With the flourish of Web, online review is becoming a more and useful important information resource for people. As result, automatic mining summarization has become hot research topic recently. Different from traditional text summarization, aims at extracting features on which reviewers express their opinions determining whether are positive or negative. In this paper, we focus specific domain - movie review. A multi-knowledge based approach proposed, integrates WordNet, statistical...

We examine the problem of keyboard acoustic emanations. present a novel attack taking as input 10-minute sound recording user typing English text using and recovering up to 96% typed characters. There is no need for training recordings labeled with corresponding clear text. A recognizer bootstrapped from can even recognize random such passwords: In our experiments, 90% 5-character passwords only letters be generated in fewer than 20 attempts by an adversary; 80% 10-character 75 adversary....

Managing data and computation is at the heart of datacenter computing. Manual management can lead to loss, wasteful consumption storage, laborious bookkeeping. Lack proper result in lost opportunities share common computations across multiple jobs or compute results incrementally.Nectar a system designed address aforementioned problems. It automates unifies within datacenter. In Nectar, are treated interchangeably by associating with its computation. Derived datasets, which computations,...

Terry Yue Zhuo, Zhuang Li, Yujin Huang, Fatemeh Shiri, Weiqing Wang, Gholamreza Haffari, Yuan-Fang Li. Proceedings of the 17th Conference European Chapter Association for Computational Linguistics. 2023.

Anonymous routing protects user communication from identification by third-party observers. Existing anonymous layers utilize Chaum-Mixes for anonymity relaying traffic through relay nodes called mixes. The source defines a static forwarding path which is relayed to the destination. resulting fragile and shortlived: failure of one mix in breaks results data loss jitter before new constructed. In this paper, we propose Cashmere, resilient layer built on structured peer-to-peer overlay....

Standard state-machine replication involves consensus on a sequence of totally ordered requests through, for example, the Paxos protocol. Such sequential execution model is becoming outdated prevalent multi-core servers. Highly concurrent executions architectures introduce non-determinism related to thread scheduling and lock contentions, fundamentally break assumption in replication. This tension between concurrency consistency not inherent because total-ordering merely simplifying...

This paper proposes an image encryption scheme based on a discrete-time alternating quantum walk (AQW) and the advanced standard (AES). We use properties to improve AES algorithm, which uses keystream generator related AQW parameters generate probability distribution matrix. Some singular values of matrix are extracted as key algorithm. The Rcon algorithm is replaced with elements Then, ascending order size clone scrambles mapping rules S-box ShiftRow transformations in plaintext XOR...

Alignment tuning is crucial for ensuring large language models (LLMs) behave ethically and helpfully. Current alignment approaches require high-quality annotations significant training resources. This paper proposes a low-cost, tuning-free method using in-context learning (ICL) to enhance LLM alignment. Through an analysis of ICL demos, we identified style as key factor influencing capabilities explicitly restyled exemplars based on this stylistic framework. Additionally, combined the demos...

GSM network is the most worldwide mobile communication nowadays. Based on SIEMENS MC35 module, general techniques of with are depicted, including initialization terminal equipment, sending and reading short messages (SMS), SMS to group users, management phonebook SIM card, furthermore, a flexible solution real-time proposed. Finally, application cases given for module.

Despite significant advancements in multi-label text classification, the ability of existing models to generalize novel and seldom-encountered complex concepts, which are compositions elementary ones, remains underexplored. This research addresses this gap. By creating unique data splits across three benchmarks, we assess compositional generalization classification models. Our results show that these often fail concepts encountered infrequently during training, leading inferior performance...

Current video text spotting methods can achieve preferable performance, powered with sufficient labeled training data. However, labeling data manually is time-consuming and labor-intensive. To overcome this, using low-cost synthetic a promising alternative. This paper introduces novel synthesis technique called FlowText, which utilizes optical flow estimation to synthesize large amount of at low cost for robust spotters. Unlike existing that focus on image-level synthesis, FlowText...

This work describes an effective spelling check approach for Chinese OCR with a new multi-knowledge based statistical language model. model combines the conventional n-gram and LSA (latent semantic analysis) model, so both local information (syntax) global (semantic) are utilized. Furthermore, similar characters used in Viterbi search process to expand candidate list order add more possible correct results. With our approach, best recognition accuracy rate increases from 79.3% 91.9%, which...

There has been an increasing interest in incorporating Artificial Intelligence (AI) into Defence and military systems to complement augment human intelligence capabilities. However, much work still needs be done toward achieving effective human-machine partnership. This is aimed at enhancing communications by developing a capability for automatically translating natural language machine-understandable (e.g., SQL queries). Techniques this goal typically involve building semantic parser...

Semantic parsing is the task of translating natural language utterances into machine-readable meaning representations. Currently, most semantic methods are not able to utilize contextual information (e.g. dialogue and comments history), which has a great potential boost systems. To address this issue, context dependent recently drawn lot attention. In survey, we investigate progress on for parsing, together with current datasets tasks. We then point out open problems challenges future...

Semantic parsing maps natural language (NL) utterances into logical forms (LFs), which underpins many advanced NLP problems. parsers gain performance boosts with deep neural networks, but inherit vulnerabilities against adversarial examples. In this paper, we provide the first empirical study on robustness of semantic in presence attacks. Formally, adversaries are considered to be perturbed utterance-LF pairs, whose have exactly same meanings as original ones. A scalable methodology is...

In this work, we investigate the problems of semantic parsing in a few-shot learning setting. setting, are provided with k utterance-logical form pairs per new predicate. The state-of-the-art neural parsers achieve less than 25% accuracy on benchmark datasets when = 1. To tackle problem, proposed to i) apply designated meta-learning method train model; ii) regularize attention scores alignment statistics; iii) smoothing technique pretraining. As result, our consistently outperforms all...

This paper investigates continual learning for semantic parsing. In this setting, a neural parser learns tasks sequentially without accessing full training data from previous tasks. Direct application of the SOTA algorithms to problem fails achieve comparable performance with re-training models all seen because they have not considered special properties structured outputs yielded by parsers. Therefore, we propose TotalRecall, method designed parsers two aspects: i) sampling memory replay...

Multilingual semantic parsing aims to leverage the knowledge from high-resource languages improve low-resource parsing, yet commonly suffers data imbalance problem. Prior works propose utilize translations by either humans or machines alleviate such issues. However, human are expensive, while machine cheap but prone error and bias. In this work, we an active learning approach that exploits strengths of both iteratively adding small batches into machine-translated training set. Besides, novel...

Image segmentation based on continual learning exhibits a critical drop of performance, mainly due to catastrophic forgetting and background shift, as they are required incorporate new classes continually. In this paper, we propose simple, yet effective Continual Segmentation method with incremental Dynamic Query (CISDQ), which decouples the representation both old knowledge lightweight query embedding. CISDQ includes three contributions: 1) We define <italic...