POLIBITS, Vol. 60, pp. 35-41, 2019.
Abstract: NER, n-grams, text representation, graph-based representation, named entity recognitionen|ok 06 TeX - 8M116 - MICAI_villaverde/MICAI.pdf|Spain|A Comparison of Adaptive and Non-Adaptive Ensemble Methods for Classification Systems|Mónica Villaverde, David Aledo, David Pérez, Félix Moreno|Intelligent systems are increasingly common in many sensor applications, specially nowadays due to the growing of the Internet of Things (IoT). Communication among sensors is important to collect information, but also could be a useful mechanism to enhance the system reliability, since sensors can build an ensemble to improve the system behavior. Moreover, adaptive capabilities are also critical when the ensemble is working in a changing environment conditions. In this work, a two-stage architecture is applied. The first stage is in charge of making a classification from partial information of the entire target, or from different points of view, using an Artificial Neural Network (ANN) as a classifier. Additionally, the second stage aggregates all the estimations given by the ensemble in order to obtain the final classification. In this work four different ensembles methods for the second stage are analyzed: (i) ANN, (ii) plurality majority, (iii) basic weighted majority (iv) and stochastic weighted majority. The majority voting algorithm and the ANN are compared against our two proposed adaptive algorithms which are based on weighted majority and do not require previous training. This comparison not only takes into account the accuracy of each algorithm but also adaptation. Results show that ANN is the most accurate proposal whereas the most innovate proposed stochastic weighted voting is the most adaptive one.|Ensemble methods, wireless sensor network, artificial neural network, voting algorithms
Keywords: The identification of indirect relationships between texts from different sources makes the task of text mining useful when the goal is to obtain the most valuable information from a set of texts. That is why in the field of information retrieval the correct recognition of named entities plays an important role when extracting valuable information in large amounts of text. Therefore, it is important to propose techniques that improve the NER classifiers in order to achieve the correct recognition of named entities. In this work, a graph structure for storage and enrichment of named entities is proposed. It makes use of synonyms and domain-specific ontologies in the area of computing. The performance of the proposed structure is measured and compared with other NER classifiers in the experiments carried out.
PDF: Named Entity Recognition Based on a Graph Structure
PDF: Named Entity Recognition Based on a Graph Structure
https://doi.org/10.17562/PB-60-5
Table of contents of POLIBITS 60