Text-to-SQL translates natural language questions into executable SQL queries, enabling intuitive database access for non-experts. While large language models achieve strong performance on Text-to-SQL with prompting, they still struggle with complex queries that involve deeply nested logic or multiple clauses. A widely used approach employs SQL skeletons--intermediate representations of query logic--to streamline generation, but existing methods are limited by their reliance on a single structural hypothesis and lack of progressive reasoning. To overcome these limitations, we propose LEAF-SQL, a novel framework that reframes skeleton prediction as a coarse-to-fine tree search process. LEAF-SQL enables systematic exploration of diverse structural hypotheses with adaptive refinement. Several key techniques are employed in LEAF-SQL: (1) a three-level skeleton hierarchy to guide the search, (2) a Skeleton Formulation Agent to generate diverse candidates, and (3) a Skeleton Evaluation Agent to efficiently prune the search space. This integrated design yields skeleton candidates that are both structurally diverse and granularity-adaptive, providing a stronger foundation for the SQL generation. Extensive experiments show that LEAF-SQL consistently improves the performance of various LLM backbones. On the official hidden test set of the challenging BIRD benchmark, our method achieves 71.6 execution accuracy, which outperforms leading search-based and skeleton-based methods, affirming its effectiveness for complex queries.

Recent advancements in event argument extraction (EAE) involve incorporating useful auxiliary information into models during training and inference, such as retrieved instances and event templates. These methods face two challenges: (1) the retrieval results may be irrelevant and (2) templates are developed independently for each event without considering their possible relationship. In this work, we propose DEGAP to address these challenges through a simple yet effective components: dual prefixes, i.e. learnable prompt vectors, where the instance-oriented prefix and template-oriented prefix are trained to learn information from different event instances and templates. Additionally, we propose an event-guided adaptive gating mechanism, which can adaptively leverage possible connections between different events and thus capture relevant information from the prefix. Finally, these event-guided prefixes provide relevant information as cues to EAE model without retrieval. Extensive experiments demonstrate that our method achieves new state-of-the-art performance on four datasets (ACE05, RAMS, WIKIEVENTS, and MLEE). Further analysis shows the impact of different components.

Event extraction is a long-standing and challenging task in natural language processing, and existing studies mainly focus on extracting events within sentences. However, a significant problem that has not been carefully investigated is whether an “event topic” can be identified to represent the main aspects of extracted events. This article formulates the “topic event” extraction problem, aiming to identify a representative event from extracted ones. Specifically, after defining the topic event, we develop a multifocal graph-based framework to handle the extraction task. To enrich the associations of events and their tokens, we construct four event graphs, including the event subgraph and three event-associated graphs (i.e., event dependency parsing graph, event organization graph, and event share token graph), that reflect the internal and external structures of events, respectively. Subsequently, we design a multi-attention event-graph neural network to capture these event graph structures and improve event subgraph embedding. Finally, the output embeddings in the last layer of each channel are concatenated and fed into a fully connected network for topic event recognition. Extensive experiments validate the effectiveness of our method, and the results confirm its superiority over state-of-the-art baselines. In-depth analyses explore the essential factors (e.g., graph structures, attentions, feature generation method, etc.) determining the extraction performance.

Existing models on event detection share three-fold limitations, including (1) insufficient consideration of the structures between dependency relations, (2) limited exploration of the directed-edge semantics, and (3) issues in strengthening the event core arguments. To tackle these problems, we propose a dependency structure-enhanced event detection framework. In addition to the traditional token dependency parsing tree, denoted as TDG, our model considers the dependency edges in it as new nodes and constructs a dependency relation graph (DRG). DRG allows the embedding representations of dependency relations to be updated as nodes rather than edges in a graph neural network. Moreover, the levels of core argument nodes in the two graphs are adjusted by dependency relation types in TDG to enhance their status. Subsequently, the two graphs are further encoded and jointly trained in graph attention networks (GAT). Importantly, we design an interaction strategy of node embedding for the two graphs and refine the attention coefficient computational method to encode the semantic meaning of directed edges. Extensive experiments are conducted to validate the effectiveness of our method, and the results confirm its superiority over the state-of-the-art baselines. Our model outperforms the best benchmark with the F1 score increased by 3.5 and 3.4 percentage points on ACE2005 English and Chinese corpus.

Large language models (LLMs) with prompting have achieved encouraging results on many natural language processing (NLP) tasks based on task-tailored promptings. Text-to-SQL is a critical task that generates SQL queries from natural language questions. However, prompting on LLMs haven’t show superior performance on Text-to-SQL task due to the absence of tailored promptings. In this work, we propose three promptings specifically designed for Text-to-SQL: SL-prompt, CC-prompt, and SL+CC prompt. SL-prompt is designed to guide LLMs to identify relevant tables; CC-prompt directs LLMs to generate SQL clause by clause; and SL+CC prompt is proposed to combine the strengths of these above promptings. The three prompting strategies makes three solutions for Text-to-SQL. Then, another prompting strategy, the RS-prompt is proposed to direct LLMs to select the best answer from the results of the solutions. We conducted extensive experiments, and experimental results show that our method achieved an execution accuracy of 86.2% and a test-suite accuracy of 76.9%, which is 1.1%, and 2.7% higher than the current state-of-the-art Text-to-SQL methods, respectively. The results confirmed that the proposed promptings enhanced the capabilities of LLMs on Text-to-SQL. Experimental results also show that the granularity of schema linking and the order of clause generation have great impact on the performance, which are considered little in previous research.

Document-level event argument extraction aims to identify argument spans and predict the roles they play in the event from a single document. To enhance argument extraction by better leveraging relevant context information, most existing methods focus solely on context information from the separate perspective of either event or role, without concurrently considering the influence of both event and role. Additionally, existing methods struggle to effectively handle the many-to-many relationship between arguments and roles. Therefore, this paper introduces the CasDEE (Cascade Constrained Pooling and Relevance Evaluation for Document-level Event Argument Extraction) model, which filters context information relevant to the target event and its roles following a constraining pathway of event, role, and argument boundary, and integrates it into the representations of both the candidate argument and the role. To tackle the many-to-many relationship between arguments and roles, CasDEE first utilizes a relevance function to evaluate the relevance of each span-role pair, and then selects all span-role pairs whose relevance are greater than the given threshold as the final argument extraction result. Extensive experiments show that CasDEE achieves state-of-the-art performance on RAMS and WikiEvents datasets. Further experiments reveal the superior performance of CasDEE in addressing long-range dependencies, handling the many-to-many relationship between arguments and roles, and multiple events extraction.

With the continued proliferation of wireless communication and mobile devices equipped with built-in GPS sensors, burgeoning applications of location-based services are springing up, such as mobile crowdsourcing applications (MCS), which are revolutionizing our daily lives. However, collecting and sharing continuously spatio-temporal data to a service provider of MCS applications will incur users' concerns about their privacy. In this paper, we study the problem of locally differentially private data aggregation over distributed spatial-temporal infinite streams. To this end, we proposed a LDP-based framework for dealing with the problem. Firstly, we propose a novel model of (w, ε)-Clustering-based Local Differential Privacy ((w, ε)-CLDP) to capture the temporal and spatial correlations in spatio-temporal infinite stream while guaranteeing stringent differential privacy. Secondly, we develop an efficient GRR-based Local Budget Absorption (LBA) mechanism as a building block for achieving (w, ε)-CLDP and present its privacy analysis. On this basis, we present a framework of real-time spatio-temporal data aggregation over distributed infinite streams with an untrusted server. Lastly, we conduct experiments on two real-world datasets to validate our framework. The results manifest that the LBA-based framework is optimal in data utility for real-time spatio-temporal data aggregation with a rigorous privacy guarantee.

With the increasing scale of mental health problems in modern society, the scarcity of professional assistance is alarming, especially in developing countries. To address this, some online forums have emerged to provide users with useful information and help. However, a user grappling with mental health problems often struggles to find the needed information and assistance on these forums. This is primarily due to the limitations of existing search approaches that often fail to take the characteristics of mental health text into account. In this paper, we propose a new task of mental-health-oriented question-answering (MHQA) which aims to retrieve the appropriate responses for a question post by incorporating the important criteria related to mental health. Our proposed approach, JMS-QA, matches the question post and candidate responses while jointly detecting their latent mental health signals. This enables the method to incorporate mental health signals into its representations. To test the effectiveness of our approach, we create a new dataset for MHQA and conduct experiments on it. The experimental results show that JMS-QA outperforms existing state-of-the-art methods.

Document-level event extraction is a long-standing challenging information retrieval problem involving a sequence of sub-tasks: entity extraction, event type judgment, and event type-specific multi-event extraction. However, addressing the problem as multiple learning tasks leads to increased model complexity. Also, existing methods insufficiently utilize the correlation of entities crossing different events, resulting in limited event extraction performance. This article introduces a novel framework for document-level event extraction, incorporating a new data structure called token-event-role and a multi-channel argument role prediction module. The proposed data structure enables our model to uncover the primary role of tokens in multiple events, facilitating a more comprehensive understanding of event relationships. By leveraging the multi-channel prediction module, we transform entity and multi-event extraction into a single task of predicting token–event pairs, thereby reducing the overall parameter size and enhancing model efficiency. The results demonstrate that our approach outperforms the state-of-the-art method by 9.5 percentage points in terms of the F 1 score, highlighting its superior performance in event extraction. Furthermore, an ablation study confirms the significant value of the proposed data structure in improving event extraction tasks, further validating its importance in enhancing the overall performance of the framework.

Learning topic hierarchies from a multi-domain corpus is crucial in topic modeling as it reveals valuable structural information embedded within documents. Despite the extensive literature on hierarchical topic models, effectively discovering inter-topic correlations and differences among subtopics at the same level in the topic hierarchy, obtained from multiple domains, remains an unresolved challenge. This article proposes an enhanced nested Chinese restaurant process (nCRP), nCRP+, by introducing an additional mechanism based on Chinese restaurant franchise (CRF) for aspect-sharing pattern extraction in the original nCRP. Subsequently, by employing the distribution extracted from nCRP+ as the prior distribution for topic hierarchy in the hierarchical Dirichlet processes (HDP), we develop a hierarchical topic model for multi-domain corpus, named rHDP. We describe the model with the analogy of Chinese restaurant franchise based on the central kitchen and propose a hierarchical Gibbs sampling scheme to infer the model. Our method effectively constructs well-established topic hierarchies, accurately reflecting diverse parent-child topic relationships, explicit topic aspect sharing correlations for inter-topics, and differences between these shared topics. To validate the efficacy of our approach, we conduct experiments using a renowned public dataset and an online collection of Chinese financial documents. The experimental results confirm the superiority of our method over the state-of-the-art techniques in identifying multi-domain topic hierarchies, according to multiple evaluation metrics.

Qizhi Wan, Changxuan Wan, Keli Xiao, Dexi Liu, Chenliang Li, Bolong Zheng, Xiping Liu, Rong Hu. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Data-to-text generation takes structured data as input and produces text that sufficiently describes the data as output. Recently, it has received a lot of attention from both research field and industry. However, as a critical data form, time series is less discussed in this domain. This paper proposes Repr2Seq, a data-to-text generation model for time series. To better capture the structure and core information of time series, Repr2Seq obtains representation vectors using time series representation learning methods, which are then fed into a neural network-based model to generate text sequences. To demonstrate the effectiveness of Repr2Seq, a dataset consisting of stock price series and corresponding comments is proposed. Experiments show that Repr2Seq achieves significant improvement over standard approaches and leads to satisfactory results. We also conduct experiments to investigate the effect of hyperparameters on the model and detect performance improvements in various settings.

In event-based social networks (EBSN), group event recommendation has become an important task for groups to quickly find events that they are interested in. Existing methods on group event recommendation either consider just one type of information, explicit or implicit, or separately model the explicit and implicit information. However, these methods often generate a problem of data sparsity or of model vector redundancy. In this paper, we present a Graph Multi-head Attention Network (GMAN) model for group event recommendation that integrates the explicit and implicit information in EBSN. Specifically, we first construct a user-explicit graph based on the user's explicit information, such as gender, age, occupation and the interactions between users and events. Then we build a user-implicit graph based on the user's implicit information, such as friend relationships. The incorporated both explicit and implicit information can effectively describe the user's interests and alleviate the data sparsity problem. Considering that there may be a correlation between the user's explicit and implicit information in EBSN, we take the user's explicit vector representation as the input of the implicit information aggregation when modeling with graph neural networks. This unified user modeling can solve the aforementioned problem of user model vector redundancy and is also suitable for event modeling. Furthermore, we utilize a multi-head attention network to learn richer implicit information vectors of users and events from multiple perspectives. Finally, in order to get a higher level of group vector representation, we use a vanilla attention mechanism to fuse different user vectors in the group. Through experimenting on two real-world Meetup datasets, we demonstrate that GMAN model consistently outperforms state-of-the-art methods on group event recommendation.

Mental-health-oriented question-answering (MH-QA) aims at retrieving an appropriate response for a question post involving mental health issues, which will be used to assist counsellors to reply to the support seeker. This task is different from the general QA task because many additional criteria such as emotions are involved. In this paper, we propose the Multi-Feature Graph Convolutional Network model (MF-GCN) to integrate not only semantic features, but also mental health related features and context features, to match question posts and responses. Different types of feature are exploited through multiple graph convolutional networks. A new dataset is constructed for MH-QA to evaluate our model. Experimental results show that our model with mental health features significantly outperforms a wide range of state-of-the-art models without them.

Sharing a service among multiple users could bring benefit to users by reducing their service price and also benefit service providers by allowing them to make more profit. Shared services usually have a capacity limit. To construct a win-win situation between users and providers through service sharing, it is necessary to reasonably organize users with similar service requests into groups under the limits of group sizes. This paper explores methods to enable user grouping for sharing services with the limit on the group size. Based on user request descriptions and evaluation, a connection relation is built to present which two users could be in a group. Semi-groups can be derived from the connection relation and two grouping schemes satisfying the group size limit are further determined through proposed algorithms to meet different service expectations. Finally, a case study and a simulation are given to demonstrate the effectiveness of the proposed methods in grouping users to provide higher service benefits to both users and providers.

Summary The switched reluctance motor (SRM) has the high level of vibration and noise, whose application is greatly limited in many fields. In this article, a novel stator for a SRM is designed to reduce torque ripple and is suitable for one‐directional or low‐speed reverse rotation. On the one hand, a shoe is added to the stator pole side in the rotating forward direction from the completely unaligned to the aligned position. On the other hand, a slot is set on the top of stator tooth. First, causes of vibration and torque ripple are analyzed. Second, it is indicated by the equivalent magnetic circuit (EMC) method that the air gap permeability of the model with the single‐sided pole shoe is larger than that of the conventional model in the nonoverlapping region between the stator and rotor. The influences of the angle and height of shoe on the air gap magnetic density, radial force, and torque ripple are analyzed. Third, the results from EMC method show that the radial force can be reduced by opening a slot at the top of stator pole arc. The width and depth of slot which has effect on the radial force and torque ripple are analyzed. Finally, finite element analysis results show that the decreases of 34.52% for the radial force wave peak and 43.39% for the torque ripple are achieved at the cost of a decrease of 8.54% for the average torque of the single‐sided pole and the slot model of the stator pole compared with the conventional model. Some experiments for the two prototypes verify that the torque ripple and radial force can be reduced by adopting the new model.

With the rapid development of wireless communication technologies, location-based social networks (LBSNs) like foursquare and Gowalla have become very popular. Point of interest (POI) recommendation is a kind of important recommendation in LBSNs for enhancing user experiences. Unlike online social networks, LBSNs have a great deal of check-in data and comment information, which can provide valuable information for POI recommendation. In this paper, a novel recommendation strategy using tensor factorization is proposed for improving accurate rate of POI recommendation. Firstly, the latent dirichlet allocation(LDA) topic model is used to extract topic information and generate topic probability distribution of each POI based on comment information from users. Secondly, the check-in data of each user is divided into multiple data slices corresponding to each hour of a day. By connecting with the topic distributions of the visited POIs of each user, a user-topic-time tensor is conducted to present the potential preferences of all users. Finally, a higher order singular value decomposition (HOSVD) algorithm is employed to decompose the third-order tensor, to get dense preference information for POI recommendation. The experiments on a real dataset show that the proposed approach have better performance than the baseline methods.

Nowadays, Radio Frequency Identification (RFID) technology has been widely employed in the fields of object positioning, tracking and monitoring. However, there are a large number of redundant data generated in RFID systems due to duplicate detection and cross detection. Since RFID data is usually streaming, uncertain and mobile data, traditional static data and data stream filtering strategies cannot be applied to filter the RFID data effectively. In the paper, we first present a three-phase filtering framework under a block-based sliding window model. Aiming to filter the temporal redundant events, we propose an approximate Probability Synthesis Bloom Filter (PSBF) and discuss its filter principle, update rules and error rate in details. Comparing with the existing RFID filters, PSBF can not only filter the redundant probabilistic events, but also can calculate object existential probabilities with temporal decaying, and handle with the situations of location movement and staying at the overlapping areas among multiple readers correctly. The experiments on the simulated dataset show that the proposed filter outperforms the state-of-the-art filtering method.

This paper studies the generic location-aware rank query (GLRQ) over a set of location-aware objects. A GLRQ is composed of a spatial location, a set of keywords, a query predicate, and a ranking function formulated on location, text and other attributes. The result consists of k objects satisfying the predicate ranked according to the ranking function. An example is a query searching for the restaurants that 1) are nearby, 2) offer “American” food, and 3) have high ratings (rating > 4.0). Such queries can not be processed efficiently using existing techniques. In this work, we propose a novel framework called LINQ for efficient processing of GLRQs. To handle the predicate and the attribute-based scoring, we devise a new index structure called synopses tree, which contains the synopses of different subsets of the dataset. The synopses tree enables pruning of search space according to the satisfiability of the predicate. To process the query constraints over the location and keywords, the framework integrates the synopses tree with the spatio-textual index such as IR-tree. The framework therefore is capable of processing the GLRQs efficiently and holistically. We conduct extensive experiments to demonstrate that our solution provides excellent query performance.

Keyword search is an effective paradigm for information discovery and has been introduced recently to query XML documents. In this paper, we address the problem of returning clustered results for keyword search on XML documents. We first propose a novel semantics for answers to an XML keyword query. The core of the semantics is the conceptually related relationship between keyword matches, which is based on the conceptual relationship between nodes in XML trees. Then, we propose a new clustering methodology for XML search results, which clusters results according to the way they match the given query. Two approaches to implement the methodology are discussed. The first approach is a conventional one which does clustering after search results are retrieved; the second one clusters search results actively, which has characteristics of clustering on the fly. The generated clusters are then organized into a cluster hierarchy with different granularities to enable users locate the results of interest easily and precisely. Experimental results demonstrate the meaningfulness of the proposed semantics as well as the efficiency of the proposed methods.

In XML retrieval, nodes with different tags play different roles in XML documents and then tags should be reflected in the relevance ranking. An automatic method is proposed in this paper to infer the weights of tags. We first investigate 15 features about tags, and then select five of them based on the correlations between these features and manual tag weights. Using these features, a tag weight assignment model, ATG, is designed. We evaluate the performance of ATG on two real data sets, IEEECS and Wikipedia from two different perspectives. One is to evaluate the quality of the model by measuring the correlation between weights generated by our model and those given by experts. The other is to test the effectiveness of the model in improving retrieval performance. Experimental results show that the tag weights generated by ATG are highly correlated with the manually assigned weights and the ATG model improves retrieval effectiveness significantly.

XML documents can be retrieved by means of not only content-only (CO) queries, but also content-and-structure (CAS) queries. Though promising better retrieval precision, CAS queries introduce several new challenges. To address these challenges, we propose a novel approach for XML CAS retrieval. The distinctive feature of the approach is that it adopts a content-oriented point of view. Specifically, the approach first decomposes a CAS query into several fragments, then retrieves results for each query fragment in a content-centric way, and finally scores each answer node. The approach is adaptive to versatile homogeneous and heterogeneous data environments. To assess the relevance of retrieval results to a query fragment, we present a scoring strategy that measures relevance from both content and structure perspectives. In addition, an effective approach is proposed to infer answer nodes based on the CAS query and document structure. An efficient algorithm is also presented for CAS retrieval. Finally, we demonstrate the effectiveness of the proposed methods through comprehensive experimental studies.