Never-Ending Language Learning Research Articles

Learning Bayesian Networks from the Knowledge of a Never-Ending Learning System

Never-Ending Learning (NEL) is a new paradigm of machine learning in which computer systems can learn continuously and incrementally. The first never-ending learning system reported in literature is named NELL (Never-Ending Language Learning). NELL has built a Knowledge Base (KB) containing many types of knowledge. In this paper, we propose to learn Bayesian networks from the knowledge stored in the NELLs KB aiming to assist in the development of expert systems in the future. In addition, Bayesian networks have shown to be promising to solve the problem of representing semantic relations and extending the NELLs initial ontology. In the initial experiments, we have built a dataset on the domain of diseases from relations existing in NELL and applied two learning algorithms of Bayesian networks named K2 and DMBC. The empirical results have shown that this proposal is promising.

Augmenting Linked Data Ontologies with New Object Properties

Although several RDF knowledge bases are available through the Linked Open Data (LOD) initiative, the ontology schema of such linked datasets is not very rich. In particular, they lack object properties. The problem of finding new object properties between any two given classes has not been investigated in detail in the context of Linked Data. In the first part of this paper, we present DARO (Detecting Arbitrary Relations for enriching Ontology of Linked Data)—an unsupervised solution to enrich the LOD cloud with new object properties (and their instances) between two given classes. DARO first identifies text patterns from the web corpus that can potentially represent relations between individuals. These text patterns are then clustered based on semantic similarity to capture the object properties between the two given classes. We have empirically evaluated our approach on several pairs of classes and found that the system can indeed be used for enriching the linked datasets with new object properties and their instances. We have compared DARO with newOntExt which is an offshoot of the NELL (Never-Ending Language Learning) effort. Our experiments reveal that DARO gives better results than newOntExt as a recall-oriented system. In the second part of the paper, we propose a methodology to predict potential pairs of classes which could be connected by object properties but are not yet connected. We claim that evidence obtained from external textual resources and their Word2Vec representations can be made use of, for this purpose. Our approach gives results that are complementary to those given by the traditional techniques found in the literature. Hence our method can be used in combination with the traditional techniques for maximum benefits.

Coupling for Coreference Resolution in a Never-ending Learning System

The Never-Ending Language Learning (NELL) is a system that attempts learning to learn from the Web every day, in an autonomously way. Keeping a high precision is the key to keep the NELL’s learning active and better day-by-day. One of the challenges for NELL system is to properly identify different noun phrases that denote the same concept in order to maintain the cohesion of the knowledge base. This article investigates the coupling as an approach for improving coreference resolution on NELL. For that, several coupled algorithms considering semantic and morphologic features were compared with results previously obtained with no use of coupling. The results presented in this article confirm empirically that coupling strategy is a useful and good approach to achieve better coverage and accuracy in NELL’s knowledge base.

Human and Machine Cultures of Reading: A Cognitive-Assemblage Approach

The concept “cultures of reading” should be expanded to include machines that read. Machine reading is exemplified by the computer system called Never-Ending Language Learning (NELL). Because NELL lacks real-world experience, its semantic comprehension is limited to forming categories of words. This process illustrates a major difference between human and machine reading: whereas human reading involves causal reasoning, machine reading relies more on correlations. Human-machine hybrid reading, for example the kind done with an e-book, can be understood as a cognitive assemblage through which information, interpretations, and meanings circulate. The introduction of mechanical cognition into printing can be seen in the Paige Compositor, from 1878. The transition from electromechanical cognition to more flexible digital and electronic computational media marks the movement from print, understood as a technology involving the arrangement of type pieces to impress ink on paper, to postprint, in which inked products originate as computer files. This change, which signals an ontological rupture in writing and reading practices, is addressed through a cognitive-assemblage approach emphasizing the distribution of cognition among technical and human actors.