Xiaohui Yu

Kaiwen Chen, Yueting Chen, Xiaohui Yu et al.

Large language models (LLMs) have revolutionized natural language interfaces for databases, particularly in text-to-SQL conversion. However, current approaches often generate unreliable outputs when faced with ambiguity or insufficient context. We present Reliable Text-to-SQL (RTS), a novel framework that enhances query generation reliability by incorporating abstention and human-in-the-loop mechanisms. RTS focuses on the critical schema linking phase, which aims to identify the key database elements needed for generating SQL queries. It autonomously detects potential errors during the answer generation process and responds by either abstaining or engaging in user interaction. A vital component of RTS is the Branching Point Prediction (BPP) which utilizes statistical conformal techniques on the hidden layers of the LLM model for schema linking, providing probabilistic guarantees on schema linking accuracy. We validate our approach through comprehensive experiments on the BIRD benchmark, demonstrating significant improvements in robustness and reliability. Our findings highlight the potential of combining transparent-box LLMs with human-in-the-loop processes to create more robust natural language interfaces for databases. For the BIRD benchmark, our approach achieves near-perfect schema linking accuracy, autonomously involving a human when needed. Combined with query generation, we demonstrate that near-perfect schema linking and a small query generation model can almost match SOTA accuracy achieved with a model orders of magnitude larger than the one we use.

Zhaoyue Chen, Nick Koudas, Zhe Zhang et al.

Machine Learning (ML) applications are proliferating in the enterprise. Relational data which are prevalent in enterprise applications are typically normalized; as a result, data has to be denormalized via primary/foreign-key joins to be provided as input to ML algorithms. In this paper, we study the implementation of popular nonlinear ML models, Gaussian Mixture Models (GMM) and Neural Networks (NN), over normalized data addressing both cases of binary and multi-way joins over normalized relations. For the case of GMM, we show how it is possible to decompose computation in a systematic way both for binary joins and for multi-way joins to construct mixture models. We demonstrate that by factoring the computation, one can conduct the training of the models much faster compared to other applicable approaches, without any loss in accuracy. For the case of NN, we propose algorithms to train the network taking normalized data as the input. Similarly, we present algorithms that can conduct the training of the network in a factorized way and offer performance advantages. The redundancy introduced by denormalization can be exploited for certain types of activation functions. However, we demonstrate that attempting to explore this redundancy is helpful up to a certain point; exploring redundancy at higher layers of the network will always result in increased costs and is not recommended. We present the results of a thorough experimental evaluation, varying several parameters of the input relations involved and demonstrate that our proposals for the training of GMM and NN yield drastic performance improvements typically starting at 100%, which become increasingly higher as parameters of the underlying data vary, without any loss in accuracy.

Meng Chen, Yang Liu, Xiaohui Yu

In this paper, we solve the problem of predicting the next locations of the moving objects with a historical dataset of trajectories. We present a Next Location Predictor with Markov Modeling (NLPMM) which has the following advantages: (1) it considers both individual and collective movement patterns in making prediction, (2) it is effective even when the trajectory data is sparse, (3) it considers the time factor and builds models that are suited to different time periods. We have conducted extensive experiments in a real dataset, and the results demonstrate the superiority of NLPMM over existing methods.

Meng Chen, Xiaohui Yu, Yang Liu

Traffic problems have seriously affected people's life quality and urban development, and forecasting the short-term traffic congestion is of great importance to both individuals and governments. However, understanding and modeling the traffic conditions can be extremely difficult, and our observations from real traffic data reveal that (1) similar traffic congestion patterns exist in the neighboring time slots and on consecutive workdays; (2) the levels of traffic congestion have clear multiscale properties. To capture these characteristics, we propose a novel method named PCNN based on deep Convolutional Neural Network, modeling Periodic traffic data for short-term traffic congestion prediction. PCNN has two pivotal procedures: time series folding and multi-grained learning. It first temporally folds the time series and constructs a two-dimensional matrix as the network input, such that both the real-time traffic conditions and past traffic patterns are well considered; then with a series of convolutions over the input matrix, it is able to model the local temporal dependency and multiscale traffic patterns. In particular, the global trend of congestion can be addressed at the macroscale; whereas more details and variations of the congestion can be captured at the microscale. Experimental results on a real-world urban traffic dataset confirm that folding time series data into a two-dimensional matrix is effective and PCNN outperforms the baselines significantly for the task of short-term congestion prediction.

Meng Chen, Xiaoyi Jia, Elizabeth Gorbonos et al.

With the booming of personalized recipe sharing networks (e.g., Yummly), a deluge of recipes from different cuisines could be obtained easily. In this paper, we aim to solve a problem which many home-cooks encounter when searching for recipes online. Namely, finding recipes which best fit a handy set of ingredients while at the same time follow healthy eating guidelines. This task is especially difficult since the lions share of online recipes have been shown to be unhealthy. In this paper we propose a novel framework named NutRec, which models the interactions between ingredients and their proportions within recipes for the purpose of offering healthy recommendation. Specifically, NutRec consists of three main components: 1) using an embedding-based ingredient predictor to predict the relevant ingredients with user-defined initial ingredients, 2) predicting the amounts of the relevant ingredients with a multi-layer perceptron-based network, 3) creating a healthy pseudo-recipe with a list of ingredients and their amounts according to the nutritional information and recommending the top similar recipes with the pseudo-recipe. We conduct the experiments on two recipe datasets, including Allrecipes with 36,429 recipes and Yummly with 89,413 recipes, respectively. The empirical results support the framework's intuition and showcase its ability to retrieve healthier recipes.

Meng Chen, Xiaohui Yu, Yang Liu

The widespread use of positioning devices (e.g., GPS) has given rise to a vast body of human movement data, often in the form of trajectories. Understanding human mobility patterns could benefit many location-based applications. In this paper, we propose a novel generative model called TraLFM via latent factor modeling to mine human mobility patterns underlying traffic trajectories. TraLFM is based on three key observations: (1) human mobility patterns are reflected by the sequences of locations in the trajectories; (2) human mobility patterns vary with people; and (3) human mobility patterns tend to be cyclical and change over time. Thus, TraLFM models the joint action of sequential, personal and temporal factors in a unified way, and brings a new perspective to many applications such as latent factor analysis and next location prediction. We perform thorough empirical studies on two real datasets, and the experimental results confirm that TraLFM outperforms the state-of-the-art methods significantly in these applications.

Yueting Chen, Xiaohui Yu, Nick Koudas

Recent advances in Computer Vision and Deep Learning made possible the efficient extraction of a schema from frames of streaming video. As such, a stream of objects and their associated classes along with unique object identifiers derived via object tracking can be generated, providing unique objects as they are captured across frames. In this paper we initiate a study of temporal queries involving objects and their co-occurrences in video feeds. For example, queries that identify video segments during which the same two red cars and the same two humans appear jointly for five minutes are of interest to many applications ranging from law enforcement to security and safety. We take the first step and define such queries in a way that they incorporate certain physical aspects of video capture such as object occlusion. We present an architecture consisting of three layers, namely object detection/tracking, intermediate data generation and query evaluation. We propose two techniques,MFS and SSG, to organize all detected objects in the intermediate data generation layer, which effectively, given the queries, minimizes the number of objects and frames that have to be considered during query evaluation. We also introduce an algorithm called State Traversal (ST) that processes incoming frames against the SSG and efficiently prunes objects and frames unrelated to query evaluation, while maintaining all states required for succinct query evaluation. We present the results of a thorough experimental evaluation utilizing both real and synthetic data establishing the trade-offs between MFS and SSG. We stress various parameters of interest in our evaluation and demonstrate that the proposed query evaluation methodology coupled with the proposed algorithms is capable to evaluate temporal queries over video feeds efficiently, achieving orders of magnitude performance benefits.