Effective information retrieval (IR) from vast datasets relies on advanced techniques to extract relevant information in response to queries.Recent advancements in dense retrieval have showcased remarkable efficacy compared to traditional sparse retrieval methods.To further enhance retrieval performance, knowledge distillation techniques, often leveraging robust crossencoder rerankers, have been extensively explored.However, existing approaches primarily distill knowledge from pointwise rerankers, which assign absolute relevance scores to documents, thus facing challenges related to inconsistent comparisons.This paper introduces Pairwise Relevance Distillation (PAIRDISTILL) to leverage pairwise reranking, offering finegrained distinctions between similarly relevant documents to enrich the training of dense retrieval models.Our experiments demonstrate that PAIRDISTILL outperforms existing methods, achieving new state-of-the-art results across multiple benchmarks.This highlights the potential of PAIRDISTILL in advancing dense retrieval techniques effectively. 1

This paper introduces the Ninth Dialog System Technology Challenge (DSTC-9). This edition of the DSTC focuses on applying end-to-end dialog technologies for four distinct tasks in dialog systems, namely, 1. Task-oriented dialog Modeling with Unstructured Knowledge Access, 2. Multi-domain task-oriented dialog, 3. Interactive evaluation of dialog and 4. Situated interactive multimodal dialog. This paper describes the task definition, provided datasets, baselines, and evaluation setup for each track. We also summarize the results of the submitted systems to highlight the general trends of the state-of-the-art technologies for the tasks.

Conversational search provides a natural interface for information retrieval (IR).Recent approaches have demonstrated promising results in applying dense retrieval to conversational IR.However, training dense retrievers requires large amounts of in-domain paired data.This hinders the development of conversational dense retrievers, as abundant in-domain conversations are expensive to collect.In this paper, we propose CONVERSER, a framework for training conversational dense retrievers with at most 6 examples of in-domain dialogues.Specifically, we utilize the in-context learning capability of large language models to generate conversational queries given a passage in the retrieval corpus.Experimental results on conversational retrieval benchmarks OR-QuAC and TREC CAsT 19 show that the proposed CON-VERSER achieves comparable performance to fully-supervised models, demonstrating the effectiveness of our proposed framework in fewshot conversational dense retrieval.1

Prior work has demonstrated that data augmentation is useful for improving dialogue state tracking.However, there are many types of user utterances, while the prior method only considered the simplest one for augmentation, raising the concern about poor generalization capability.In order to better cover diverse dialogue acts and control the generation quality, this paper proposes controllable user dialogue act augmentation (CUDA-DST) to augment user utterances with diverse behaviors.With the augmented data, different state trackers gain improvement and show better robustness, achieving the state-of-the-art performance on MultiWOZ 2.1.1

Given the clinical notes written in electronic health records (EHRs), it is challenging to predict the diagnostic codes which is formulated as a multi-label classification task.The large set of labels, the hierarchical dependency, and the imbalanced data make this prediction task extremely hard.Most existing work built a binary prediction for each label independently, ignoring the dependencies between labels.To address this problem, we propose a two-stage framework to improve automatic ICD coding by capturing the label correlation.Specifically, we train a label set distribution estimator to rescore the probability of each label set candidate generated by a base predictor.This paper is the first attempt at learning the label set distribution as a reranking module for medical code prediction.In the experiments, our proposed framework is able to improve upon best-performing predictors on the benchmark MIMIC datasets.1

Pre-trained language models have achieved huge improvement on many NLP tasks.However, these methods are usually designed for written text, so they do not consider the properties of spoken language.Therefore, this paper aims at generalizing the idea of language model pre-training to lattices generated by recognition systems.We propose a framework that trains neural lattice language models to provide contextualized representations for spoken language understanding tasks.The proposed two-stage pre-training approach reduces the demands of speech data and has better efficiency.Experiments on intent detection and dialogue act recognition datasets demonstrate that our proposed method consistently outperforms strong baselines when evaluated on spoken inputs. 1

In modular dialogue systems, natural language understanding (NLU) and natural language generation (NLG) are two critical components, where NLU extracts the semantics from the given texts and NLG is to construct corresponding natural language sentences based on the input semantic representations.However, the dual property between understanding and generation has been rarely explored.The prior work (Su et al., 2019) is the first attempt that utilized the duality between NLU and NLG to improve the performance via a dual supervised learning framework.However, the prior work still learned both components in a supervised manner; instead, this paper introduces a general learning framework to effectively exploit such duality, providing flexibility of incorporating both supervised and unsupervised learning algorithms to train language understanding and generation models in a joint fashion.The benchmark experiments demonstrate that the proposed approach is capable of boosting the performance of both NLU and NLG. 1

Employing pre-trained language models (LM) to extract contextualized word representations has achieved state-of-the-art performance on various NLP tasks. However, applying this technique to noisy transcripts generated by automatic speech recognizer (ASR) is concerned. Therefore, this paper focuses on making contextualized representations more ASR-robust. We propose a novel confusion-aware fine-tuning method to mitigate the impact of ASR errors on pre-trained LMs. Specifically, we fine-tune LMs to produce similar representations for acoustically confusable words that are obtained from word confusion networks (WCNs) produced by ASR. Experiments on multiple benchmark datasets show that the proposed method significantly improves the performance of spoken language understanding when performing on ASR transcripts.

Natural language understanding (NLU) and natural language generation (NLG) are both critical research topics in the NLP and dialogue fields.Natural language understanding is to extract the core semantic meaning from the given utterances, while natural language generation is opposite, of which the goal is to construct corresponding sentences based on the given semantics.However, such dual relationship has not been investigated in literature.This paper proposes a novel learning framework for natural language understanding and generation on top of dual supervised learning, providing a way to exploit the duality.The preliminary experiments show that the proposed approach boosts the performance for both tasks, demonstrating the effectiveness of the dual relationship.1

Lattices are compact representations that encode multiple hypotheses, such as speech recognition results or different word segmentations. It is shown that encoding lattices as opposed to 1-best results generated by automatic speech recognizer (ASR) boosts the performance of spoken language understanding (SLU). Recently, pre-trained language models with the transformer architecture have achieved the state-of-the-art results on natural language understanding, but their ability of encoding lattices has not been explored. Therefore, this paper aims at adapting pre-trained transformers to lattice inputs in order to perform understanding tasks specifically for spoken language. Our experiments on the benchmark ATIS dataset show that fine-tuning pre-trained transformers with lattice inputs yields clear improvement over fine-tuning with 1-best results. Further evaluation demonstrates the effectiveness of our methods under different acoustic conditions11The code is available at https://github.com/MiuLab/Lattice-SLU.

Compared with current distributed networks, Software defined networking is an innovative technology to manage networks by a centralized controller with global visibility. OpenFlow is one kind of SDN protocols. It provides flow-level controllability of network traffic to simplify the management of the administrators of data center. Flow control rules used by OpenFlow in switch flow table are called flow entries. Modern commodity switches use TCAM to implement flow table because of its high matching speed. However, the general TCAM size of switch only accommodates up to thousands of flow entries due to its high cost-to-density ratio and high power consumption. It indicates that the maximal number of flow that can be controlled is restricted to the TCAM size of switch. That leads to several scalability issues, especially in a data center. In this paper, we propose a topology based flow entry management (TFEM) technique, which is a scalable and memory-efficient solution by exploiting the hierarchical topology characteristic to aggregate the uses of flow entries for most of the small flows in a data center. With the proposed downlink and uplink separated flow tables structure, our results show that TFEM can reduce the average number of flow entries by 50%~78% for fat-tree topology against the traditional SDN without any switch hardware modification.

This paper examines the use of two-way training in multiple-input multiple-output (MIMO) wireless systems to discriminate the channel estimation (and, thus, data detection) performance between two receivers, namely, a legitimate receiver (LR) and an unauthorized receiver (UR). This work extends upon the discriminatory channel estimation (DCE) proposed in our prior work, where it was previously assumed that training signals can only be sent by the transmitter. The DCE design criterion is to minimize the channel estimation error at the LR while confining the channel estimation error at the UR above a minimum level. In the case of two-way training, training signals can first be transmitted on the reverse link to enable channel estimation at the transmitter and allow the transmitter to insert artificial noise (AN) along with the training signal in the forward link to disrupt the training at the UR, while minimizing the interference on the LR. The optimal power allocation between training and AN signals is devised for systems that are subject to both average and peak power constraints. Numerical results demonstrate the efficacy of the proposed two-way training scheme when used in discriminating the performances between LR and UR.

This paper presents a DSP core for multi-core media SoC, which is optimized to execute a set of signal processing tasks very efficiently. The fully-programmable core has a data-centric instruction set and a corresponding latency-insensitive micro-architecture, where the hardware design is optimized concurrently with its automatic software generator. The proposed DSP core has 3/spl times/ performance (in cycles) of those found in commercial dual-core application processors with similar computing resources. The silicon implementation in UMC 0.18/spl mu/m 1P6M CMOS technology operates at 314MHz and consumes only 52mW average power.