Search

There is little evidence on the right approach on how to delineate the sub-networks in a labour market. The subject of research in this paper is computational influence identification of the labour force transitions between different professional occupations in the Slovenian labour network from 2010 to 2020. We use community detection algorithm to identify occupation groups and apply influence analysis on the Slovenian labour network from 2010 to 2020. This directly supports the decision-makers and employment services in identifying job opportunities for job-seekers based. The main contribution is using influence analysis to detect occupations and communities that had the most significant impact on the Slovenian labour market. The research is the first work to successfully apply community and influence analysis in the Slovenian labour network to the best of our knowledge. The paper carries several important implications, primarily highlighting the usage of existing data to increase employment levels.

Managing employees in the organization is a complex task, and it produces significant operational expenses, especially when an organization has strict business and security policies of a periodical employees’ (re)deployment within the organization. Every activity in that process has its cost and ownership. Those activities are often performed manually and are triggered by certain internal communications protocols. Using a real case study of an actual large organization, we conducted several business processes (BP) simulations and analyzed long-lasting processes that are expensive and less effective. Here we propose business process improvements (BPI) in the human resource (HR) and IT domain by (1) optimizing the organization's BP, (2) integrating a decision support system (DSS) with human resources management (HRM) software and identity and access management (IDAM) software. The proposed integration reduces the time and costs of bringing an adequate employee into service. This article describes the improvement process of managing employees in a big organization, which is documented, simulated, and analyzed through three improvement stages to decrease organizational expenses and increase organizational income, image, and concurrency. The simulation results of the improved BP show that the organization’s management structure may easily estimate investments in the IT integrations of DSS, HRM, and IDAM software to improve business efficiency and raising of an organization’s income. Integration of the DEX model in the DSS platform, and integration of DSS with HRM and IDAM software into IS results in 82,74% in time reduction, 87,64% of operational expenses savings, and 82,69% compensation of the projected loss of income.

Human behaviour in various circumstances mirrors the corresponding brain connectivity patterns, which are suitably represented by functional brain networks. While the objective analysis of these networks by graph theory tools deepened our understanding of brain functions, the multi-brain structures and connections underlying human social behaviour remain largely unexplored. In this study, we analyse the aggregate graph that maps coordination of EEG signals previously recorded during spoken communications in two groups of six listeners and two speakers. Applying an innovative approach based on the algebraic topology of graphs, we analyse higher-order topological complexes consisting of mutually interwoven cliques of a high order to which the identified functional connections organise. Our results reveal that the topological quantifiers provide new suitable measures for differences in the brain activity patterns and inter-brain synchronisation between speakers and listeners. Moreover, the higher topological complexity correlates with the listener’s concentration to the story, confirmed by self-rating, and closeness to the speaker’s brain activity pattern, which is measured by network-to-network distance. The connectivity structures of the frontal and parietal lobe consistently constitute distinct clusters, which extend across the listener’s group. Formally, the topology quantifiers of the multi-brain communities exceed the sum of those of the participating individuals and also reflect the listener’s rated attributes of the speaker and the narrated subject. In the broader context, the presented study exposes the relevance of higher topological structures (besides standard graph measures) for characterising functional brain networks under different stimuli.

The growing amount of text electronically available has placed text type classification among the most exciting issues in the field of exploratory data mining. This talk presents an preliminary approach to text type classification by features of linguistic co-occurrence networks. Text can be represented as a complex network of linked words: each individual word is a node and interactions amongst words are links. The aim of our work-in-progress presented in this talk is to investigate the idea of replacing the standard natural language processing feature sets with linguistic network measures for the purpose of text type classification. This talk tackles the problem of binary classification of two different text types. Our dataset is consisted of 150 equal-sized Croatian texts divided in two classes: 75 literature texts and 75 blog texts. Literature texts represent segments from 7 different books written in or translated to Croatian language, while blog texts are collected from two very popular Croatian blogs. The trait which prompted us to do the classification of this particular text types is the linguistic distinction between book and blog. We constructed 150 different co-occurrence networks (one for each text in the dataset), all weighted and directed. Words are nodes linked if they are co- occurring as neighbors to each other in a sentence. The weight of the link is proportional to the overall co-occurrence frequencies of the corresponding word pairs within a text. For each network we computed a set of 10 measures (number of components, average degree, average path length, clustering coefficient, transitivity, degree assortativity, density, reciprocity, average in-selectivity, average out-selectivity), which are used as feature set for classification. All features are rescaled to [0 − 1] in order to make them independent of each other. We preformed a series of classification experiments using various types of classification algorithms and methods (support vector machine, classification trees, Naive Bayes, k-nearest neighbor, LDA, QDA). The performance of each classifier was evaluated with corresponding methods, such as misclassification error measures, confusion matrices and receiver operating characteristic curves. All classification experiments show very good classification accuracy, while the average in- and out- selectivity measures act as the most useful features in predicting the correct text type and reducing the misclassification rate. Precision and recall measures and ROC curves indicate that the node selectivity measures are the only measures from the feature set that can capture the structural differences between two classes of networks.

In this paper we analyse network motifs in the co-occurrence directed networks constructed from five different texts (four books and one portal) in the Croatian language. After preparing the data and network construction, we perform the network motif analysis. We analyse the motif frequencies and Z-scores in the five networks. We present the triad significance profile for five datasets. Furthermore, we compare our results with the existing results for the linguistic networks. Firstly, we show that the triad significance profile for the Croatian language is very similar with the other languages and all the networks belong to the same family of networks. However, there are certain differences between the Croatian language and other analysed languages. We conclude that this is due to the free word-order of the Croatian language.

Viewed as a unique, biologically-based human faculty, language has been recognized as the reflection of the human cognitive capacities, both in terms of its structure and its computational characteristics. Studying languages at intra- and cross-linguistic levels is of paramount importance in relation to our biological, cultural, historical and social beings. Hence, human languages, besides still being our main tools of communication, reflect our history and culture. Language can be seen as a complex adaptive system, evolving in parallel with our society. The analysis and understanding of language and its complexity is of crucial interest in the modern globalizing world. Various aspects of natural language systems can be represented as complex networks, whose vertices depict linguistic units, while links/edges model their morphosyntactic, semantic and/or pragmatic interactions. Thus the language network can be constructed at various linguistic levels: syntactic, semantic, phonetic, syllabic ect. So far there have been efforts to model the phenomena of various language subsystems and examine their unique function through complex networks. Still, the present endeavors in linguistic network research focus on isolated linguistic subsystems lacking to explain (or even explore) the mechanism of their mutual interaction, interplay or inheritance. Obtaining such findings is critical for deepening our understanding of conceptual universalities in natural languages, especially to shed light on the cognitive representation of the language in the human brain. Therefore, one of the main open questions in linguistic networks is explaining how different language subsystems mutually interact. The complexity of any natural language is contained in the interplay among several language levels. Below the word-level, it is possible to explore the type of phonology, morphology and syllabic subsystem complexity. The phonology subsystem complexity is reflected in the morphology subsystem complexity. On the word- level, the morphology subsystem complexity reflects in the complexity of the word order, syntactic rules and the ambiguity of lexis. Since the word order can be considered as the primary factor (but not the only) that determines linguistic structure, it is important to explore the subsystems’ interactions by which it is influenced. The overview of language networks will be revealed by means of the Croatian by showing the properties of five language subsystems: co-occurrence, syntax and shuffled layers (on the word level) and syllabic and grapheme layer (on the sub-word level).

Classifier comprehensibility is a decisive factor for practical classifier applications ; however it is ill-defined and hence difficult to measure. Most algorithms use comprehensibility metrics based on classifier complexity – such as the number of leaves in a classification tree – despite evidence that they do not correspond to comprehensibility well. A classifier comprehensibility survey was therefore designed in order to derive exhaustive comprehensibility metrics better reflecting the human sense of classifier comprehensibility. This paper presents an implementation of a classification-tree comprehensibility survey based on the suggested design and empirically verifies the assumptions on which the survey design is based: the chosen respondent performance metrics measured while solving the chosen tasks can be used to indirectly but objectively measure the influence of chosen tree properties on their comprehensibility.

In this paper we analyse the selectivity measure calculated from the complex network in the task of the automatic keyword extraction. Texts, collected from different web sources (portals, forums), are represented as directed and weighted co-occurrence complex networks of words. Words are nodes and links are established between two nodes if they are directly co-occurring within the sentence. We test different centrality measures for ranking nodes - keyword candidates. The promising results are achieved using the selectivity measure. Then we propose an approach which enables extracting word pairs according to the values of the in/out selectivity and weight measures combined with filtering., 10 pages

In this paper, we introduce selectivity-based keyword extraction as a new unsupervised method for graph-based keyword extraction. Node selectivity measure is defined as the average weight distribution on the links of a single node and used in procedure of keyword candidate extraction. In particular, we propose extracting three word long keyword sequence and proving that the obtained results compare favourably with previously published results. Experiments were conducted on Croatian news articles dataset with keywords annotated by human experts. The selectivity-based keyword extraction method achieved the average F2 score of 25.32% on isolated documents and F2 score of 42.07% on a document collection. Proposed method is derived solely from statistical and structural information, which are reflected in the topological properties of text network. Furthermore, comparative results indicate that our simple graph-based method provides results that are comparable with more complex supervised and unsupervised methods, as well as with human annotators.