双语知识库中关联实例的多策略提取机制

双语库是翻译记忆系统最重要的组成部分之一。从有限规模的双语库中提取更多的符合用户当前翻译需要的关联实例是翻译记忆技术研究的主要内容,本文首先对当前基于单一方法的实例检索算法存在的局限性进行了分析,并在对双语库进行知识化表示的基础上,提出了基于多策略的关联实例提取机制,即综合运用句子句法结构匹配、句子编辑距离计算、句子短语片段匹配、词汇语义泛化、基于扩展信息(如: 句子来源、所属专业、应用频度等信息)的优选等策略进行关联实例提取。试验结果表明,该方法有效提高了关联实例的召回数量和质量,明显改善了对用户的辅助效果。     

Addressing the Ontology Acquisition Bottleneck Through Reverse Ontological Engineering

The use of ontologies in knowledge engineering arose as a solution to the difficulties associated with acquiring knowledge, commonly referred to as the knowledge acquisition bottleneck. The knowledge-level model represented in an ontology provides a much more structured and principled approach compared with earlier transfer-of-symbolic-knowledge approaches but brings with it a new problem, which can be termed the ontology-acquisition (and maintenance) bottleneck. Each ontological approach offers a different structure, different terms and different meanings for those terms. The unifying theme across approaches is the considerable effort associated with developing, validating and connecting ontologies. We propose an approach to engineering ontologies by retrospectively and automatically discovering them from existing data and knowledge sources in the organization. The method offered assists in the identification of similar and different terms and includes strategies for developing a shared ontology. The approach uses a human-centered, concept-based knowledge processing technique, known as formal concept analysis, to generate an ontology from examples. To assist classification of examples and to identify the salient features of the example, we use a rapid and incremental knowledge acquisition and representation technique, known as ripple-down rules. The method can be used as an alternative or complement to other approaches.     

An emotion understanding framework for intelligent agents based on episodic and semantic memories

Emotional intelligence is the ability to process information about one’s own emotions and the emotions of others. It involves perceiving emotions, understanding emotions, managing emotions and using emotions in thought processes and in other activities. Emotion understanding is the cognitive activity of using emotions to infer why an agent is in an emotional state and which actions are associated with the emotional state. For humans, knowledge about emotions includes, in part, emotional experiences (episodic memory) and abstract knowledge about emotions (semantic memory). In accordance with the need for more sophisticated agents, the current research aims to increase the emotional intelligence of software agents by introducing and evaluating an emotion understanding framework for intelligent agents. The framework organizes the knowledge about emotions using episodic memory and semantic memory. Its episodic memory learns by storing specific details of emotional events experienced firsthand or observed. Its semantic memory is a lookup table of emotion-related facts combined with semantic graphs that learn through abstraction of additional relationships among emotions and actions from episodic memory. The framework is simulated in a multi-agent system in which agents attempt to elicit target emotions in other agents. They learn what events elicit emotions in other agents through interaction and observation. To evaluate the importance of different memory components, we run simulations with components “lesioned”. We show that our framework outperformed Q-learning, a standard method for machine learning.     

Squeezing More Past Knowledge for Online Class-Incremental Continual Learning

Continual learning (CL) studies the problem of learning to accumulate knowledge over time from a stream of data. A crucial challenge is that neural networks suffer from performance degradation on previously seen data, known as catastrophic forgetting, due to allowing parameter sharing. In this work, we consider a more practical online class-incremental CL setting, where the model learns new samples in an online manner and may continuously experience new classes. Moreover, prior knowledge is unavailable during training and evaluation. Existing works usually explore sample usages from a single dimension, which ignores a lot of valuable supervisory information. To better tackle the setting, we propose a novel replay-based CL method, which leverages multi-level representations produced by the intermediate process of training samples for replay and strengthens supervision to consolidate previous knowledge. Specifically, besides the previous raw samples, we store the corresponding logits and features in the memory. Furthermore, to imitate the prediction of the past model, we construct extra constraints by leveraging multi-level information stored in the memory. With the same number of samples for replay, our method can use more past knowledge to prevent interference. We conduct extensive evaluations on several popular CL datasets, and experiments show that our method consistently outperforms state-of-the-art methods with various sizes of episodic memory. We further provide a detailed analysis of these results and demonstrate that our method is more viable in practical scenarios.      

Knowledge Compilation Using the Extension Rule

In this paper, we define a new class of tractable theories: EPCCL theories. Using EPCCL theories as a target language, we propose a new method for knowledge compilation. It is different from existing approaches in that both the compilation and the querying are based on the extension rule, a newly introduced inference rule. With our compilation method, arbitrary queries about the compiled knowledge base can be answered in linear time in the size of the compiled knowledge base. For some theories, the compilation can be done very efficiently, and the size of the compiled theory is small. Furthermore, our method suggests a new family of knowledge compilation methods.     

Transforming examples for multiclass boosting

AdaBoost.M2 and AdaBoost.MH are boosting algorithms for learning from multiclass datasets. They have received less attention than other boosting algorithms because they require base classifiers that can handle the pseudoloss or Hamming loss, respectively. The difficulty with these loss functions is that each example is associated with k weights, where k is the number of classes. We address this issue by transforming an m-example dataset with k weights per example into a dataset with km examples and one weight per example. Minimising error on the transformed dataset is equivalent to minimising loss on the original dataset. Resampling the transformed dataset can be used for time efficiency and base classifiers that cannot handle weighted examples. We empirically apply the transformation on several multiclass datasets using naive Bayes and decision trees as base classifiers. Our experiment shows that it is competitive with AdaBoost.ECC, a boosting algorithm using output coding.     

基于知识管理的 BIM 模型建筑设计合规性自动检查系统研究

建筑设计合规性自动检查对保证建筑信息模型(BIM)符合设计规范要求,增加规范 检查自动化程度具有重要意义。结合合规性检查理论与专家系统方法,提出了以BIM 模型为检 查对象的合规性自动检查系统框架,以规则知识与推理机制分开的方式实现合规性检查过程。 以《住宅设计规范》为例,对规范中的条文进行知识分析,总结出规范知识表达式,构建规则 库和规则库访问机制;建立了逻辑策略下推理机制,将规则库中的规则信息与BIM 信息进行推 理,输出检查结果;最后构建了合规性检查系统验证平台,通过BIM 模型实例完成模型数据提 取及规则推理的过程,实现了合规性检查的功能,验证了合规性检查方法框架。该方法在一定 程度上能够指导后续的合规性检查相关研究,有效提高BIM 模型的建筑设计合规性检查效率, 保证检查质量,促进建筑工程领域信息化的发展。     

Translating a knowledge base into HTML

The problem treated in this paper is the slow response of inference engines, especially in a multi-user environment. In this paper, we present a solution that caches all the possible answers to all the possible consultations, by means of translating the entire knowledge base into a set of HTML documents. We prove that the consumption of memory in this case is acceptable (depends linearly on the number of rules in the initial knowledge base), and that the translational algorithm has polynomial complexity. This solution outperforms any other possible solution, since in this case the infering time becomes equal to zero.     

Maximal consistent extensions of information systems relative to their theories

Knowledge encoded in information systems can be represented by different sets of rules generated by these systems. One can consider sets of deterministic, nondeterministic or probabilistic rules. Such sets of rules can be treated as theories of information systems. Any such a theory generated from a given information system corresponds to a subjective view on knowledge encoded in this information system. Such theories can be used for solving different problems. For example, the maximal consistent extensions of information systems were studied for synthesis of concurrent processes specified by information systems. In this approach, the maximal consistent extension of a given information system consists of all objects perceived by means of attributes which are consistent with the theory including all the so called true and realizable deterministic rules extracted from the original information system. In this paper, we report results on the maximal consistent extensions of information systems relative to some other theories of information systems, e.g., theories consisting of rules such as true and realizable inhibitory rules, true inhibitory rules, and true deterministic rules. We also discuss algorithmic problems related to the maximal consistent extensions. In particular, from the obtained results it follows that solutions based on these new sets of rules, e.g., on inhibitory rules can be of higher quality than in the case of deterministic rules.     

One-shot learning of object categories

Learning visual models of object categories notoriously requires hundreds or thousands of training examples. We show that it is possible to learn much information about a category from just one, or a handful, of images. The key insight is that, rather than learning from scratch, one can take advantage of knowledge coming from previously learned categories, no matter how different these categories might be. We explore a Bayesian implementation of this idea. Object categories are represented by probabilistic models. Prior knowledge is represented as a probability density function on the parameters of these models. The posterior model for an object category is obtained by updating the prior in the light of one or more observations. We test a simple implementation of our algorithm on a database of 101 diverse object categories. We compare category models learned by an implementation of our Bayesian approach to models learned from by maximum likelihood (ML) and maximum a posteriori (MAP) methods. We find that on a database of more than 100 categories, the Bayesian approach produces informative models when the number of training examples is too small for other methods to operate successfully.     

Criteria of progress for information systems design theories

According to Kuhn, science and progress are strongly interrelated. In this paper, we define criteria of progress for design theories. A broad analysis of the literature on information systems design science reveals that there is no consensus on the criteria of progress for design theories. We therefore analyze different concepts of progress for natural science theories. Based on well-founded criteria stemming from the philosophy of science and referring to natural science theories, we develop a set of criteria of progress for design theories. In summary, our analysis results in six criteria of progress for design theories: A design theory is partially progressive compared to another if it is ceteris paribus (1) more useful, (2) internally more consistent, (3) externally more consistent, (4) more general, (5) simpler, or (6) more fruitful of further research. Although the measurement of these criteria is not the focus of this paper, the problem of measurement cannot be totally neglected. We therefore discuss different methods for measuring the criteria based on different concepts of truth: the correspondence theory of truth, the coherence theory of truth, and the consensus theory of truth. We finally show the applicability of the criteria with an example.     

OntoSTEP: Enriching product model data using ontologies

The representation and management of product lifecycle information is critical to any manufacturing organization. Different modeling languages are used at different lifecycle stages, for example STEP’s EXPRESS may be used at a detailed design stage, while UML may be used for initial design stages. It is necessary to consolidate product information created using these different languages to build a coherent knowledge base. In this paper, we present an approach to enable the translation of STEP schema and its instances to Ontology Web Language (OWL). This gives a model–which we call OntoSTEP–that can easily be integrated with any OWL ontologies to create a semantically rich model. As an example, we combine geometry information represented in STEP with non-geometry information, such as function and behavior, represented using the NIST’s Core Product Model (CPM). A plug-in for Protégé is developed to automate the different steps of the translation. As additional benefits, reasoning, inference procedures, and queries can be performed on enriched legacy CAD models. We describe the rules for the translation from EXPRESS to OWL, and illustrate the benefits of OWL translation with an example. We will also describe how these mapping rules can be implemented through meta-model based transformations, which can be used to map other languages to OWL.     

Extraction of rules from natural objects for automated mechanicalprocessing

In process applications, fast and accurate extraction of complex information from an object for the purpose of mechanical processing of that object, is often required. In this paper, a general rule-based approach is developed using a database of measurable geometric “features” and associated complex information. The rules relate the features to the complex processing information. During the on-line processing, the object features are measured and passed into the rule base. The output from the rule base is the complex information that is needed to process the object. A methodology is developed to generate probabilistic rules for the rule base using multivariate probability densities. A knowledge integration scheme is also developed which combines statistical knowledge with expert knowledge in order to improve the reliability and efficiency of information extraction. The rule generation methodology is implemented in a knowledge-based vision system for process information recognition. As an illustrative example, the problem of efficient head removal in an automated salmon processing plant is considered     

Sketch theory as a framework for knowledge management

In this paper, we give an overview of sketch theory as a knowledge management framework and discuss its strengths relative to logic, the semantic web and relational algebra. Sketch theory, for example, supports modularity among meta-data, instance data and uncertainty. It also provides a notion of constraint-preserving map. We explore Q-trees as a technique for inference with sketches and compare it to logical deduction. Ideas can be formulated in distinct ways even within a fixed formalism. We illustrate solution of this alignment problem using sketches and the notion of Morita equivalence of logical theories. Sketch theory provides rich notions of contextual view with which we compute illustrative examples. Finally, we outline a program for advancing sketch theory as a complement to other knowledge management technologies and discuss transformations between sketches and other models.     

Using a relational database to support explanation in aknowledge-based system

In knowledge-based system (KBS) applications where the number of facts in the knowledge base is very large, the amount of information that is needed for effective explanation support can become too voluminous to be stored in main memory. The authors present an approach to modeling and managing the information needed for explanation rising a relational database. It is shown how different types of explanation can then be produced by appropriate queries on this database. The authors formulate representative queries for some major types of explanation using ESQL, an extension of structured query language (SQL)     

一种融合实体语义知识的实体集合扩展方法

实体集合扩展是开放式信息抽取的一个重要问题,该问题研究如何从一个语义类的若干实体(称为种子)出发,得到该类别的更多实体。现有实体集合扩展方法主要使用上下文模板或种子在语料中的分布信息进行抽取,其缺点是无法解决种子的歧义问题,而该问题会影响方法的有效性。在该文中,作者提出了一种融合实体语义知识的实体集合扩展方法,通过引入语义知识来解决种子歧义性问题。新方法通过使用Wikipedia实现了语义知识的引入,并把基于语义知识的扩展方法和基于模板的扩展方法相融合。实验表明,与单纯基于上下文方法相比,该文方法在准确率上提升了18.5%,召回率上提升了6.8%,MAP值上提升了22.8%。     

Quantitative knowledge acquisition for expert systems

A common problem in the design of expert systems is the definition of rules from data obtained in system operation or simulation. Whilte it is relatively easy to collect data and to log the comments of human operators engaged in experiments, generalizing such information to a set of rules has not previously been a straightforward task. This paper presents a statistical method for generating rule bases from numerical data, motivated by an example based on aircraft navigation with multiple sensors. The specific objective is to design an expert system that selects a satisfactory suite of measurements from a dissimilar, redundant set, given an arbitrary navigation geometry and possible sensor failures. This paper describes the systematic development of a Navigation Sensor Management (NSM) Expert System from Kalman Filter covariance data. The development method invokes two statistical techniques: Analysis of Variance (ANOVA) and the ID3 algorithm. The ANOVA technique indicates whether variations of problem parameters give statistically different covariance results, and the ID3 algorithm identifies the relationships between the problem parameters using probabilistic knowledge extracted from a simulation example set. ANOVA results show that statistically different position accuracies are obtained when different navigation aids are used, the number of navigation aids is changed, the trajectory is varied, or the performance history is altered. By indicating that these four factors significantly affect the decision metric, an appropriate parameter framework was designed, and a simulation example base was created. The example base contained over 900 training examples from nearly 300 simulations. The ID3 algorithm was then applied to the example base, yielding classification “rules” in the form of decision trees. The NSM expert system consists of seventeen decision trees that predict the performance of a specified integrated navigation sensor configuration. The performance of these decision trees was assessed on two arbitrary trajectories, and the performance results are presented using a predictive metric. The test trajectories used to evaluate the system's performance show that the NSM Expert adapts to new situations and provides reasonable estimates of sensor configuration performance.     

Plausible Justification Trees: A Framework for Deep and Dynamic Integration of Learning Strategies

This article describes a framework for the deep and dynamic integration of learning strategies. The framework is based on the idea that each single-strategy learning method is ultimately the result of certain elementary inferences (like deduction, analogy, abduction, generalization, specialization, abstraction, concretion, etc.). Consequently, instead of integrating learning strategies at a macro level, we propose to integrate the different inference types that generate individual learning strategies. The article presents a concept-learning and theory-revision method that was developed in this framework. It allows the system to learn from one or from several (positive and/or negative) examples, and to both generalize and specialize its knowledge base. The method integrates deeply and dynamically different learning strategies, depending on the relationship between the input information and the knowledge base. It also behaves as a single-strategy learning method whenever the applicability conditions of such a method are satisfied.