arXiv subjects

Tianyu Liu

Publications and source records attributed to Tianyu Liu.

At least 199 records · Page 11Linked to original sources

arXiv2022

SmartSales: Sales Script Extraction and Analysis from Sales Chatlog

In modern sales applications, automatic script extraction and management greatly decrease the need for human labor to collect the winning sales scripts, which largely boost the success rate for sales and can be shared across the sales teams. In this work, we present the SmartSales system to serve both the sales representatives and managers to attain the sales insights from the large-scale sales chatlog. SmartSales consists of three modules: 1) Customer frequently asked questions (FAQ) extraction aims to enrich the FAQ knowledge base by harvesting high quality customer question-answer pairs from the chatlog. 2) Customer objection response assists the salespeople to figure out the typical customer objections and corresponding winning sales scripts, as well as search for proper sales responses for a certain customer objection. 3) Sales manager dashboard helps sales managers to monitor whether a specific sales representative or team follows the sales standard operating procedures (SOP). The proposed prototype system is empowered by the state-of-the-art conversational intelligence techniques and has been running on the Tencent Cloud to serve the sales teams from several different areas.

cs.CL↗

arXiv2022

Star-Convex Constrained Optimization for Visibility Planning with Application to Aerial Inspection

The visible capability is critical in many robot applications, such as inspection and surveillance, etc. Without the assurance of the visibility to targets, some tasks end up not being complete or even failing. In this paper, we propose a visibility guaranteed planner by star-convex constrained optimization. The visible space is modeled as star convex polytope (SCP) by nature and is generated by finding the visible points directly on point cloud. By exploiting the properties of the SCP, the visibility constraint is formulated for trajectory optimization. The trajectory is confined in the safe and visible flight corridor which consists of convex polytopes and SCPs. We further make a relaxation to the visibility constraints and transform the constrained trajectory optimization problem into an unconstrained one that can be reliably and efficiently solved. To validate the capability of the proposed planner, we present the practical application in site inspection. The experimental results show that the method is efficient, scalable, and visibility guaranteed, presenting the prospect of application to various other applications in the future.

cs.RO↗

arXiv2022

A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation

Large pretrained generative models like GPT-3 often suffer from hallucinating non-existent or incorrect content, which undermines their potential merits in real applications. Existing work usually attempts to detect these hallucinations based on a corresponding oracle reference at a sentence or document level. However ground-truth references may not be readily available for many free-form text generation applications, and sentence- or document-level detection may fail to provide the fine-grained signals that would prevent fallacious content in real time. As a first step to addressing these issues, we propose a novel token-level, reference-free hallucination detection task and an associated annotated dataset named HaDes (HAllucination DEtection dataSet). To create this dataset, we first perturb a large number of text segments extracted from English language Wikipedia, and then verify these with crowd-sourced annotations. To mitigate label imbalance during annotation, we utilize an iterative model-in-loop strategy. We conduct comprehensive data analyses and create multiple baseline models.

cs.CL↗

arXiv2022

Cerberus Transformer: Joint Semantic, Affordance and Attribute Parsing

Multi-task indoor scene understanding is widely considered as an intriguing formulation, as the affinity of different tasks may lead to improved performance. In this paper, we tackle the new problem of joint semantic, affordance and attribute parsing. However, successfully resolving it requires a model to capture long-range dependency, learn from weakly aligned data and properly balance sub-tasks during training. To this end, we propose an attention-based architecture named Cerberus and a tailored training framework. Our method effectively addresses the aforementioned challenges and achieves state-of-the-art performance on all three tasks. Moreover, an in-depth analysis shows concept affinity consistent with human cognition, which inspires us to explore the possibility of weakly supervised learning. Surprisingly, Cerberus achieves strong results using only 0.1%-1% annotation. Visualizations further confirm that this success is credited to common attention maps across tasks. Code and models can be accessed at https://github.com/OPEN-AIR-SUN/Cerberus.

cs.CV↗

arXiv2022

Length L-function for Network-Constrained Point Data

Network constrained points are referred to as points restricted to road networks, such as taxi pick up and drop off locations. A significant pattern of network constrained points is referred to as an aggregation; e.g., the aggregation of pick up points may indicate a high taxi demand in a particular area. Although the network K function using the shortest path network distance has been proposed to detect point aggregation, its statistical unit is still radius based. R neighborhood, in particular, has inconsistent network length owing to the complex configuration of road networks which cause unfair counts and identification errors in networks (e.g., the length of the r neighborhood located at an intersection is longer than that on straight roads, which may include more points). In this study, we derived the length L function for network constrained points to identify the aggregation by designing a novel neighborhood as the statistical unit; the total length of this is consistent throughout the network. Compared to the network K function, our method can detect a true to life aggregation scale, identify the aggregation with higher network density, as well as identify the aggregations that the network K function cannot. We validated our method using taxi trips pick up location data within Zhongguancun Area in Beijing, analyzing differences in maximal aggregation between workdays and weekends to understand taxi demand in the morning and evening peak.

stat.OT↗

arXiv2022

Premise-based Multimodal Reasoning: Conditional Inference on Joint Textual and Visual Clues

It is a common practice for recent works in vision language cross-modal reasoning to adopt a binary or multi-choice classification formulation taking as input a set of source image(s) and textual query. In this work, we take a sober look at such an unconditional formulation in the sense that no prior knowledge is specified with respect to the source image(s). Inspired by the designs of both visual commonsense reasoning and natural language inference tasks, we propose a new task termed Premise-based Multi-modal Reasoning(PMR) where a textual premise is the background presumption on each source image. The PMR dataset contains 15,360 manually annotated samples which are created by a multi-phase crowd-sourcing process. With selected high-quality movie screenshots and human-curated premise templates from 6 pre-defined categories, we ask crowd-source workers to write one true hypothesis and three distractors (4 choices) given the premise and image through a cross-check procedure. Besides, we generate adversarial samples to alleviate the annotation artifacts and double the size of PMR. We benchmark various state-of-the-art (pretrained) multi-modal inference models on PMR and conduct comprehensive experimental analyses to showcase the utility of our dataset.

cs.CL↗

arXiv2022

Demystify Optimization and Generalization of Over-parameterized PAC-Bayesian Learning

PAC-Bayesian is an analysis framework where the training error can be expressed as the weighted average of the hypotheses in the posterior distribution whilst incorporating the prior knowledge. In addition to being a pure generalization bound analysis tool, PAC-Bayesian bound can also be incorporated into an objective function to train a probabilistic neural network, making them a powerful and relevant framework that can numerically provide a tight generalization bound for supervised learning. For simplicity, we call probabilistic neural network learned using training objectives derived from PAC-Bayesian bounds as {\it PAC-Bayesian learning}. Despite their empirical success, the theoretical analysis of PAC-Bayesian learning for neural networks is rarely explored. This paper proposes a new class of convergence and generalization analysis for PAC-Bayes learning when it is used to train the over-parameterized neural networks by the gradient descent method. For a wide probabilistic neural network, we show that when PAC-Bayes learning is applied, the convergence result corresponds to solving a kernel ridge regression when the probabilistic neural tangent kernel (PNTK) is used as its kernel. Based on this finding, we further characterize the uniform PAC-Bayesian generalization bound which improves over the Rademacher complexity-based bound for non-probabilistic neural network. Finally, drawing the insight from our theoretical results, we propose a proxy measure for efficient hyperparameters selection, which is proven to be time-saving.

cs.LG↗

arXiv2021

An experimental approach to mapping chemical bonds in nanostructured materials

We introduce a number of techniques in quantitative convergent-beam electron diffraction under development by our group and discuss the basis for measuring interatomic electrostatic potentials (and therefore also electron densities), localised at sub-nanometre scales, with sufficient accuracy and precision to map chemical bonds in and around nanostructures in nanostructured materials. This has never been possible as experimental measurements of bonding have always been restricted to homogeneous single-phased crystals.

cond-mat.mtrl-sci↗

arXiv2021

Semi-supervised Implicit Scene Completion from Sparse LiDAR

Recent advances show that semi-supervised implicit representation learning can be achieved through physical constraints like Eikonal equations. However, this scheme has not yet been successfully used for LiDAR point cloud data, due to its spatially varying sparsity. In this paper, we develop a novel formulation that conditions the semi-supervised implicit function on localized shape embeddings. It exploits the strong representation learning power of sparse convolutional networks to generate shape-aware dense feature volumes, while still allows semi-supervised signed distance function learning without knowing its exact values at free space. With extensive quantitative and qualitative results, we demonstrate intrinsic properties of this new learning system and its usefulness in real-world road scenes. Notably, we improve IoU from 26.3% to 51.0% on SemanticKITTI. Moreover, we explore two paradigms to integrate semantic label predictions, achieving implicit semantic completion. Code and models can be accessed at https://github.com/OPEN-AIR-SUN/SISC.

cs.CV↗

arXiv2021

Explicit Interaction Network for Aspect Sentiment Triplet Extraction

Aspect Sentiment Triplet Extraction (ASTE) aims to recognize targets, their sentiment polarities and opinions explaining the sentiment from a sentence. ASTE could be naturally divided into 3 atom subtasks, namely target detection, opinion detection and sentiment classification. We argue that the proper subtask combination, compositional feature extraction for target-opinion pairs, and interaction between subtasks would be the key to success. Prior work, however, may fail on `one-to-many' or `many-to-one' situations or derive non-existent sentiment triplets due to defective subtask formulation, sub-optimal feature representation or the lack of subtask interaction. In this paper, we divide ASTE into target-opinion joint detection and sentiment classification subtasks, which is in line with human cognition, and correspondingly utilize sequence encoder and table encoder to handle them. Table encoder extracts sentiment at token-pair level, so that the compositional feature between targets and opinions can be easily captured. To establish explicit interaction between subtasks, we utilize the table representation to guide the sequence encoding, and inject the sequence features back into the table encoder. Experiments show that our model outperforms state-of-the-art methods on six popular ASTE datasets.

cs.CL↗

arXiv2021

First Target and Opinion then Polarity: Enhancing Target-opinion Correlation for Aspect Sentiment Triplet Extraction

Aspect Sentiment Triplet Extraction (ASTE) aims to extract triplets from a sentence, including target entities, associated sentiment polarities, and opinion spans which rationalize the polarities. Existing methods are short on building correlation between target-opinion pairs, and neglect the mutual interference among different sentiment triplets. To address these issues, we utilize a two-stage framework to enhance the correlation between targets and opinions: at stage one, we extract targets and opinions through sequence tagging; then we append a group of artificial tags named Perceivable Pair, which indicate the span of a specific target-opinion tuple, to the input sentence to obtain closer correlated target-opinion pair representation. Meanwhile, we reduce the negative interference between triplets by restricting tokens' attention field. Finally, the polarity is identified according to the representation of the Perceivable Pair. We conduct experiments on four datasets, and the experimental results show the effectiveness of our model.

cs.CL↗

arXiv2021

A novel similarity measure for mining missing links in long-path networks

Network information mining is the study of the network topology, which answers a large number of application-based questions towards the structural evolution and the function of a real system. For example, the questions can be related to how the real system evolves or how individuals interact with each other in social networks. Although the evolution of the real system may seem to be found regularly, capturing patterns on the whole process of the evolution is not trivial. Link prediction is one of the most important technologies in network information mining, which can help us understand the real system's evolution law. Link prediction aims to uncover missing links or quantify the likelihood of the emergence of nonexistent links from known network structures. Currently, widely existing methods of link prediction almost focus on short-path networks that usually have a myriad of close triangular structures. However, these algorithms on highly sparse or long-path networks have poor performance. Here, we propose a new index that is associated with the principles of Structural Equivalence and Shortest Path Length ($SESPL$) to estimate the likelihood of link existence in long-path networks. Through 548 real networks test, we find that $SESPL$ is more effective and efficient than other similarity-based predictors in long-path networks. We also exploit the performance of $SESPL$ predictor and embedding-based approaches via machine learning techniques, and the performance of $SESPL$ can achieve a gain of 44.09\% over $GraphWave$ and 7.93\% over $Node2vec$. Finally, according to the matrix of Maximal Information Coefficient ($MIC$) between all the similarity-based predictors, $SESPL$ is a new independent feature to the space of traditional similarity features.

physics.soc-ph↗

arXiv2021

Behind the Scenes: An Exploration of Trigger Biases Problem in Few-Shot Event Classification

Few-Shot Event Classification (FSEC) aims at developing a model for event prediction, which can generalize to new event types with a limited number of annotated data. Existing FSEC studies have achieved high accuracy on different benchmarks. However, we find they suffer from trigger biases that signify the statistical homogeneity between some trigger words and target event types, which we summarize as trigger overlapping and trigger separability. The biases can result in context-bypassing problem, i.e., correct classifications can be gained by looking at only the trigger words while ignoring the entire context. Therefore, existing models can be weak in generalizing to unseen data in real scenarios. To further uncover the trigger biases and assess the generalization ability of the models, we propose two new sampling methods, Trigger-Uniform Sampling (TUS) and COnfusion Sampling (COS), for the meta tasks construction during evaluation. Besides, to cope with the context-bypassing problem in FSEC models, we introduce adversarial training and trigger reconstruction techniques. Experiments show these techniques help not only improve the performance, but also enhance the generalization ability of models.

cs.CL↗

arXiv2021

Identifying Aggregation Artery Architecture of constrained Origin-Destination flows using Manhattan L-function

The movement of humans and goods in cities can be represented by constrained flow, which is defined as the movement of objects between origin and destination in road networks. Flow aggregation, namely origins and destinations aggregated simultaneously, is one of the most common patterns, say the aggregated origin-to-destination flows between two transport hubs may indicate the great traffic demand between two sites. Developing a clustering method for constrained flows is crucial for determining urban flow aggregation. Among existing methods about identifying flow aggregation, L-function of flows is the major one. Nevertheless, this method depends on the aggregation scale, the key parameter detected by Euclidean L-function, it does not adapt to road network. The extracted aggregation may be overestimated and dispersed. Therefore, we propose a clustering method based on L-function of Manhattan space, which consists of three major steps. The first is to detect aggregation scales by Manhattan L-function. The second is to determine core flows possessing highest local L-function values at different scales. The final step is to take the intersection of core flows neighbourhoods, the extent of which depends on corresponding scale. By setting the number of core flows, we could concentrate the aggregation and thus highlight Aggregation Artery Architecture (AAA), which depicts road sections that contain the projection of key flow cluster on the road networks. Experiment using taxi flows showed that AAA could clarify resident movement type of identified aggregated flows. Our method also helps selecting locations for distribution sites, thereby supporting accurate analysis of urban interactions.

cs.CG↗

arXiv2021

Pseudo Landau levels, negative strain resistivity, and enhanced thermopower in twisted graphene nanoribbons

As a canonical response to the applied magnetic field, the electronic states of a metal are fundamentally reorganized into Landau levels. In Dirac metals, Landau levels can be expected without magnetic fields, provided that an inhomogeneous strain is applied to spatially modulate electron hoppings in a similar way as the Aharonov-Bohm phase. We here predict that a twisted zigzag nanoribbon of graphene exhibits strain-induced pseudo Landau levels of unexplored but analytically solvable dispersions at low energies. The presence of such dispersive pseudo Landau levels results in a negative strain resistivity characterizing the $(1+1)$-dimensional chiral anomaly if partially filled and can greatly enhance the thermopower when fully filled.

cond-mat.mes-hall↗

arXiv2021

Field-Tunable One-Sided Higher-Order Topological Hinge States in Dirac Semimetals

Recently, higher-order topological matter and 3D quantum Hall effects have attracted great attention. The Fermi-arc mechanism of the 3D quantum Hall effect proposed in Weyl semimetals is characterized by the one-sided hinge states, which do not exist in all the previous quantum Hall systems and more importantly pose a realistic example of the higher-order topological matter. The experimental effort so far is in the Dirac semimetal Cd$_3$As$_2$, where however time-reversal symmetry leads to hinge states on both sides of the top/bottom surfaces, instead of the aspired one-sided hinge states. We propose that under a tilted magnetic field, the hinge states in Cd$_3$As$_2$-like Dirac semimetals can be one-sided, highly tunable by field direction and Fermi energy, and robust against weak disorder. Furthermore, we propose a scanning tunneling Hall measurement to detect the one-sided hinge states. Our results will be insightful for exploring not only the quantum Hall effects beyond two dimensions, but also other higher-order topological insulators in the future.

cond-mat.mes-hall↗

arXiv2021

Quantum Hall effect originated from helical edge states in Cd$_3$As$_2$

The recent experimental observations of the quantum Hall effect in 3D topological semimetals have attracted great attention, but there are still debates on its origin. We systematically study the dependence of the quantum Hall effect in topological semimetals on the thickness, Fermi energy, and growth direction, taking into account the contributions from the Fermi-arc surface states, confinement-induced bulk subbands, and helical side-surface edge states. In particular, we focus on the intensively studied Dirac semimetal Cd$_{3}$As$_{2}$ and its slabs grown along experimentally accessible directions, including [001], [110], and [112]. We reveal an ignored mechanism from the Zeeman splitting of the helical edge states, which along with Fermi-arc 3D quantum Hall effect, may give a non-monotonic dependence of the Hall conductance plateaus on the magnetic field in the most experimentally studied [112] direction slab. Our results will be insightful for exploring the quantum Hall effects beyond two dimensions.

cond-mat.mes-hall↗

arXiv2021

Application of the unified control and detection framework to detecting stealthy integrity cyber-attacks on feedback control systems

This draft addresses issues of detecting stealthy integrity cyber-attacks on automatic control systems in the unified control and detection framework. A general form of integrity cyber-attacks that cannot be detected using the well-established observer-based technique is first introduced as kernel attacks. The well-known replay, zero dynamics and covert attacks are special forms of the kernel attacks. Existence conditions for the kernel attacks are presented. It is demonstrated, in the unified framework of control and detection, that all kernel attacks can be structurally detected when not only the observer-based residual, but also the control signal based residual signals are generated and used for the detection purpose. Based on the analytical results, two schemes for detecting the kernel attacks are then proposed, which allow reliable attack detection without loss of control performance. While the first scheme is similar to the well-established moving target method and auxiliary system aided detection scheme, the second detector is realised with encrypted transmissions of control and monitoring signals in the feedback control system that prevents adversary to gain system knowledge by means of eavesdropping attacks. Both schemes are illustrated by examples of detecting replay, zero dynamics and covert attacks and an experimental study on a three-tank control system.

eess.SY↗