Kazuma Hashimoto

Tian Xie, Xinyi Yang, Angela S. Lin et al.

Creating a system that can have meaningful conversations with humans to help accomplish tasks is one of the ultimate goals of Artificial Intelligence (AI). It has defined the meaning of AI since the beginning. A lot has been accomplished in this area recently, with voice assistant products entering our daily lives and chat bot systems becoming commonplace in customer service. At first glance there seems to be no shortage of options for dialogue systems. However, the frequently deployed dialogue systems today seem to all struggle with a critical weakness - they are hard to build and harder to maintain. At the core of the struggle is the need to script every single turn of interactions between the bot and the human user. This makes the dialogue systems more difficult to maintain as the tasks become more complex and more tasks are added to the system. In this paper, we propose Converse, a flexible tree-based modular task-oriented dialogue system. Converse uses an and-or tree structure to represent tasks and offers powerful multi-task dialogue management. Converse supports task dependency and task switching, which are unique features compared to other open-source dialogue frameworks. At the same time, Converse aims to make the bot building process easy and simple, for both professional and non-professional software developers. The code is available at https://github.com/salesforce/Converse.

Karthik Raman, Iftekhar Naim, Jiecao Chen et al. · apple-ml, cmu

Pretrained, large, generative language models (LMs) have had great success in a wide range of sequence tagging and structured prediction tasks. Casting a sequence tagging task as a Seq2Seq one requires deciding the formats of the input and output sequences. However, we lack a principled understanding of the trade-offs associated with these formats (such as the effect on model accuracy, sequence length, multilingual generalization, hallucination). In this paper, we rigorously study different formats one could use for casting input text sentences and their output labels into the input and target (i.e., output) of a Seq2Seq model. Along the way, we introduce a new format, which we show to to be both simpler and more effective. Additionally the new format demonstrates significant gains in the multilingual settings -- both zero-shot transfer learning and joint training. Lastly, we find that the new format is more robust and almost completely devoid of hallucination -- an issue we find common in existing formats. With well over a 1000 experiments studying 14 different formats, over 7 diverse public benchmarks -- including 3 multilingual datasets spanning 7 languages -- we believe our findings provide a strong empirical basis in understanding how we should tackle sequence tagging tasks.

Haopeng Zhang, Semih Yavuz, Wojciech Kryscinski et al. · salesforce

Abstractive summarization systems leveraging pre-training language models have achieved superior results on benchmark datasets. However, such models have been shown to be more prone to hallucinate facts that are unfaithful to the input context. In this paper, we propose a method to remedy entity-level extrinsic hallucinations with Entity Coverage Control (ECC). We first compute entity coverage precision and prepend the corresponding control code for each training example, which implicitly guides the model to recognize faithfulness contents in the training phase. We further extend our method via intermediate fine-tuning on large but noisy data extracted from Wikipedia to unlock zero-shot summarization. We show that the proposed method leads to more faithful and salient abstractive summarization in supervised fine-tuning and zero-shot settings according to our experimental results on three benchmark datasets XSum, Pubmed, and SAMSum of very different domains and styles.

Tong Niu, Kazuma Hashimoto, Yingbo Zhou et al. · salesforce

Aligning parallel sentences in multilingual corpora is essential to curating data for downstream applications such as Machine Translation. In this work, we present OneAligner, an alignment model specially designed for sentence retrieval tasks. This model is able to train on only one language pair and transfers, in a cross-lingual fashion, to low-resource language pairs with negligible degradation in performance. When trained with all language pairs of a large-scale parallel multilingual corpus (OPUS-100), this model achieves the state-of-the-art result on the Tateoba dataset, outperforming an equally-sized previous model by 8.0 points in accuracy while using less than 0.6% of their parallel data. When finetuned on a single rich-resource language pair, be it English-centered or not, our model is able to match the performance of the ones finetuned on all language pairs under the same data budget with less than 2.0 points decrease in accuracy. Furthermore, with the same setup, scaling up the number of rich-resource language pairs monotonically improves the performance, reaching a minimum of 0.4 points discrepancy in accuracy, making it less mandatory to collect any low-resource parallel data. Finally, we conclude through empirical results and analyses that the performance of the sentence alignment task depends mostly on the monolingual and parallel data size, up to a certain size threshold, rather than on what language pairs are used for training or evaluation.

Lingyu Gao, Aditi Chaudhary, Krishna Srinivasan et al. · apple-ml, cmu

In-context learning (ICL) i.e. showing LLMs only a few task-specific demonstrations has led to downstream gains with no task-specific fine-tuning required. However, LLMs are sensitive to the choice of prompts, and therefore a crucial research question is how to select good demonstrations for ICL. One effective strategy is leveraging semantic similarity between the ICL demonstrations and test inputs by using a text retriever, which however is sub-optimal as that does not consider the LLM's existing knowledge about that task. From prior work (Lyu et al., 2023), we already know that labels paired with the demonstrations bias the model predictions. This leads us to our hypothesis whether considering LLM's existing knowledge about the task, especially with respect to the output label space can help in a better demonstration selection strategy. Through extensive experimentation on three text classification tasks, we find that it is beneficial to not only choose semantically similar ICL demonstrations but also to choose those demonstrations that help resolve the inherent label ambiguity surrounding the test example. Interestingly, we find that including demonstrations that the LLM previously mis-classified and also fall on the test example's decision boundary, brings the most performance gain.

Semih Yavuz, Kazuma Hashimoto, Yingbo Zhou et al.

Fusion-in-decoder (Fid) (Izacard and Grave, 2020) is a generative question answering (QA) model that leverages passage retrieval with a pre-trained transformer and pushed the state of the art on single-hop QA. However, the complexity of multi-hop QA hinders the effectiveness of the generative QA approach. In this work, we propose a simple generative approach (PathFid) that extends the task beyond just answer generation by explicitly modeling the reasoning process to resolve the answer for multi-hop questions. By linearizing the hierarchical reasoning path of supporting passages, their key sentences, and finally the factoid answer, we cast the problem as a single sequence prediction task. To facilitate complex reasoning with multiple clues, we further extend the unified flat representation of multiple input documents by encoding cross-passage interactions. Our extensive experiments demonstrate that PathFid leads to strong performance gains on two multi-hop QA datasets: HotpotQA and IIRC. Besides the performance gains, PathFid is more interpretable, which in turn yields answers that are more faithfully grounded to the supporting passages and facts compared to the baseline Fid model.

Man Luo, Kazuma Hashimoto, Semih Yavuz et al.

While both extractive and generative readers have been successfully applied to the Question Answering (QA) task, little attention has been paid toward the systematic comparison of them. Characterizing the strengths and weaknesses of the two readers is crucial not only for making a more informed reader selection in practice but also for developing a deeper understanding to foster further research on improving readers in a principled manner. Motivated by this goal, we make the first attempt to systematically study the comparison of extractive and generative readers for question answering. To be aligned with the state-of-the-art, we explore nine transformer-based large pre-trained language models (PrLMs) as backbone architectures. Furthermore, we organize our findings under two main categories: (1) keeping the architecture invariant, and (2) varying the underlying PrLMs. Among several interesting findings, it is important to highlight that (1) the generative readers perform better in long context QA, (2) the extractive readers perform better in short context while also showing better out-of-domain generalization, and (3) the encoder of encoder-decoder PrLMs (e.g., T5) turns out to be a strong extractive reader and outperforms the standard choice of encoder-only PrLMs (e.g., RoBERTa). We also study the effect of multi-task learning on the two types of readers varying the underlying PrLMs and perform qualitative and quantitative diagnosis to provide further insights into future directions in modeling better readers.

Kazuma Hashimoto, Iftekhar Naim, Karthik Raman

Sequence labeling is a core task in text understanding for IE/IR systems. Text generation models have increasingly become the go-to solution for such tasks (e.g., entity extraction and dialog slot filling). While most research has focused on the labeling accuracy, a key aspect -- of vital practical importance -- has slipped through the cracks: understanding model confidence. More specifically, we lack a principled understanding of how to reliably gauge the confidence of a model in its predictions for each labeled span. This paper aims to provide some empirical insights on estimating model confidence for generative sequence labeling. Most notably, we find that simply using the decoder's output probabilities \textbf{is not} the best in realizing well-calibrated confidence estimates. As verified over six public datasets of different tasks, we show that our proposed approach -- which leverages statistics from top-$k$ predictions by a beam search -- significantly reduces calibration errors of the predictions of a generative sequence labeling model.

Kazuma Hashimoto, Karthik Raman

Sequential labeling is a fundamental NLP task, forming the backbone of many applications. Supervised learning of Seq2Seq models has shown great success on these problems. However, the training objectives are still significantly disconnected with the metrics and desiderata we care about in practice. For example, a practical sequence tagging application may want to optimize for a certain precision-recall trade-off (of the top-k predictions) which is quite different from the standard objective of maximizing the likelihood of the gold labeled sequence. Thus to bridge this gap, we propose GROOT -- a simple yet effective framework for Generative Reward Optimization Of Text sequences. GROOT works by training a generative sequential labeling model to match the decoder output distribution with that of the (black-box) reward function. Using an iterative training regime, we first generate prediction candidates, then correct errors in them, and finally contrast those candidates (based on their reward values). As demonstrated via extensive experiments on four public benchmarks, GROOT significantly improves all reward metrics. Furthermore, GROOT leads to improvements of the overall decoder distribution as evidenced by the quality gains of the top-$k$ candidates.

Kazuma Hashimoto, Karthik Raman, Michael Bendersky

In-Context Learning (ICL) is an emergent capability of Large Language Models (LLMs). Only a few demonstrations enable LLMs to be used as blackbox for new tasks. Previous studies have shown that using LLMs' outputs as labels is effective in training models to select demonstrations. Such a label is expected to estimate utility of a demonstration in ICL; however, it has not been well understood how different labeling strategies affect results on target tasks. This paper presents an analysis on different utility functions by focusing on LLMs' output probability given ground-truth output, and task-specific reward given LLMs' prediction. Unlike the previous work, we introduce a novel labeling method, incremental utility, which estimates how much incremental knowledge is brought into the LLMs by a demonstration. We conduct experiments with instruction-tuned LLMs on binary/multi-class classification, segmentation, and translation across Arabic, English, Finnish, Japanese, and Spanish. Our results show that (1) the probability is effective when the probability values are distributed across the whole value range (on the classification tasks), and (2) the downstream metric is more robust when nuanced reward values are provided with long outputs (on the segmentation and translation tasks). We then show that the proposed incremental utility further helps ICL by contrasting how the LLMs perform with and without the demonstrations.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Jianguo Zhang, Kazuma Hashimoto, Yao Wan et al.

Pre-trained Transformer-based models were reported to be robust in intent classification. In this work, we first point out the importance of in-domain out-of-scope detection in few-shot intent recognition tasks and then illustrate the vulnerability of pre-trained Transformer-based models against samples that are in-domain but out-of-scope (ID-OOS). We construct two new datasets, and empirically show that pre-trained models do not perform well on both ID-OOS examples and general out-of-scope examples, especially on fine-grained few-shot intent detection tasks. To figure out how the models mistakenly classify ID-OOS intents as in-scope intents, we further conduct analysis on confidence scores and the overlapping keywords, as well as point out several prospective directions for future work. Resources are available on https://github.com/jianguoz/Few-Shot-Intent-Detection.

Arjun R. Akula, Kazuma Hashimoto, Krishna Srinivasan et al. · apple-ml, cmu

The emergence of long-context large language models (LLMs) has enabled the use of hundreds, or even thousands, of demonstrations for in-context learning (ICL) - a previously impractical regime. This paper investigates whether traditional ICL selection strategies, which balance the similarity of ICL examples to the test input (using a text retriever) with diversity within the ICL set, remain effective when utilizing a large number of demonstrations. Our experiments demonstrate that, while longer contexts can accommodate more examples, simply increasing the number of demonstrations does not guarantee improved performance. Smart ICL selection remains crucial, even with thousands of demonstrations. To further enhance ICL in this setting, we introduce Refract ICL, a novel ICL selection algorithm specifically designed to focus LLM attention on challenging examples by strategically repeating them within the context and incorporating zero-shot predictions as error signals. Our results show that Refract ICL significantly improves the performance of extremely long-context models such as Gemini 1.5 Pro, particularly on tasks with a smaller number of output classes.

Aditi Chaudhary, Karthik Raman, Krishna Srinivasan et al.

Query-document relevance prediction is a critical problem in Information Retrieval systems. This problem has increasingly been tackled using (pretrained) transformer-based models which are finetuned using large collections of labeled data. However, in specialized domains such as e-commerce and healthcare, the viability of this approach is limited by the dearth of large in-domain data. To address this paucity, recent methods leverage these powerful models to generate high-quality task and domain-specific synthetic data. Prior work has largely explored synthetic data generation or query generation (QGen) for Question-Answering (QA) and binary (yes/no) relevance prediction, where for instance, the QGen models are given a document, and trained to generate a query relevant to that document. However in many problems, we have a more fine-grained notion of relevance than a simple yes/no label. Thus, in this work, we conduct a detailed study into how QGen approaches can be leveraged for nuanced relevance prediction. We demonstrate that -- contrary to claims from prior works -- current QGen approaches fall short of the more conventional cross-domain transfer-learning approaches. Via empirical studies spanning 3 public e-commerce benchmarks, we identify new shortcomings of existing QGen approaches -- including their inability to distinguish between different grades of relevance. To address this, we introduce label-conditioned QGen models which incorporates knowledge about the different relevance. While our experiments demonstrate that these modifications help improve performance of QGen techniques, we also find that QGen approaches struggle to capture the full nuance of the relevance label space and as a result the generated queries are not faithful to the desired relevance label.

Ye Liu, Kazuma Hashimoto, Yingbo Zhou et al.

Dense neural text retrieval has achieved promising results on open-domain Question Answering (QA), where latent representations of questions and passages are exploited for maximum inner product search in the retrieval process. However, current dense retrievers require splitting documents into short passages that usually contain local, partial, and sometimes biased context, and highly depend on the splitting process. As a consequence, it may yield inaccurate and misleading hidden representations, thus deteriorating the final retrieval result. In this work, we propose Dense Hierarchical Retrieval (DHR), a hierarchical framework that can generate accurate dense representations of passages by utilizing both macroscopic semantics in the document and microscopic semantics specific to each passage. Specifically, a document-level retriever first identifies relevant documents, among which relevant passages are then retrieved by a passage-level retriever. The ranking of the retrieved passages will be further calibrated by examining the document-level relevance. In addition, hierarchical title structure and two negative sampling strategies (i.e., In-Doc and In-Sec negatives) are investigated. We apply DHR to large-scale open-domain QA datasets. DHR significantly outperforms the original dense passage retriever and helps an end-to-end QA system outperform the strong baselines on multiple open-domain QA benchmarks.

Mingfei Gao, Zeyuan Chen, Nikhil Naik et al.

We propose a novel framework to conduct field extraction from forms with unlabeled data. To bootstrap the training process, we develop a rule-based method for mining noisy pseudo-labels from unlabeled forms. Using the supervisory signal from the pseudo-labels, we extract a discriminative token representation from a transformer-based model by modeling the interaction between text in the form. To prevent the model from overfitting to label noise, we introduce a refinement module based on a progressive pseudo-label ensemble. Experimental results demonstrate the effectiveness of our framework.

Xi Ye, Semih Yavuz, Kazuma Hashimoto et al.

Existing KBQA approaches, despite achieving strong performance on i.i.d. test data, often struggle in generalizing to questions involving unseen KB schema items. Prior ranking-based approaches have shown some success in generalization, but suffer from the coverage issue. We present RnG-KBQA, a Rank-and-Generate approach for KBQA, which remedies the coverage issue with a generation model while preserving a strong generalization capability. Our approach first uses a contrastive ranker to rank a set of candidate logical forms obtained by searching over the knowledge graph. It then introduces a tailored generation model conditioned on the question and the top-ranked candidates to compose the final logical form. We achieve new state-of-the-art results on GrailQA and WebQSP datasets. In particular, our method surpasses the prior state-of-the-art by a large margin on the GrailQA leaderboard. In addition, RnG-KBQA outperforms all prior approaches on the popular WebQSP benchmark, even including the ones that use the oracle entity linking. The experimental results demonstrate the effectiveness of the interplay between ranking and generation, which leads to the superior performance of our proposed approach across all settings with especially strong improvements in zero-shot generalization.

Shrimai Prabhumoye, Kazuma Hashimoto, Yingbo Zhou et al.

Document grounded generation is the task of using the information provided in a document to improve text generation. This work focuses on two different document grounded generation tasks: Wikipedia Update Generation task and Dialogue response generation. Our work introduces two novel adaptations of large scale pre-trained encoder-decoder models focusing on building context driven representation of the document and enabling specific attention to the information in the document. Additionally, we provide a stronger BART baseline for these tasks. Our proposed techniques outperform existing methods on both automated (at least 48% increase in BLEU-4 points) and human evaluation for closeness to reference and relevance to the document. Furthermore, we perform comprehensive manual inspection of the generated output and categorize errors to provide insights into future directions in modeling these tasks.

Govardana Sachithanandam Ramachandran, Kazuma Hashimoto, Caiming Xiong

The recent success of reinforcement learning's (RL) in solving complex tasks is most often attributed to its capacity to explore and exploit an environment where it has been trained. Sample efficiency is usually not an issue since cheap simulators are available to sample data on-policy. On the other hand, task oriented dialogues are usually learnt from offline data collected using human demonstrations. Collecting diverse demonstrations and annotating them is expensive. Unfortunately, use of RL methods trained on off-policy data are prone to issues of bias and generalization, which are further exacerbated by stochasticity in human response and non-markovian belief state of a dialogue management system. To this end, we propose a batch RL framework for task oriented dialogue policy learning: causal aware safe policy improvement (CASPI). This method gives guarantees on dialogue policy's performance and also learns to shape rewards according to intentions behind human responses, rather than just mimicking demonstration data; this couple with batch-RL helps overall with sample efficiency of the framework. We demonstrate the effectiveness of this framework on a dialogue-context-to-text Generation and end-to-end dialogue task of the Multiwoz2.0 dataset. The proposed method outperforms the current state of the art on these metrics, in both case. In the end-to-end case, our method trained only on 10\% of the data was able to out perform current state in three out of four evaluation metrics.

Keisuke Shirai, Kazuma Hashimoto, Akiko Eriguchi et al.

Neural text generation models conditioning on given input (e.g. machine translation and image captioning) are usually trained by maximum likelihood estimation of target text. However, the trained models suffer from various types of errors at inference time. In this paper, we propose to suppress an arbitrary type of errors by training the text generation model in a reinforcement learning framework, where we use a trainable reward function that is capable of discriminating between references and sentences containing the targeted type of errors. We create such negative examples by artificially injecting the targeted errors to the references. In experiments, we focus on two error types, repeated and dropped tokens in model-generated text. The experimental results show that our method can suppress the generation errors and achieve significant improvements on two machine translation and two image captioning tasks.

Jian-Guo Zhang, Kazuma Hashimoto, Wenhao Liu et al.

Intent detection is one of the core components of goal-oriented dialog systems, and detecting out-of-scope (OOS) intents is also a practically important skill. Few-shot learning is attracting much attention to mitigate data scarcity, but OOS detection becomes even more challenging. In this paper, we present a simple yet effective approach, discriminative nearest neighbor classification with deep self-attention. Unlike softmax classifiers, we leverage BERT-style pairwise encoding to train a binary classifier that estimates the best matched training example for a user input. We propose to boost the discriminative ability by transferring a natural language inference (NLI) model. Our extensive experiments on a large-scale multi-domain intent detection task show that our method achieves more stable and accurate in-domain and OOS detection accuracy than RoBERTa-based classifiers and embedding-based nearest neighbor approaches. More notably, the NLI transfer enables our 10-shot model to perform competitively with 50-shot or even full-shot classifiers, while we can keep the inference time constant by leveraging a faster embedding retrieval model.

Shiyang Li, Semih Yavuz, Kazuma Hashimoto et al.

Dialogue state trackers have made significant progress on benchmark datasets, but their generalization capability to novel and realistic scenarios beyond the held-out conversations is less understood. We propose controllable counterfactuals (CoCo) to bridge this gap and evaluate dialogue state tracking (DST) models on novel scenarios, i.e., would the system successfully tackle the request if the user responded differently but still consistently with the dialogue flow? CoCo leverages turn-level belief states as counterfactual conditionals to produce novel conversation scenarios in two steps: (i) counterfactual goal generation at turn-level by dropping and adding slots followed by replacing slot values, (ii) counterfactual conversation generation that is conditioned on (i) and consistent with the dialogue flow. Evaluating state-of-the-art DST models on MultiWOZ dataset with CoCo-generated counterfactuals results in a significant performance drop of up to 30.8% (from 49.4% to 18.6%) in absolute joint goal accuracy. In comparison, widely used techniques like paraphrasing only affect the accuracy by at most 2%. Human evaluations show that COCO-generated conversations perfectly reflect the underlying user goal with more than 95% accuracy and are as human-like as the original conversations, further strengthening its reliability and promise to be adopted as part of the robustness evaluation of DST models.

Kazuma Hashimoto, Raffaella Buschiazzo, James Bradbury et al.

This paper presents a high-quality multilingual dataset for the documentation domain to advance research on localization of structured text. Unlike widely-used datasets for translation of plain text, we collect XML-structured parallel text segments from the online documentation for an enterprise software platform. These Web pages have been professionally translated from English into 16 languages and maintained by domain experts, and around 100,000 text segments are available for each language pair. We build and evaluate translation models for seven target languages from English, with several different copy mechanisms and an XML-constrained beam search. We also experiment with a non-English pair to show that our dataset has the potential to explicitly enable $17 \times 16$ translation settings. Our experiments show that learning to translate with the XML tags improves translation accuracy, and the beam search accurately generates XML structures. We also discuss trade-offs of using the copy mechanisms by focusing on translation of numerical words and named entities. We further provide a detailed human analysis of gaps between the model output and human translations for real-world applications, including suitability for post-editing.

Andre Esteva, Anuprit Kale, Romain Paulus et al.

The COVID-19 global pandemic has resulted in international efforts to understand, track, and mitigate the disease, yielding a significant corpus of COVID-19 and SARS-CoV-2-related publications across scientific disciplines. As of May 2020, 128,000 coronavirus-related publications have been collected through the COVID-19 Open Research Dataset Challenge. Here we present CO-Search, a retriever-ranker semantic search engine designed to handle complex queries over the COVID-19 literature, potentially aiding overburdened health workers in finding scientific answers during a time of crisis. The retriever is built from a Siamese-BERT encoder that is linearly composed with a TF-IDF vectorizer, and reciprocal-rank fused with a BM25 vectorizer. The ranker is composed of a multi-hop question-answering module, that together with a multi-paragraph abstractive summarizer adjust retriever scores. To account for the domain-specific and relatively limited dataset, we generate a bipartite graph of document paragraphs and citations, creating 1.3 million (citation title, paragraph) tuples for training the encoder. We evaluate our system on the data of the TREC-COVID information retrieval challenge. CO-Search obtains top performance on the datasets of the first and second rounds, across several key metrics: normalized discounted cumulative gain, precision, mean average precision, and binary preference.

Lichao Sun, Kazuma Hashimoto, Wenpeng Yin et al.

There is an increasing amount of literature that claims the brittleness of deep neural networks in dealing with adversarial examples that are created maliciously. It is unclear, however, how the models will perform in realistic scenarios where \textit{natural rather than malicious} adversarial instances often exist. This work systematically explores the robustness of BERT, the state-of-the-art Transformer-style model in NLP, in dealing with noisy data, particularly mistakes in typing the keyboard, that occur inadvertently. Intensive experiments on sentiment analysis and question answering benchmarks indicate that: (i) Typos in various words of a sentence do not influence equally. The typos in informative words make severer damages; (ii) Mistype is the most damaging factor, compared with inserting, deleting, etc.; (iii) Humans and machines have different focuses on recognizing adversarial attacks.

Akari Asai, Kazuma Hashimoto, Hannaneh Hajishirzi et al.

Answering questions that require multi-hop reasoning at web-scale necessitates retrieving multiple evidence documents, one of which often has little lexical or semantic relationship to the question. This paper introduces a new graph-based recurrent retrieval approach that learns to retrieve reasoning paths over the Wikipedia graph to answer multi-hop open-domain questions. Our retriever model trains a recurrent neural network that learns to sequentially retrieve evidence paragraphs in the reasoning path by conditioning on the previously retrieved documents. Our reader model ranks the reasoning paths and extracts the answer span included in the best reasoning path. Experimental results show state-of-the-art results in three open-domain QA datasets, showcasing the effectiveness and robustness of our method. Notably, our method achieves significant improvement in HotpotQA, outperforming the previous best model by more than 14 points.

Jian-Guo Zhang, Kazuma Hashimoto, Chien-Sheng Wu et al.

Dialog state tracking (DST) is a core component in task-oriented dialog systems. Existing approaches for DST mainly fall into one of two categories, namely, ontology-based and ontology-free methods. An ontology-based method selects a value from a candidate-value list for each target slot, while an ontology-free method extracts spans from dialog contexts. Recent work introduced a BERT-based model to strike a balance between the two methods by pre-defining categorical and non-categorical slots. However, it is not clear enough which slots are better handled by either of the two slot types, and the way to use the pre-trained model has not been well investigated. In this paper, we propose a simple yet effective dual-strategy model for DST, by adapting a single BERT-style reading comprehension model to jointly handle both the categorical and non-categorical slots. Our experiments on the MultiWOZ datasets show that our method significantly outperforms the BERT-based counterpart, finding that the key is a deep interaction between the domain-slot and context information. When evaluated on noisy (MultiWOZ 2.0) and cleaner (MultiWOZ 2.1) settings, our method performs competitively and robustly across the two different settings. Our method sets the new state of the art in the noisy setting, while performing more robustly than the best model in the cleaner setting. We also conduct a comprehensive error analysis on the dataset, including the effects of the dual strategy for each slot, to facilitate future research.

Akari Asai, Akiko Eriguchi, Kazuma Hashimoto et al.

Despite recent work in Reading Comprehension (RC), progress has been mostly limited to English due to the lack of large-scale datasets in other languages. In this work, we introduce the first RC system for languages without RC training data. Given a target language without RC training data and a pivot language with RC training data (e.g. English), our method leverages existing RC resources in the pivot language by combining a competitive RC model in the pivot language with an attentive Neural Machine Translation (NMT) model. We first translate the data from the target to the pivot language, and then obtain an answer using the RC model in the pivot language. Finally, we recover the corresponding answer in the original language using soft-alignment attention scores from the NMT model. We create evaluation sets of RC data in two non-English languages, namely Japanese and French, to evaluate our method. Experimental results on these datasets show that our method significantly outperforms a back-translation baseline of a state-of-the-art product-level machine translation system.

Kazuma Hashimoto, Yoshimasa Tsuruoka

A major obstacle in reinforcement learning-based sentence generation is the large action space whose size is equal to the vocabulary size of the target-side language. To improve the efficiency of reinforcement learning, we present a novel approach for reducing the action space based on dynamic vocabulary prediction. Our method first predicts a fixed-size small vocabulary for each input to generate its target sentence. The input-specific vocabularies are then used at supervised and reinforcement learning steps, and also at test time. In our experiments on six machine translation and two image captioning datasets, our method achieves faster reinforcement learning ($\sim$2.7x faster) with less GPU memory ($\sim$2.3x less) than the full-vocabulary counterpart. The reinforcement learning with our method consistently leads to significant improvement of BLEU scores, and the scores are equal to or better than those of baselines using the full vocabularies, with faster decoding time ($\sim$3x faster) on CPUs.

Kazuma Hashimoto, Yoshimasa Tsuruoka

This paper presents a novel neural machine translation model which jointly learns translation and source-side latent graph representations of sentences. Unlike existing pipelined approaches using syntactic parsers, our end-to-end model learns a latent graph parser as part of the encoder of an attention-based neural machine translation model, and thus the parser is optimized according to the translation objective. In experiments, we first show that our model compares favorably with state-of-the-art sequential and pipelined syntax-based NMT models. We also show that the performance of our model can be further improved by pre-training it with a small amount of treebank annotations. Our final ensemble model significantly outperforms the previous best models on the standard English-to-Japanese translation dataset.

Kazuma Hashimoto, Caiming Xiong, Yoshimasa Tsuruoka et al.

Transfer and multi-task learning have traditionally focused on either a single source-target pair or very few, similar tasks. Ideally, the linguistic levels of morphology, syntax and semantics would benefit each other by being trained in a single model. We introduce a joint many-task model together with a strategy for successively growing its depth to solve increasingly complex tasks. Higher layers include shortcut connections to lower-level task predictions to reflect linguistic hierarchies. We use a simple regularization term to allow for optimizing all model weights to improve one task's loss without exhibiting catastrophic interference of the other tasks. Our single end-to-end model obtains state-of-the-art or competitive results on five different tasks from tagging, parsing, relatedness, and entailment tasks.

Yusuke Watanabe, Kazuma Hashimoto, Yoshimasa Tsuruoka

We propose a simple domain adaptation method for neural networks in a supervised setting. Supervised domain adaptation is a way of improving the generalization performance on the target domain by using the source domain dataset, assuming that both of the datasets are labeled. Recently, recurrent neural networks have been shown to be successful on a variety of NLP tasks such as caption generation; however, the existing domain adaptation techniques are limited to (1) tune the model parameters by the target dataset after the training by the source dataset, or (2) design the network to have dual output, one for the source domain and the other for the target domain. Reformulating the idea of the domain adaptation technique proposed by Daume (2007), we propose a simple domain adaptation method, which can be applied to neural networks trained with a cross-entropy loss. On captioning datasets, we show performance improvements over other domain adaptation methods.

Akiko Eriguchi, Kazuma Hashimoto, Yoshimasa Tsuruoka

Most of the existing Neural Machine Translation (NMT) models focus on the conversion of sequential data and do not directly use syntactic information. We propose a novel end-to-end syntactic NMT model, extending a sequence-to-sequence model with the source-side phrase structure. Our model has an attention mechanism that enables the decoder to generate a translated word while softly aligning it with phrases as well as words of the source sentence. Experimental results on the WAT'15 English-to-Japanese dataset demonstrate that our proposed model considerably outperforms sequence-to-sequence attentional NMT models and compares favorably with the state-of-the-art tree-to-string SMT system.

Kazuma Hashimoto, Yoshimasa Tsuruoka

We present a novel method for jointly learning compositional and non-compositional phrase embeddings by adaptively weighting both types of embeddings using a compositionality scoring function. The scoring function is used to quantify the level of compositionality of each phrase, and the parameters of the function are jointly optimized with the objective for learning phrase embeddings. In experiments, we apply the adaptive joint learning method to the task of learning embeddings of transitive verb phrases, and show that the compositionality scores have strong correlation with human ratings for verb-object compositionality, substantially outperforming the previous state of the art. Moreover, our embeddings improve upon the previous best model on a transitive verb disambiguation task. We also show that a simple ensemble technique further improves the results for both tasks.

Kazuma Hashimoto, Pontus Stenetorp, Makoto Miwa et al.

We present a novel learning method for word embeddings designed for relation classification. Our word embeddings are trained by predicting words between noun pairs using lexical relation-specific features on a large unlabeled corpus. This allows us to explicitly incorporate relation-specific information into the word embeddings. The learned word embeddings are then used to construct feature vectors for a relation classification model. On a well-established semantic relation classification task, our method significantly outperforms a baseline based on a previously introduced word embedding method, and compares favorably to previous state-of-the-art models that use syntactic information or manually constructed external resources.