestration framework for automating workflows using LLMs. MARCO addresses key challenges in utilizing LLMs for complex, multi-step task execution in a production environment. It incorporates robust guardrails to steer LLM behavior, validate outputs, and recover from errors that stem from inconsistent output formatting, function and parameter hallucination, and lack of domain knowledge. Through extensive experiments we demonstrate MARCO’s superior performance with 94.48% and 92.74% accuracy on task execution for Digital Restaurant Service Platform conversations and Retail conversations datasets respectively along with 44.91% improved latency and 33.71% cost reduction in a production setting. We also report effects of guardrails in performance gain along with comparisons of various LLM models, both open-source and proprietary. The modular and generic design of MARCO allows it to be adapted for automating workflows across domains and to execute complex tasks through multi-turn interactions.

Backward compatibility of model predictions is a desired property when updating a machine learning driven application. It allows to seamlessly improve the underlying model without introducing regression bugs. In classification tasks these bugs occur in the form of negative flips. This means an instance that was correctly classified by the old model is now classified incorrectly by the updated model. This has direct negative impact on the user experience of such systems e.g. a frequently used voice assistant query is suddenly misclassified.A common reason to update the model is when new training data becomes available and needs to be incorporated. Simply retraining the model with the updated data introduces the unwanted negative flips. We study the problem of regression during data updates and propose Backward Compatible Weight Interpolation (BCWI). This method interpolates between the weights of the old and new model and we show in extensive experiments that it reduces negative flips without sacrificing the improved accuracy of the new model. BCWI is straight forward to implement and does not increase inference cost. We also explore the use of importance weighting during interpolation and averaging the weights of multiple new models in order to further reduce negative flips.

Dialogue act annotations are important to improve response generation quality in task-oriented dialogue systems. However, it can be challenging to use dialogue acts to control response generation in a generalizable way because different datasets and tasks may have incompatible annotations. While alternative methods that utilize latent action spaces or reinforcement learning do not require explicit annotations, they may lack interpretability or face difficulties defining task-specific rewards. In this work, we present a novel end-to-end latent dialogue act model (DiactTOD) that represents dialogue acts in a latent space. DiactTOD, when pre-trained on a large corpus, is able to predict and control dialogue acts to generate controllable responses using these latent representations in a zero-shot fashion. Our approach demonstrates state-of-the-art performance across a wide range of experimental settings on the MultiWOZ dataset, including zero-shot, few-shot, and full data fine-tuning with both end-to-end and policy optimization configurations.

In executable task-oriented semantic parsing, the system aims to translate users’ utterances in natural language to machine-interpretable programs (API calls) that can be executed according to pre-defined API specifications. With the popularity of Large Language Models (LLMs), in-context learning offers a strong baseline for such scenarios, especially in data-limited regimes. However, LLMs are known to hallucinate and therefore pose a formidable challenge in constraining generated content. Thus, it remains uncertain if LLMs can effectively perform task-oriented utterance-to-API generation, where respecting the API’s structural and task-specific constraints is crucial. In this work, we seek to measure, analyze and mitigate such constraints violations. First, we identify the categories of various constraints in obtaining API-semantics from task-oriented utterances, and define fine-grained metrics that complement traditional ones. Second, we leverage these metrics to conduct a detailed error analysis of constraints violations seen in state-of-the-art LLMs, which motivates us to investigate two popular mitigation strategies– Semantic-Retrieval of Demonstrations (SRD) and API-aware Constrained Decoding (API-CD). Our experiments show that these strategies are effective at reducing constraints violations and improving the quality of the generated API calls, but require careful consideration given their implementation complexity and latency.

When upgrading neural models to a newer version, new errors that were not encountered in the legacy version can be introduced, known as regression errors. This inconsistent behavior during model upgrade often outweighs the benefits of accuracy gain and hinders the adoption of new models. To mitigate regression errors from model upgrade, distillation and ensemble have proven to be viable solutions without significant compromise in performance. Despite the progress, these approaches attained an incremental reduction in regression which is still far from achieving backward-compatible model upgrade. In this work, we propose a novel method, Gated Fusion, that promotes backward compatibility via learning to mix predictions between old and new models. Empirical results on two distinct model upgrade scenarios show that our method reduces the number of regression errors by 62% on average, outperforming the strongest baseline by an average of 25%.

Despite growing interest in applications based on natural customer support conversations,there exist remarkably few publicly available datasets that reflect the expected characteristics of conversations in these settings. Existing task-oriented dialogue datasets, which were collected to benchmark dialogue systems mainly in written human-to-bot settings, are not representative of real customer support conversations and do not provide realistic benchmarks for systems that are applied to natural data. To address this gap, we introduce NatCS, a multi-domain collection of spoken customer service conversations. We describe our process for collecting synthetic conversations between customers and agents based on natural language phenomena observed in real conversations. Compared to previous dialogue datasets, the conversations collected with our approach are more representative of real human-to-human conversations along multiple metrics. Finally, we demonstrate potential uses of NatCS, including dialogue act classification and intent induction from conversations as potential applications, showing that dialogue act annotations in NatCS provide more effective training data for modeling real conversations compared to existing synthetic written datasets. We publicly release NatCS to facilitate research in natural dialog systems

Intent classification (IC) plays an important role in task-oriented dialogue systems. However, IC models often generalize poorly when training without sufficient annotated examples for each user intent. We propose a novel pre-training method for text encoders that uses contrastive learning with intent psuedo-labels to produce embeddings that are well-suited for IC tasks, reducing the need for manual annotations. By applying this pre-training strategy, we also introduce Pre-trained Intent-aware Encoder (PIE), which is designed to align encodings of utterances with their intent names. Specifically, we first train a tagger to identify key phrases within utterances that are crucial for interpreting intents. We then use these extracted phrases to create examples for pre-training a text encoder in a contrastive manner. As a result, our PIE model achieves up to 5.4% and 4.0% higher accuracy than the previous state-of-the-art pre-trained text encoder for the N-way zero- and one-shot settings on four IC datasets.

With increasing demand for and adoption of virtual assistants, recent work has investigated ways to accelerate bot schema design through the automatic induction of intents or the induction of slots and dialogue states. However, a lack of dedicated benchmarks and standardized evaluation has made progress difficult to track and comparisons between systems difficult to make. This challenge track, held as part of the Eleventh Dialog Systems Technology Challenge, introduces a benchmark that aims to evaluate methods for the automatic induction of customer intents in a realistic setting of customer service interactions between human agents and customers. We propose two subtasks for progressively tackling the automatic induction of intents and corresponding evaluation methodologies. We then present three datasets suitable for evaluating the tasks and propose simple baselines. Finally, we summarize the submissions and results of the challenge track, for which we received submissions from 34 teams.

Dialogue meaning representation formulates natural language utterance semantics in their conversational context in an explicit and machine-readable form. Previous work typically follows the intent-slot framework, which is easy for annotation yet limited in scalability for complex linguistic expressions. A line of works alleviates the representation issue by introducing hierarchical structures but challenging to express complex compositional semantics, such as negation and coreference. We propose Dialogue Meaning Representation (DMR), a pliable and easily extendable representation for task-oriented dialogue. Our representation contains a set of nodes and edges to represent rich compositional semantics. Moreover, we propose an inheritance hierarchy mechanism focusing on domain extensibility. Additionally, we annotated DMR-FastFood, a multi-turn dialogue dataset with more than 70k utterances, with DMR. We propose two evaluation tasks to evaluate different dialogue models and a novel coreference resolution model GNNCoref for the graph-based coreference resolution task. Experiments show that DMR can be parsed well with pre-trained Seq2Seq models, and GNNCoref outperforms the baseline models by a large margin.The dataset and code are available at https://github.com/amazon-research/dialogue-meaning-representation

Pre-trained language models (PLM) have advanced the state-of-the-art across NLP applications, but lack domain-specific knowledge that does not naturally occur in pre-training data. Previous studies augmented PLMs with symbolic knowledge for different downstream NLP tasks. However, knowledge bases (KBs) utilized in these studies are usually large-scale and static, in contrast to small, domain-specific, and modifiable knowledge bases that are prominent in real-world task-oriented dialogue (TOD) systems. In this paper, we showcase the advantages of injecting domain-specific knowledge prior to fine-tuning on TOD tasks. To this end, we utilize light-weight adapters that can be easily integrated with PLMs and serve as a repository for facts learned from different KBs. To measure the efficacy of proposed knowledge injection methods, we introduce Knowledge Probing using Response Selection (KPRS) – a probe designed specifically for TOD models. Experiments on KPRS and the response generation task show improvements of knowledge injection with adapters over strong baselines.

In text classification tasks, useful information is encoded in the label names. Label semantic aware systems have leveraged this information for improved text classification performance during fine-tuning and prediction. However, use of label-semantics during pre-training has not been extensively explored. We therefore propose Label Semantic Aware Pre-training (LSAP) to improve the generalization and data efficiency of text classification systems. LSAP incorporates label semantics into pre-trained generative models (T5 in our case) by performing secondary pre-training on labeled sentences from a variety of domains. As domain-general pre-training requires large amounts of data, we develop a filtering and labeling pipeline to automatically create sentence-label pairs from unlabeled text. We perform experiments on intent (ATIS, Snips, TOPv2) and topic classification (AG News, Yahoo! Answers). LSAP obtains significant accuracy improvements over state-of-the-art models for few-shot text classification while maintaining performance comparable to state of the art in high-resource settings.

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.

Frame-based state representation is widely used in modern task-oriented dialog systems to model user intentions and slot values. However, a fixed design of domain ontology makes it difficult to extend to new services and APIs. Recent work proposed to use natural language descriptions to define the domain ontology instead of tag names for each intent or slot, thus offering a dynamic set of schema. In this paper, we conduct in-depth comparative studies to understand the use of natural language description for schema in dialog state tracking. Our discussion mainly covers three aspects: encoder architectures, impact of supplementary training, and effective schema description styles. We introduce a set of newly designed bench-marking descriptions and reveal the model robustness on both homogeneous and heterogeneous description styles in training and evaluation.

Recent studies have proposed different methods to improve multilingual word representations in contextualized settings including techniques that align between source and target embedding spaces. For contextualized embeddings, alignment becomes more complex as we additionally take context into consideration. In this work, we propose using Optimal Transport (OT) as an alignment objective during fine-tuning to further improve multilingual contextualized representations for downstream cross-lingual transfer. This approach does not require word-alignment pairs prior to fine-tuning that may lead to sub-optimal matching and instead learns the word alignments within context in an unsupervised manner. It also allows different types of mappings due to soft matching between source and target sentences. We benchmark our proposed method on two tasks (XNLI and XQuAD) and achieve improvements over baselines as well as competitive results compared to similar recent works.

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.

Behavior of deep neural networks can be inconsistent between different versions. Regressions during model update are a common cause of concern that often over-weigh the benefits in accuracy or efficiency gain. This work focuses on quantifying, reducing and analyzing regression errors in the NLP model updates. Using negative flip rate as regression measure, we show that regression has a prevalent presence across tasks in the GLUE benchmark. We formulate the regression-free model updates into a constrained optimization problem, and further reduce it into a relaxed form which can be approximately optimized through knowledge distillation training method. We empirically analyze how model ensemble reduces regression. Finally, we conduct CheckList behavioral testing to understand the distribution of regressions across linguistic phenomena, and the efficacy of ensemble and distillation methods.

Intent classification (IC) and slot filling (SF) are core components in most goal-oriented dialogue systems. Current IC/SF models perform poorly when the number of training examples per class is small. We propose a new few-shot learning task, few-shot IC/SF, to study and improve the performance of IC and SF models on classes not seen at training time in ultra low resource scenarios. We establish a few-shot IC/SF benchmark by defining few-shot splits for three public IC/SF datasets, ATIS, TOP, and Snips. We show that two popular few-shot learning algorithms, model agnostic meta learning (MAML) and prototypical networks, outperform a fine-tuning baseline on this benchmark. Prototypical networks achieves significant gains in IC performance on the ATIS and TOP datasets, while both prototypical networks and MAML outperform the baseline with respect to SF on all three datasets. In addition, we demonstrate that joint training as well as the use of pre-trained language models, ELMo and BERT in our case, are complementary to these few-shot learning methods and yield further gains.

Summarizing data samples by quantitative measures has a long history, with descriptive statistics being a case in point. However, as natural language processing methods flourish, there are still insufficient characteristic metrics to describe a collection of texts in terms of the words, sentences, or paragraphs they comprise. In this work, we propose metrics of diversity, density, and homogeneity that quantitatively measure the dispersion, sparsity, and uniformity of a text collection. We conduct a series of simulations to verify that each metric holds desired properties and resonates with human intuitions. Experiments on real-world datasets demonstrate that the proposed characteristic metrics are highly correlated with text classification performance of a renowned model, BERT, which could inspire future applications.

Pre-trained language models that learn contextualized word representations from a large un-annotated corpus have become a standard component for many state-of-the-art NLP systems. Despite their successful applications in various downstream NLP tasks, the extent of contextual impact on the word representation has not been explored. In this paper, we present a detailed analysis of contextual impact in Transformer- and BiLSTM-based masked language models. We follow two different approaches to evaluate the impact of context: a masking based approach that is architecture agnostic, and a gradient based approach that requires back-propagation through networks. The findings suggest significant differences on the contextual impact between the two model architectures. Through further breakdown of analysis by syntactic categories, we find the contextual impact in Transformer-based MLM aligns well with linguistic intuition. We further explore the Transformer attention pruning based on our findings in contextual analysis.

This paper describes the system submission of our team Amazon to the shared task on Cross Framework Meaning Representation Parsing (MRP) at the 2019 Conference for Computational Language Learning (CoNLL). Via extensive analysis of implicit alignments in AMR, we recategorize five meaning representations (MRs) into two classes: Lexical- Anchoring and Phrasal-Anchoring. Then we propose a unified graph-based parsing framework for the lexical-anchoring MRs, and a phrase-structure parsing for one of the phrasal- anchoring MRs, UCCA. Our system submission ranked 1st in the AMR subtask, and later improvements show promising results on other frameworks as well.

Hierarchical neural networks are often used to model inherent structures within dialogues. For goal-oriented dialogues, these models miss a mechanism adhering to the goals and neglect the distinct conversational patterns between two interlocutors. In this work, we propose Goal-Embedded Dual Hierarchical Attentional Encoder-Decoder (G-DuHA) able to center around goals and capture interlocutor-level disparity while modeling goal-oriented dialogues. Experiments on dialogue generation, response generation, and human evaluations demonstrate that the proposed model successfully generates higher-quality, more diverse and goal-centric dialogues. Moreover, we apply data augmentation via goal-oriented dialogue generation for task-oriented dialog systems with better performance achieved.

The need for high-quality, large-scale, goal-oriented dialogue datasets continues to grow as virtual assistants become increasingly wide-spread. However, publicly available datasets useful for this area are limited either in their size, linguistic diversity, domain coverage, or annotation granularity. In this paper, we present strategies toward curating and annotating large scale goal oriented dialogue data. We introduce the MultiDoGO dataset to overcome these limitations. With a total of over 81K dialogues harvested across six domains, MultiDoGO is over 8 times the size of MultiWOZ, the other largest comparable dialogue dataset currently available to the public. Over 54K of these harvested conversations are annotated for intent classes and slot labels. We adopt a Wizard-of-Oz approach wherein a crowd-sourced worker (the “customer”) is paired with a trained annotator (the “agent”). The data curation process was controlled via biases to ensure a diversity in dialogue flows following variable dialogue policies. We provide distinct class label tags for agents vs. customer utterances, along with applicable slot labels. We also compare and contrast our strategies on annotation granularity, i.e. turn vs. sentence level. Furthermore, we compare and contrast annotations curated by leveraging professional annotators vs the crowd. We believe our strategies for eliciting and annotating such a dialogue dataset scales across modalities and domains and potentially languages in the future. To demonstrate the efficacy of our devised strategies we establish neural baselines for classification on the agent and customer utterances as well as slot labeling for each domain.

In recent years the use of electronic medical records has accelerated resulting in large volumes of medical data when a patient visits a healthcare facility. As a first step towards reimbursement healthcare institutions need to associate ICD-10 billing codes to these documents. This is done by trained clinical coders who may use a computer assisted solution for shortlisting of codes. In this work, we present our work to build a machine learning based scalable system for predicting ICD-10 codes from electronic medical records. We address data imbalance issues by implementing two system architectures using convolutional neural networks and logistic regression models. We illustrate the pros and cons of those system designs and show that the best performance can be achieved by leveraging the advantages of both using a system combination approach.

Extracting instances of sentiment-oriented relations from user-generated web documents is important for online marketing analysis. Unlike previous work, we formulate this extraction task as a structured prediction problem and design the corresponding inference as an integer linear program. Our latent structural SVM based model can learn from training corpora that do not contain explicit annotations of sentiment-bearing expressions, and it can simultaneously recognize instances of both binary (polarity) and ternary (comparative) relations with regard to entity mentions of interest. The empirical evaluation shows that our approach significantly outperforms state-of-the-art systems across domains (cameras and movies) and across genres (reviews and forum posts). The gold standard corpus that we built will also be a valuable resource for the community.

In this paper, we report on first attempts and findings to analyzing German patient records, using a hybrid parsing architecture and a combination of two relation extraction strategies. On a practical level, we are interested in the extraction of concepts and relations among those concepts, a necessary cornerstone for building medical information systems. The parsing pipeline consists of a morphological analyzer, a robust chunk parser adapted to Latin phrases used in medical diagnosis, a repair rule stage, and a probabilistic context-free parser that respects the output from the chunker. The relation extraction stage is a combination of two systems: SProUT, a shallow processor which uses hand-written rules to discover relation instances from local text units and DARE which extracts relation instances from complete sentences, using rules that are learned in a bootstrapping process, starting with semantic seeds. Two small experiments have been carried out for the parsing pipeline and the relation extraction stage.

This paper introduces the CLIMB (Comparative Libraries of Implementations with Matrix Basis) methodology and grammars. The basic idea behind CLIMB is to use code generation as a general methodology for grammar development in order to create a more systematic approach to grammar development. The particular method used in this paper is closely related to the LinGO Grammar Matrix. Like the Grammar Matrix, resulting grammars are HPSG grammars that can map bidirectionally between strings and MRS representations. The main purpose of this paper is to provide insight into the process of using CLIMB for grammar development. In addition, we describe three projects that make use of this methodology or have concrete plans to adapt CLIMB in the future: CLIMB for Germanic languages, CLIMB for Slavic languages and CLIMB to combine two grammars of Mandarin Chinese. We present the first results that indicate feasibility and development time improvements for creating a medium to large coverage precision grammar.

We present the ongoing development of MCG, a linguistically deep and precise grammar for Mandarin Chinese together with its accompanying treebank, both based on the linguistic framework of HPSG, and using MRS as the semantic representation. We highlight some key features of our grammar design, and review a number of challenging phenomena, with comparisons to alternative linguistic treatments and implementations. One of the distinguishing characteristics of our approach is the tight integration of grammar and treebank development. The two-step treebank annotation procedure benefits from the efficiency of the discriminant-based annotation approach, while giving the annotators full freedom of producing extra-grammatical structures. This not only allows the creation of a precise and full-coverage treebank with an imperfect grammar, but also provides prompt feedback for grammarians to identify the errors in the grammar design and implementation. Preliminary evaluation and error analysis shows that the grammar already covers most of the core phenomena for Mandarin Chinese, and the treebank annotation procedure reaches a stable speed of 35 sentences per hour with satisfying quality.

This paper presents a question generation system based on the approach of semantic rewriting. The state-of-the-art deep linguistic parsing and generation tools are employed to convert (back and forth) between the natural language sentences and their meaning representations in the form of Minimal Recursion Semantics (MRS). By carefully operating on the semantic structures, we show a principled way of generating questions without ad-hoc manipulation of the syntactic structures. Based on the (partial) understanding of the sentence meaning, the system generates questions which are semantically grounded and purposeful. And with the support of deep linguistic grammars, the grammaticality of the generation results is warranted. Further, with a specialized ranking model, the linguistic realizations from the general purpose generation model are further refined for our the question generation task. The evaluation results from QGSTEC2010 show promising prospects of the proposed approach.

In this paper, we describe our hybrid parsing model on the Mandarin Chinese processing. In particular, we work on the Tsinghua Chinese Treebank (TCT), whose annotation has both constitutes and the head information of each constitute. The model we design combines the mainstream constitute parsing and dependency parsing. We present in detail 1) how to (partially) encode the head information into the constitute parsing, 2) how to encode constitute information into the dependency parsing, and 3) how to restore the head information using the dependency structure. For each of them, we take different strategies to deal with different cases. In an open shared task evaluation, we achieve an f1-score of 85.23% for the constitute parsing, 82.35% with partial head information, and 74.27% with complete head information. The error analysis shows the challenge of restoring multiple-headed constitutes and also some potentials to use the dependency structure to guide the constitute parsing, which will be our future work to explore.

The aim of this paper is twofold. We focus, on the one hand, on the task of dynamically annotating English compound nouns, and on the other hand we propose disambiguation methods and techniques which facilitate the annotation task. Both the aforementioned are part of a larger on-going effort which aims to create HPSG annotation for the texts from theWall Street Journal (henceforward WSJ) sections of the Penn Treebank (henceforward PTB) with the help of a hand-written large-scale and wide-coverage grammar of English, the English Resource Grammar (henceforward ERG; Flickinger (2002)). As we show in this paper, such annotations are very rich linguistically, since apart from syntax they also incorporate semantics, which does not only ensure that the treebank is guaranteed to be a truly sharable, re-usable and multi-functional linguistic resource, but also calls for the necessity of a better disambiguation of the internal (syntactic) structure of larger units of words, such as compound nouns, since this has an impact on the representation of their meaning, which is of utmost interest if the linguistic annotation of a given corpus is to be further understood as the practice of adding interpretative linguistic information of the highest quality in order to give “added value” to the corpus.

The task of parse disambiguation has gained in importance over the last decade as the complexity of grammars used in deep linguistic processing has been increasing. In this paper we propose to employ the fine-grained HPSG formalism in order to investigate the contribution of deeper linguistic knowledge to the task of ranking the different trees the parser outputs. In particular, we focus on the incorporation of semantic features in the disambiguation component and the stability of our model cross domains. Our work is carried out within DELPH-IN (http://www.delph-in.net), using the LinGo Redwoods and the WeScience corpora, parsed with the English Resource Grammar and the PET parser.

In this paper we propose a partial parsing model which achieves robust parsing with a large HPSG grammar. Constraint-based precision grammars, like the HPSG grammar we are using for the experiments reported in this paper, typically lack robustness, especially when applied to real world texts. To maximally recover the linguistic knowledge from an unsuccessful parse, a proper selection model must be used. Also, the efficiency challenges usually presented by the selection model must be answered. Building on the work reported in (Zhang et al., 2007), we further propose a new partial parsing model that splits the parsing process into two stages, both of which use the bottom-up chart-based parsing algorithm. The algorithm is implemented and a preliminary experiment shows promising results.

In this work, we examine and attempt to extend the coverage of a German HPSG grammar. We use the grammar to parse a corpus of newspaper text and evaluate the proportion of sentences which have a correct attested parse, and analyse the cause of errors in terms of lexical or constructional gaps which prevent parsing. Then, using a maximum entropy model, we evaluate prediction of lexical types in the HPSG type hierarchy for unseen lexemes. By automatically adding entries to the lexicon, we observe that we can increase coverage without substantially decreasing precision.

In this paper, we report on methods to detect and repair lexical errors for deep grammars. The lack of coverage has for long been the major problem for deep processing. The existence of various errors in the hand-crafted large grammars prevents their usage in real applications. The manual detection and repair of errors requires asignificant amount of human effort. An experiment with the British National Corpus shows about 70% of the sentences contain unknownword(s) for the English Resource Grammar. With the help of error mining methods, many lexical errors are discovered, which cause a large part of the parsing failures. Moreover, with a lexical type predictor based on a maximum entropy model, new lexical entries are automatically generated. The contribution of various features for the model is evaluated. With the disambiguated full parsing results, the precision of the predictor is enhanced significantly.