Ontology learning

Probabilistic Relational Models (PRMs) extend Bayesian networks (BNs) with the notion of class of relational databases. Because of their richness, learning them is a difficult task. In this paper, we propose a method that learns a PRM from data using the semantic knowledge of an ontology describing these data in order to make the learning easier. To present our approach, we describe an implementation based on an ontology of transformation processes and compare its performance to that of a method that learns a PRM directly from data. We show that, even with small datasets, our approach of learning a PRM using an ontology is more efficient.

We present a novel approach to learning taxonomic relations between terms by considering multiple and heterogeneous sources of evidence. In order to derive an optimal combination of these sources, we exploit a machine-learning approach, representing all the sources of evidence as first-order features and training standard classifiers. We consider in particular different features derived from Word-Net, an approach matching Hearst-style patterns in a corpus and on the Web as well as further methods mentioned in the literature. In particular, we explore different classifiers as well as various strategies for dealing with unbalanced datasets. We evaluate our approach by comparing the results with a reference taxonomy for the tourism domain.

We present a novel approach to learning taxonomic relations between terms by considering multiple and heterogeneous sources of evidence. In order to derive an optimal combination of these sources, we exploit a machine-learning ap- proach, representing all the sources of evidence as first-order features and training standard classifiers. We consider in particular different features derived from Word- Net, an approach

We present a novel approach to learning taxonomic relations between terms by considering multiple and heterogeneous sources of evidence. In order to derive an optimal combination of these sources, we exploit a machine-learning approach, representing all the sources of evidence as first-order features and training standard classifiers. We consider in particular different features derived from Word-Net, an approach matching Hearst-style patterns in a corpus and on the Web as well as further methods mentioned in the literature. In particular, we explore different classifiers as well as various strategies for dealing with unbalanced datasets. We evaluate our approach by comparing the results with a reference taxonomy for the tourism domain.

Ontologies now play an important role for enabling the semantic web. They provide a source of precisely defined terms e.g. for knowledge-intensive applications. The terms are used for concise communication across people and applications. Typically the development of ontologies involves collaborative efforts of multiple persons. OntoEdit is an ontology editor that integrates numerous aspects of ontology engineering. This paper focuses on collaborative development of ontologies with OntoEdit which is guided by a comprehensive methodology.

This project aims at creating a network of distributed interoperable semantic services for building more complex ones. These services will be available in semantic Web service libraries, so that they can be invoked by other systems (e.g., semantic portals, software agents, etc.). Thus, to accomplish this objective, the project proposes: a) To create specific technology for developing and composing Semantic Web Services. b) To migrate the WebODE ontology development workbench to this new distributed interoperable semantic service architecture. c) To develop new semantic services (ontology learning, ontology mappings, incremental ontology evaluation, and ontology evolution). d) To develop technological support that eases semantic portal interoperability, using Web services and Semantic Web Services. The project results will be open source, so as to improve their technological transfer. The quality of these results is ensured by a benchmarking process.

ABSTRACT. This introduction tracks the evolution of the definition and role of discourse issues in NLP from the knowledge-intensive “discourse understanding” methods of the 80's to the recent concern with “accessing contents” in vast document bases via data-intensive ...

This work analyses a corpus made of the titles of research projects belonging to the last four European Commission Framework Programmes (FP4, FP5, FP6, FP7) during a time span of nearly two decades (1994-2012). The starting point is the idea of creating a corpus of titles which would constitute a terminological niche, a sort of “cluster map” offering an overall vision on the terms used and the links between them. Moreover, by performing a terminological comparison over a period of time it is possible to trace the presence of obsolete words in outdated research areas as well as of neologisms in the most recent fields. Within this scenario, the minimal purpose is to build a corpus of titles of European projects belonging to the several Framework Programmes in order to obtain a terminological mapping of relevant words in the various research areas: particularly significant would be those terms spread across different domains or those extremely tied to a specific domain. A term could ac...

Ontologies are one of the most used representations to model the domain knowledge. An ontology consists of a set of concepts connected by semantic relations. The construction and evolution of an ontology are complex and timeconsuming tasks. This paper presents DYNAMO-MAS, an Adaptive Multi-Agent System (AMAS) that automates these tasks by co-constructing an ontology from texts with an ontologist. Terms and concepts of a given domain are agentified and they act, according to the AMAS approach, by solving the non cooperative situations they locally perceive at runtime. These agents cooperate to determine their position in the AMAS (that is the ontology) thanks to (i) lexical relations between terms, (ii) some adaptive mechanisms enabling addition, removing or moving of new terms, of concepts and of relations in the ontology as well as (iii) feedbacks from the ontologist about the propositions given by the AMAS. This paper focuses on the instantiation of the AMAS approach to this difficult problem. It presents the architecture of DYNAMO-MAS, and details the cooperative behaviors of the two types of agents we defined for ontology evolution. Finally evaluations made on three different ontologies are given in order to show the genericity of our solution.

Manual ontology development and evolution are complex and time-consuming tasks, even when textual documents are used as knowledge sources in addition to human expertise or existing ontologies. Processing natural language in text produces huge amounts of linguistic data that need to be filtered out and structured. To support both of these tasks, we have developed DYNAMO-MAS, an interactive tool based on an adaptive multi-agent system (adaptive MAS or AMAS) that builds and evolves ontologies from text. DYNAMO-MAS is a partner system to build ontologies; the ontologist interacts with the system to validate or modify its outputs. This paper presents the architecture of DYNAMO-MAS, its operating principles and its evaluation on three case studies.

Recent developments towards knowledge-based applications in general and Semantic Web applications in particular are leading to an increased interest in ontologies and in dynamic methods for developing and maintaining them. As human language is a primary mode of knowledge transfer, ontology learning from relevant text collections has been among the most successful strategies in this work. Such methods mostly combine a certain level of linguistic analysis with statistical and/or machine learning approaches to find potentially interesting concepts and relations between them. Here, we discuss a formalization of this process (in the specific context of the OntoLT tool for ontology learning from text) in order to arrive at a better definition of this task, which we hope to be of use in a more principled comparison of different approaches. As ontology representation formalisms we will consider those that have a model-theoretic semantics, with OWL (and subsets of OWL) being appropriate candidates.

Recent developments towards knowledge-based applications in general and Semantic Web applications in particular are leading to an increased interest in ontologies and in dynamic methods for developing and maintaining them. As human language is a primary mode of knowledge transfer, ontology learning from relevant text collections has been among the most successful strategies in this work. Such methods mostly combine a certain level of linguistic analysis with statistical and/or machine learning approaches to find potentially interesting concepts and relations between them. Here, we discuss a formalization of this process (in the specific context of the OntoLT tool for ontology learning from text) in order to arrive at a better definition of this task, which we hope to be of use in a more principled comparison of different approaches. As ontology representation formalisms we will consider those that have a model-theoretic semantics, with OWL (and subsets of OWL) being appropriate candidates.

A person ontology comprising concepts, attributes and relationships of people has a number of applications in data protection, didentification, population of knowledge graphs for business intelligence and fraud prevention. While artificial neural networks have led to improvements in Entity Recognition, Entity Classification, and Relation Extraction, creating an ontology largely remains a manual process, because it requires a fixed set of semantic relations between concepts. In this work, we present a system for automatically populating a person ontology graph from unstructured data using neural models for Entity Classification and Relation Extraction. We introduce a new dataset for these tasks and discuss our results.

The objective of the research presented in this article is to find representational mechanisms for relating and integrating the collaborative learning elements present in real practical environments, create an integrated ontology that considers and relates these elements, and make use of it to define new collaborative learning scenarios. It is therefore necessary to identify the key ideas underlying the notion of ontology that will be essential in subsequent application development: a list of the basic elements that give rise to a common vocabulary for collaborative learning, and the relationship and dependencies between them. The Activity Theory is used as a theoretical framework for organising the elements in the ontology. This ontology gives rise to the structured elements that form the concpetual structure for the definition and construction of CSCL environments, and the analysis and assessment of group collaboration.

The study focuses on designing an optimized database system using an Entity-Relationship (ER) model and its relational mapping to ensure data integrity and minimize redundancy. Key entities identified include Student, Course, Professor, Department, and Enrollment, with Enrollment being a weak entity dependent on Student and Course. Attributes are categorized as primary keys, multi-valued, and derived attributes to enhance data organization. Relationship modeling defines one-to-one, one-to-many, and many-to-many relationships to establish clear connections between entities. The relational schema is structured using normalization techniques, including First Normal Form (1NF), to remove multi-valued attributes and improve efficiency. Various SQL join operations such as INNER JOIN, LEFT JOIN, RIGHT JOIN, FULL OUTER JOIN, SELF JOIN, and CROSS JOIN are implemented to retrieve meaningful data relationships. The methodology ensures consistency, referential integrity, and streamlined database operations, demonstrating an effective approach to structured data management. IV. ER MODEL STEPS FOR CREATING AN ER MODEL 1. IDENTIFY ENTITIES The first step in designing an ER model is to identify the key entities in the system. Entities represent real-world objects or concepts that have distinct identities. 2. DEFINE ATTRBUTES FOR EACH ENTITY Relation Names Each entity has a set of attributes that define its characteristics. Attributes can be of different types:  Key Attribute: A unique identifier for an entity.  Simple Attribute: A single-valued characteristic (e.g., Name, DOB).  Composite Attribute: Attributes that can be divided into sub-parts.  Derived Attribute: An attribute that is derived from other attributes.

This work analyses a corpus made of the titles of research projects belonging to the last four European Commission Framework Programmes (FP4, FP5, FP6, FP7) during a time span of nearly two decades (1994-2012). The starting point is the idea of creating a corpus of titles which would constitute a terminological niche, a sort of “cluster map” offering an overall vision on the terms used and the links between them. Moreover, by performing a terminological comparison over a period of time it is possible to trace the presence of obsolete words in outdated research areas as well as of neologisms in the most recent fields. Within this scenario, the minimal purpose is to build a corpus of titles of European projects belonging to the several Framework Programmes in order to obtain a terminological mapping of relevant words in the various research areas: particularly significant would be those terms spread across different domains or those extremely tied to a specific domain. A term could ac...

The challenge of effectively constructing ontologies from text documents remains unresolved, posing a critical gap in modern knowledge extraction methodologies. One of the primary obstacles is the lack of a standardized output format across different NLP tools, particularly text parsers, which serve as the foundational step in multi-stage knowledge extraction processes. While several widely used text parsers exist, each excels in specific functions, making it beneficial to leverage multiple parsers for more comprehensive ontology construction. However, this approach introduces the issue of reconciling their disparate output formats.
To address this challenge, we propose using a graph database to store parser outputs in a subject-predicate-object triple format, enabling seamless integration and further processing through rule-based transformations using SPARQL queries. A key advantage of this approach is the ability to execute new transformation rules dynamically, allowing for greater flexibility and efficiency in ontology generation.
As part of our research, we developed an intelligent agent in Java capable of constructing semantic graphs from natural language text using a rule-based approach. The agent was employed to evaluate the relationship between the execution time of syntax-semantic transformation rules and variables such as text corpus size and dataset sample dimensions. This evaluation was made possible through the implementation of first-level reflection for the studied transformation rule.
The results demonstrate that our approach – standardizing parser outputs via a graph database – roves effective in terms of both computational complexity and processing speed. By streamlining the ontology construction process, our method paves the way for advanced automated learning of intelligent agents based on textual information, unlocking new possibilities for modern science in the realm of knowledge extraction and representation.

Information extraction systems learn patterns for extracting pairs instantiating a given relation from text. For instance, for the relation capital of a system might learn extraction patterns such as 'Arg1 is capital of Arg2', or'The embassador of arg2 was called back to Arg1'. Lightly supervised information extraction systems learn extraction patterns by means of a bootstrapping procedure, where a set of seed pairs is used to find patterns associated with the seeds, and where the patterns thus found are used in turn to find potential ...

English. Automatic detection of antonymy is an important task in Natural Language Processing (NLP). However, currently, there is no effective measure to discriminate antonyms from synonyms because they share many common features. In this paper, we introduce APAnt, a new Average-Precision-based measure for the unsupervised identification of antonymy using Distributional Semantic Models (DSMs). APAnt makes use of Average Precision to estimate the extent and salience of the intersection among the most descriptive contexts of two target words. Evaluation shows that the proposed method is able to distinguish antonyms and synonyms with high accuracy, outperforming a baseline model implementing the co-occurrence hypothesis. Natural Language Processing (NLP), ad oggi non esistono sistemi soddisfacenti per risolvere questo problema. Gli antonimi, infatti, condividono molte caratteristiche con i sinonimi, e vengono spesso confusi con essi. In questo articolo introduciamo APAnt, una misura basata sull'Average Precision (AP) per l'identificazione automatica degli antonimi nei Modelli Distribuzionali (DSMs). APAnt fa uso dell'AP per stimare il grado e la rilevanza dell'intersezione tra i contesti più descrittivi di due parole target. I risultati dimostrano che APAnt è in grado di distinguere gli antonimi dai sinonimi con elevata precisione, superando la baseline basata sull'ipotesi della cooccorrenza.

An ontology can be seen as a hierarchical description of concepts in a specific domain. One of the key issues in ontology construction is the acquisition of ontology hierachy. Manual construction of ontology by experts is time-consuming and costly, and timely update is also difficult. In automatic or semi-automatic methods, the general procedure of ontology construction is to first obtain domain specific terms and then acquire the relations among them. Terms can be obtained through terminology extraction. Relationship among terms can then be acquired to construct an appropriate ontology hierachy. Formal Concept Analysis (FCA) is an effective tool for the acquisition and visualization of ontology structure. In using FCA to acquire ontology structure, some existing methods are inclined to use a single type of attributes such as verbs or nouns or adjectives in the context of a term. This paper investigates the use of context words including nouns, verbs, adjectives, and adverbs as well as their combination. Experiments on a general corpus and a domain specific corpus show that the combined attributes of content word types gives the best performance. Contrary to the believe that verbs serve as the best single type of attributes, nouns are better to use as they have overall better average performance over the whole spectrum of threshold values. Another experiment is conducted to examine context words which are qualified as domain specific terms from a so called Core Term List which is aquired separately. The Core Term List contains terms complying with some spedific standards as qualified domain terms. Experiments show a comparable or even better performance of using the Core Term List as context than using dynamically acquired nouns. In fact the core Term List approach gives comparable result to the combined content words. Thus, simple domain knowledge aquired in priori can serve as a good alternative attribute set especially in resource limited cases.

An ontology is a structured knowledgebase of concepts organized by relations among them. But concepts are usually mixed with their instances in the corpora for knowledge extraction. Concepts and their corresponding instances share similar features and are difficult to distinguish. In this paper, a novel approach is proposed to comprehensively obtain concepts with the help of definition sentences and Category Labels in Wikipedia pages. N-gram statistics and other NLP knowledge are used to help extracting appropriate concepts. The proposed method identified nearly 50,000 concepts from about 700,000 Wiki pages. The precision reaching 78.5% makes it an effective approach to mine concepts from Wikipedia for ontology construction.

The success of the Semantic Web research is dependent upon the construction of complete and reliable domain ontologies. In this paper we describe an unsupervised framework for domain ontology enrichment based on mining domain text corpora. Specifically, we enrich the hierarchical backbone of an existing ontology, i.e. its taxonomy, with new domain-specific concepts. The framework is based on an extended model of hierarchical self-organizing maps. As being founded on an unsupervised neural network architecture, the framework can be applied to different languages and domains. Terms extracted by mining a text corpus encode contextual content information, in a distributional vector space. The enrichment behaves like a classification of the extracted terms into the existing taxonomy by attaching them as hyponyms for the nodes of the taxonomy. The experiments reported are in the "Lonely Planet" tourism domain. The taxonomy and the corpus are the ones proposed in the PASCAL ontology learning and population challenge. The experimental results prove that the quality of the enrichment is considerably improved by using semantics based vector representations for the classified (newly added) terms, like the document category histograms (DCH) and the document frequency times inverse term frequency (DF-ITF) weighting scheme.

In general, Muslims all over the world have an innate tendency to hold fast to Islam’s teachings as narrated in the Qur’an and Hadith. The present study is an investigation on the utilization of this adherence for improving the standards of ethical behavior of Muslim IT professionals, particularly software engineers. The principal point of this paper is to develop the importance of ethics among software engineers in order to make them realize the impact of various immoral practices in their field e.g., property violations, general software upgrading, design methodology, software privacy, etc. The moral values put forward in code-of- conduct have been scrutinized from Islamic point of view by studying the same in light of verses in the Qur’an and Hadith of our beloved Prophet Muhammad (P.B.U.H.). It is high time for software engineers and developers to accept the dire need of a paradigm shift in software engineering that integrates divine revelation with reason. The paper, therefore,...

Ontologies stand in the heart of the Semantic Web. Nevertheless, heavyweight or formal ontologies' engineering is being commonly judged to be a tough exercise which requires time and heavy costs. Ontology Learning is thus a solution for this exigency and an approach for the 'knowledge acquisition bottleneck'. Since texts are massively available everywhere, making up of experts' knowledge and their know-how, it is of great value to capture the knowledge existing within such texts. Our approach is thus an interesting research work which tries to answer the challenge of creating concepts' hierarchies from textual data. The significance of such a solution stems from the idea by which we take advantage of the Wikipedia encyclopedia to achieve some good quality results.

To allow for a direct connection of this linguistic information for terms with corresponding classes and properties in a domain ontology, we developed a lexicon model (LingInfo) that enables the definition of LingInfo instances (each of which represents a term) for each class or property. The LingInfo model is represented by use of a meta-class, which allows for the representation of LingInfo instances with each class, where each LingInfo instance represents the linguistic features of a term for a particular class. Applications of ...

Information Retrieval (IR) is the field of computer science that deals with the storage, searching, and retrieving of the documents that satisfy the user need. Distributed Arabic Information Retrieval (DAIR) is a model enables a user to access many searchable Arabic documents reside in different locations. DAIR is more complex than the centralized Arabic IR (AIR) because it requires addressing two significant additional problems that are the resource selection and the results merging. The Arabic language is a rich in multiple meanings (senses) in a lot of words and the tasking to find the appropriate meaning is a key demand in most of the AIR applications. This paper aims to improve the efficiency of the DAIR systems through using an algorithm belong to the swarm intelligence called Artificial Bee Colony (ABC) algorithm and to improve the result quality through using the query expansion. The MABC-SDAIR algorithm is the search approach used in this work. It aims to search the most re...

The goal of Ontology Learning from Text is to learn ontologies that represent domains or applications that change often. Manually learning and updating such ontologies is too expensive. This is the reason for the Ontology Learning discipline's emergence. The leading approach to Ontology Learning from Text is the Ontology Learning Layer Cake. This approach splits the task into four or five sequential tasks. Each of the tasks may use diverse methods, ranging from uses of Linguistic knowledge to Machine Learning. The authors review the shortcomings of the Ontology Learning Layer Cake approach and conclude that the approach is not viable for Ontology Learning from Text. They suggest alternative approaches that may help learning ontologies in an efficient, effective way.

У рефераті досліджується побудова інтелектуального агента на основі принципів
системного аналізу. Розглянуто основні властивості інтелектуальних агентів, такі як
адаптивність, раціональність, навчання та комунікація. Проаналізовано роль управління,
планування та прогнозування у функціонуванні агентів, а також їх вплив на ефективність і
стійкість систем. Особлива увага приділяється архітектурі інтелектуального агента, синтезу
його складових, вибору технологій та управлінню проєктом з його створення. Крім того,
наведене обґрунтування важливості впровадження даного проекту для наукової та
навчальної діяльності кафедри.

Há muito interesse na construção de ontologias, porém, há poucos trabalhos que apontam para uma utilização de ontologias em larga escala. Algumas das razões para isso são o tempo, custo e recursos associados ao desenvolvimento. Com o objetivo de reduzir estes esforços, métodos de aprendizagem de ontologias têm sido desenvolvidos. Para isso, é necessária a automatização do processo de aquisição de conhecimento e diversas abordagens têm sido sugeridas. A utilização de textos como fonte de aquisição de conhecimento parece ser um caminho interessante, visto que a linguagem é a primeira forma de transferência de conhecimento entre os seres humanos, além de haver uma diversidade de documentos digitais disponíveis. Para dar suporte à construção semiautomática de ontologias a partir de textos, várias ferramentas foram desenvolvidas, cada uma utilizando técnicas e métodos diferentes. Este artigo tem como objetivo apresentar algumas das ferramentas disponíveis, assim como suas características...

Finance ontology is, in most cases, manually addressed. This results in a tedious development process and error prone that delay their applicability. This is why there is a need of domain ontology learning methods that built the ontology automatically and without human intervention. However, in this learning process, the discovery of non-taxonomic relationships has been recognized as one of the most difficult problems. In this paper, we propose a new methodology for learning non-taxonomic relationships and building financial ontology from scratch. Our new technique is based on using and adapting Open Information Extraction algorithms to extract and label domain relations between concepts. To evaluate our new method effectiveness, we compare the extracted non-taxonomic relations of our algorithm with related works in the same finance corpus. The results showed that our system is more accurate and more effective.

The work which will be presented in this paper is related to the building of an ontology of domain for the Arabic linguistics. We propose an approach of automatic construction that is using statistical techniques to extract elements of ontology from Arabic texts. Among these techniques we use two; the first is the "repeated segment" to identify the relevant terms that denote the concepts associated with the domain and the second is the "co-occurrence" to link these new concepts extracted to the ontology by hierarchical or nonhierarchical relations. The processing is done on a corpus of Arabic texts formed and prepared in advance.

We investigate the application of distributional semantics models for facilitating unsupervised extraction of biomedical terms from unannotated corpora. Term extraction is used as the first step of an ontology learning process that aims to (semi-)automatic annotation of biomedical concepts and relations from more than 300K PubMed titles and abstracts. We experimented with both traditional distributional semantics methods such as Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA) as well as the neural language models CBOW and Skip-gram from Deep Learning. The evaluation conducted concentrates on sepsis, a major life-threatening condition, and shows that Deep Learning models outperform LSA and LDA with much higher precision.

We investigate the application of distributional semantics models for facilitating unsupervised extraction of biomedical terms from unannotated corpora. Term extraction is used as the first step of an ontology learning process that aims to (semi-)automatic annotation of biomedical concepts and relations from more than 300K PubMed titles and abstracts. We experimented with both traditional distributional semantics methods such as Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA) as well as the neural language models CBOW and Skip-gram from Deep Learning. The evaluation conducted concentrates on sepsis, a major life-threatening condition, and shows that Deep Learning models outperform LSA and LDA with much higher precision.

This paper outlines the methods of practical preautomatic and automatic verification of evolving lexicons and an ontology used to process natural language meaning, and several approaches that can be taken to speed up the process and decrease the cost of such verification. The main methodological direction of this type of verification is increased automation in conjunction with confirmation by native speakers with basic instructions rather than by fully trained linguist knowledge engineers. It is assumed that ontological and lexical verification based on mathematical logic has been completed prior to the outlined verification.

Identifying topics and concepts associated with a set of documents is a critical task for information retrieval systems. One approach is to associate a query with a set of topics selected from a fixed ontology or vocabulary of terms. The core idea of this research is using Wikipedia articles and associated pages to make a topic ontology for this purpose. The benefit of this method is that Wikipedia is an online free-content encyclopedia which is developed through a social process and kept current by the Wikipedia community. In this paper the Persian Wikipedia has been analyzed in accordance to its articles and the category link graphs to extract a Persian pseudo-ontology. Thereafter, the created ontology has been applied through a query expansion algorithm to improve the performance of an information retrieval system. Our experiments show that it is possible to improve the precision of the information retrieval system by queries expansion based on Wikipedia.

This paper proposes a semi-supervised relation acquisition method that does not rely on extraction patterns (e.g. "X causes Y" for causal relations) but instead learns a combination of indirect evidence for the target relation-semantic word classes and partial patterns. This method can extract long tail instances of semantic relations like causality from rare and complex expressions in a large Japanese Web corpus-in extreme cases, patterns that occur only once in the entire corpus. Such patterns are beyond the reach of current pattern based methods. We show that our method performs on par with state-of-the-art pattern based methods, and maintains a reasonable level of accuracy even for instances acquired from infrequent patterns. This ability to acquire long tail instances is crucial for risk management and innovation, where an exhaustive database of high-level semantic relations like causation is of vital importance.

The paper proposes the M-OBLIGE model for building multitutor ontology-based learning environments. The model is based on local ontologies, describing the domain of each individual tutor in the environment, and external ontologies, describing more general concepts. The ontologies are used by ontology processors to decide which tutors might benefit a student who needs to learn new concepts. The model allows domain expertise to be shared, and can be used as a framework for integrating multiple tutors on the web. We show how the model can be applied to tutors in the database domain. We also illustrate the process of developing the ontologies for an existing system.

This paper describes new thesaurus construc- tion method in which class-based, small size thesauruses are constructed and merged as a whole based on domain classification system. This method has advantages in that 1) taxonomy con- struction complexity is reduced, 2) each class- based thesaurus can be reused in other domain thesaurus, and 3) term distribution per classes in target domain

In this contribution we present a new methodology to compile large language resources for domain-specific taxonomy learning. We describe the necessary stages to deal with the rich morphology of an agglutinative language, i.e. Korean, and point out a second order machine learning algorithm to unveil term similarity from a given raw text corpus. The language resource compilation described is part of a fully automatic top-down approach to construct taxonomies, without involving the human efforts which are usually required.

In this paper we describe two approaches to integrating standalone information processing techniques into a semantic application capable of acquiring and maintaining knowledge. We distinguish between integration through aggregation and through choreographing, and argue that the latter is not only simpler to realize but also provides greater benefits. The benefits were, in our experiment, related to developing a capability of maintaining and validating knowledge through an integration of down-and upstream knowledge processing tools. We describe the principles of integration and relate them to pragmatic challenges for the semantic web and to strategic directions of its evolution.

Abstract—The focus of this paper will be showing how linguistic
information can be modeled in an ontological engineering
environment for knowledge management and acquisition, and on
this basis made accessible for hierarchical and axiomatic
processing. The simplicity of Relational Network Notation models
stratal linguistic information solely with reference to sets of
interconnecting nodes. Axioms can be effortlessly declared upon
the simplicity of the notation such that the knowledge base can be
easily extended with the power of inference. Fruitful new
knowledge can thus be acquired through axiomatic inference in
terms of uncovering latent links between concepts and/or
instances in the knowledge base. With this model, various
linguistic resources, WordNet and FrameNet originally encoding
different domains of linguistic knowledge, are now capable of
interfacing with each other, retrieving and generating underlying
linguistic information, serving as a more comprehensive NLP tool.

News portals, such as Yahoo News or Google News, collect large amounts of documents from a variety of sources on a daily basis. Only a small portion of these documents can be selected and displayed on the homepage. Thus, there is a strong preference for major, recent events. In this work, we propose a scalable and accurate First Story Detection (FSD) pipeline that identifies fresh news. In comparison to other FSD systems, our method relies on relation extraction methods exploiting entities and their relations. We evaluate our pipeline using two distinct datasets from Yahoo News and Google News. Experimental results demonstrate that our method improves over the state-of-the-art systems on both datasets with constant space and time requirements.

Abstract. We propose a general methodology to build up a domain ontology from one or more domain glossaries. The particular feature of this methodology is in the parallel construction of a domain ontology and a complete domain terminology. In this paper we fully describe the methodology phases and we apply them to a real-world example from the medical domain.

A critical prerequisite for human-level cognitive systems is having a rich conceptual understanding of the world. We describe a system that learns conceptual knowledge by deep understanding of WordNet glosses. While WordNet is often criticized for having a too fine-grained approach to word senses, the set of glosses do generally capture useful knowledge about the world and encode a substantial knowledge base about everyday concepts.. Unlike previous approaches that have built ontologies of atomic concepts from the provided WordNet hierarchies, we construct complex concepts compositionally using description logic and perform reasoning to derive the best classification of knowledge. We view this work as simultaneously accomplishing two goals: building a rich semantic lexicon useful for natural language processing, and building a knowledge base that encodes commonsense knowledge.

Semantic classification of words is a highly context sensitive and somewhat moving target, hard to deal with and even harder to evaluate on an objective basis. In this paper we suggest a step-wise methodology for automatic acquisition of lexico-semantic classes and delve into the non trivial issue of how results should be evaluated against a top-down reference standard.