Technology of training a neural-network model for ontology population

The paper considers one of the subtasks of ontology learning - the ontology population, which implies the extension of existing ontology by new instances without changing the structure of its classes and relations. A brief overview of existing ontology learning approaches is presented. A highly automated technology for ontology population based on training and application of the neural-network language model to identify and extract potential instances of ontology classes from domain texts is proposed. The main stages of its application, as well as the results of its experimental evaluation and the main directions of its further improvement are considered.

Review of Tools for Semantics Extraction: Application in Tsunami Research Domain

Immense numbers of textual documents are available in a digital form. Research activities are focused on methods of how to speed up their processing to avoid information overloading or to provide formal structures for the problem solving or decision making of intelligent agents. Ontology learning is one of the directions which contributes to all of these activities. The main aim of the ontology learning is to semi-automatically, or fully automatically, extract ontologies—formal structures able to express information or knowledge. The primary motivation behind this paper is to facilitate the processing of a large collection of papers focused on disaster management, especially on tsunami research, using the ontology learning. Various tools of ontology learning are mentioned in the literature at present. The main aim of the paper is to uncover these tools, i.e., to find out which of these tools can be practically used for ontology learning in the tsunami application domain. Specific criteria are predefined for their evaluation, with respect to the “Ontology learning layer cake”, which introduces the fundamental phases of ontology learning. ScienceDirect and Web of Science scientific databases are explored, and various solutions for semantics extraction are manually “mined” from the journal articles. ProgrammableWeb site is used for exploration of the tools, frameworks, or APIs applied for the same purpose. Statistics answer the question of which tools are mostly mentioned in these journal articles and on the website. These tools are then investigated more thoroughly, and conclusions about their usage are made with respect to the tsunami domain, for which the tools are tested. Results are not satisfactory because only a limited number of tools can be practically used for ontology learning at present.

Automatic Ontology-Based Model Evolution for Learning Changes in Dynamic Environments

Knowledge engineering relies on ontologies, since they provide formal descriptions of real-world knowledge. However, ontology development is still a nontrivial task. From the view of knowledge engineering, ontology learning is helpful in generating ontologies semi-automatically or automatically from scratch. It not only improves the efficiency of the ontology development process but also has been recognized as an interesting approach for extending preexisting ontologies with new knowledge discovered from heterogenous forms of input data. Driven by the great potential of ontology learning, we present an automatic ontology-based model evolution approach to account for highly dynamic environments at runtime. This approach can extend initial models expressed as ontologies to cope with rapid changes encountered in surrounding dynamic environments at runtime. The main contribution of our presented approach is that it analyzes heterogeneous semi-structured input data for learning an ontology, and it makes use of the learned ontology to extend an initial ontology-based model. Within this approach, we aim to automatically evolve an initial ontology-based model through the ontology learning approach. Therefore, this approach is illustrated using a proof-of-concept implementation that demonstrates the ontology-based model evolution at runtime. Finally, a threefold evaluation process of this approach is carried out to assess the quality of the evolved ontology-based models. First, we consider a feature-based evaluation for evaluating the structure and schema of the evolved models. Second, we adopt a criteria-based evaluation to assess the content of the evolved models. Finally, we perform an expert-based evaluation to assess an initial and evolved models’ coverage from an expert’s point of view. The experimental results reveal that the quality of the evolved models is relevant in considering the changes observed in the surrounding dynamic environments at runtime.

A Self-Aggregated Hierarchical Topic Model for Short Texts

With the growth of the internet, short texts such as tweets from Twitter, news titles from the RSS, or comments from Amazon have become very prevalent. Many tasks need to retrieve information hidden from the content of short texts. So ontology learning methods are proposed for retrieving structured information. Topic hierarchy is a typical ontology that consists of concepts and taxonomy relations between concepts. Current hierarchical topic models are not specially designed for short texts. These methods use word co-occurrence to construct concepts and general-special word relations to construct taxonomy topics. But in short texts, word cooccurrence is sparse and lacking general-special word relations. To overcome this two problems and provide an interpretable result, we designed a hierarchical topic model which aggregates short texts into long documents and constructing topics and relations. Because long documents add additional semantic information, our model can avoid the sparsity of word cooccurrence. In experiments, we measured the quality of concepts by topic coherence metric on four real-world short texts corpus. The result showed that our topic hierarchy is more interpretable than other methods.

A Domain-Independent Ontology Learning Method Based on Transfer Learning

Ontology plays a critical role in knowledge engineering and knowledge graphs (KGs). However, building ontology is still a nontrivial task. Ontology learning aims at generating domain ontologies from various kinds of resources by natural language processing and machine learning techniques. One major challenge of ontology learning is reducing labeling work for new domains. This paper proposes an ontology learning method based on transfer learning, namely TF-Mnt, which aims at learning knowledge from new domains that have limited labeled data. This paper selects Web data as the learning source and defines various features, which utilizes abundant textual information and heterogeneous semi-structured information. Then, a new transfer learning model TF-Mnt is proposed, and the parameters’ estimation is also addressed. Although there exist distribution differences of features between two domains, TF-Mnt can measure the relevance by calculating the correlation coefficient. Moreover, TF-Mnt can efficiently transfer knowledge from the source domain to the target domain and avoid negative transfer. Experiments in real-world datasets show that TF-Mnt achieves promising learning performance for new domains despite the small number of labels in it, by learning knowledge from a proper existing domain which can be automatically selected.