Skeleton parsing for complex question answering over knowledge bases

Answering complex questions involving multiple relations over knowledge bases is a challenging task. Many previous works rely on dependency parsing. However, errors in dependency parsing would influence their performance, in particular for long complex questions. In this paper, we propose a novel skeleton grammar to represent the high-level structure of a complex question. This lightweight formalism and its BERT-based parsing algorithm help to improve the downstream dependency parsing. To show the effectiveness of skeleton, we develop two question answering approaches: skeleton-based semantic parsing (called SSP) and skeleton-based information retrieval (called SIR). In SSP, skeleton helps to improve structured query generation. In SIR, skeleton helps to improve path ranking. Experimental results show that, thanks to skeletons, our approaches achieve state-of-the-art results on three datasets: LC-QuAD 1.0, GraphQuestions, and ComplexWebQuestions 1.1.     

Structure-sensitive semantic matching for aggregate question answering over knowledge base

Aggregate question answering essentially returns answers for given questions by obtaining query graphs with unique dependencies between values and corresponding objects. Word order dependency, as the key to uniquely identify dependency of the query graph, reflects the dependencies between the words in the question. However, due to the semantic gap caused by the expression difference between questions encoded with word vectors and query graphs represented with logical formal elements, it is not trivial to match the correct query graph for the question. Most existing approaches design more expressive query graphs for complex questions and rank them just by directly calculating their similarities, ignoring the semantic gap between them. In this paper, we propose a novel Structure-sensitive Semantic Matching(SSM) approach that learns aligned representations of dependencies in questions and query graphs to eliminate their gap. First, we propose a cross-structure matching module to bridge the gap between two modalities(i.e., textual question and query graph). Then, we propose an entropy-based gated AQG filter to remove the structural noise caused by the uncertainty of dependencies. Finally, we present a two-channel query graph representation that fuses the semantics of abstract structure and grounding content of the query graph explicitly. Experimental results show that SSM could learn aligned representations of questions and query graphs to eliminate the gaps between their dependencies, and improves up to 12% (F1 score) on aggregation questions of two benchmark datasets.     

Progressive failure analysis of notched composite plate by utilizing macro mechanics approach

In this study, gradual and sudden reduction methods were combined to simulate a progressive failure in notched composite plates using a macro mechanics approach. Using the presented method, a progressive failure is simulated based on a linear softening law prior to a catastrophic failure, and thereafter, sudden reduction methods are employed for modeling a progressive failure. This combination method significantly reduces the computational cost and is also capable of simultaneously predicting the first and last ply failures (LPFs) in composite plates. The proposed method is intended to predict the first ply failure (FPF), LPF, and dominant failure modes of carbon/epoxy and glass/epoxy notched composite plates. In addition, the effects of mechanical properties and different stacking sequences on the propagation of damage in notched composite plates were studied. The results of the presented method were compared with experimental data previously reported in the literature. By comparing the numerical and experimental data, it is revealed that the proposed method can accurately simulate the failure propagation in notched composite plates at a low computational cost.     

Generating Questions Based on Semi-Automated and End-to-End Neural Network

With the emergence of large-scale knowledge base, how to use triple information to generate natural questions is a key technology in question answering systems. The traditional way of generating questions require a lot of manual intervention and produce lots of noise. To solve these problems, we propose a joint model based on semi-automated model and End-to-End neural network to automatically generate questions. The semi-automated model can generate question templates and real questions combining the knowledge base and center graph. The End-to-End neural network directly sends the knowledge base and real questions to BiLSTM network. Meanwhile, the attention mechanism is utilized in the decoding layer, which makes the triples and generated questions more relevant. Finally, the experimental results on SimpleQuestions demonstrate the effectiveness of the proposed approach.     

粒子群属性聚类的位置隐私保护

针对基于位置服务中连续查询情况下，用户自身属性信息很容易被攻击者获取，并通过关联获得用户位置隐私的情况，提出了一种利用粒子群聚类加速相似属性用户寻找，并由相似属性匿名实现用户位置泛化的隐私保护方法。该方法利用位置隐私保护中常用的可信中心服务器，通过对发送到中心服务器中的查询信息进行粒子群属性聚类，在聚类的过程中加速相似属性用户的寻找过程，由相似属性用户完成位置泛化，以此实现位置隐私保护。实验结果证明，这种基于粒子群属性聚类的隐私保护方法具有高于同类算法的隐私保护能力，以及更快的计算处理速度。     

CineCubes: Aiding data workers gain insights from OLAP queries

In this paper we demonstrate that it is possible to enrich query answering with a short data movie that gives insights to the original results of an OLAP query. Our method, implemented in an actual system, CineCubes, includes the following steps. The user submits a query over an underlying star schema. Taking this query as input, the system comes up with a set of queries complementing the information content of the original query, and executes them. For each of the query results, we execute a set of highlight extraction algorithms that identify interesting patterns and values in the data of the results. Then, the system visualizes the query results and accompanies this presentation with a text commenting on the result highlights. Moreover, via a text-to-speech conversion the system automatically produces audio for the constructed text. Each combination of visualization, text and audio practically constitutes a movie, which is wrapped as a PowerPoint presentation and returned to the user.     

SINA: Semantic interpretation of user queries for question answering on interlinked data

The architectural choices underlying Linked Data have led to a compendium of data sources which contain both duplicated and fragmented information on a large number of domains. One way to enable non-experts users to access this data compendium is to provide keyword search frameworks that can capitalize on the inherent characteristics of Linked Data. Developing such systems is challenging for three main reasons. First, resources across different datasets or even within the same dataset can be homonyms. Second, different datasets employ heterogeneous schemas and each one may only contain a part of the answer for a certain user query. Finally, constructing a federated formal query from keywords across different datasets requires exploiting links between the different datasets on both the schema and instance levels. We present Sina, a scalable keyword search system that can answer user queries by transforming user-supplied keywords or natural-languages queries into conjunctive SPARQL queries over a set of interlinked data sources. Sina uses a hidden Markov model to determine the most suitable resources for a user-supplied query from different datasets. Moreover, our framework is able to construct federated queries by using the disambiguated resources and leveraging the link structure underlying the datasets to query. We evaluate Sina over three different datasets. We can answer 25 queries from the QALD-1 correctly. Moreover, we perform as well as the best question answering system from the QALD-3 competition by answering 32 questions correctly while also being able to answer queries on distributed sources. We study the runtime of SINA in its mono-core and parallel implementations and draw preliminary conclusions on the scalability of keyword search on Linked Data.     

E?cient Video Cutout by Paint Selection

Video cutout refers to extracting moving objects from videos, which is an important step in many video editing tasks. Recent algorithms have limitations in terms of e?ciency, interaction style, and rob...     

The generic model query language GMQL – Conceptual specification,implementation, and runtime evaluation

The generic model query language GMQL is designed to query collections of conceptual models created in arbitrary graph-based modelling languages. Querying conceptual models means searching for particular model subgraphs that comply with a predefined pattern query. Such a query specifies the structural and semantic properties of the model fragment to be returned. In this paper, we derive requirements for a generic model query language from the literature and formally specify the language’s syntax and semantics. We conduct an analysis of GMQL׳s theoretical and practical runtime performance concluding that it returns query results within satisfactory time. Given its generic nature, GMQL contributes to a broad range of different model analysis scenarios ranging from business process compliance management to model translation and business process weakness detection. As GMQL returns results with acceptable runtime performance, it can be used to query large collections of hundreds or thousands of conceptual models containing not only process models, but also data models or organizational charts. In this paper, we furthermore evaluate GMQL against the backdrop of existing query approaches thereby carving out its advantages and limitations as well as pointing toward future research.     

贝叶斯网络查询语义扩展的专家发现方法

专家发现是实体检索领域的一个研究热点，针对经典专家发现模型存在索引术语独立性假设与检索性能低的缺陷，提出一种基于贝叶斯网络模型的专家发现方法。该方法模型采用四层网络结构，能够实现图形化的概率推理，同时运用词向量技术能够实现查询术语的语义扩展。实验结果显示该模型在多个评价指标上均优于经典专家发现模型，能够有效实现查询术语语义扩展，提高专家检索性能。