Online product reviews (OPR) are a commonly used medium for consumers to communicate their experiences with products during online shopping. Previous studies have investigated the helpfulness of OPRs using frequency-based, linguistic, meta-data, readability, and reviewer attributes. In this study, we explored the impact of robust contextual word embeddings, topic, and language models in predicting the helpfulness of OPRs. In addition, the wrapper-based feature selection technique is employed to select effective subsets from each type of features. Five feature generation techniques including word2vec, FastText, Global Vectors for Word Representation (GloVe), Latent Dirichlet Allocation (LDA), and Embeddings from Language Models (ELMo), were employed. The proposed framework is evaluated on two Amazon datasets (Video games and Health & personal care). The results showed that the ELMo model outperformed the six standard baselines, including the fine-tuned Bidirectional Encoder Representations from Transformers (BERT) model. In addition, ELMo achieved Mean Square Error (MSE) of 0.0887 and 0.0786 respectively on two datasets and MSE of 0.0791 and 0.0708 with the wrapper method. This results in the reduction of 1.43% and 1.63% in MSE as compared to the fine-tuned BERT model on respective datasets. However, the LDA model has a comparable performance with the fine-tuned BERT model but outperforms the other five baselines. The proposed framework demonstrated good generalization abilities by uncovering important factors of product reviews and can be evaluated on other voting platforms.

Knowledge Graphs is one of the most popular techniques for knowledge-based modelling in various subdomains of modern AI technologies ranging from natural language processing to e-commerce recommendations and cyberphysical systems. Even complex technical systems like telecommunication networks could be modelled by means of Knowledge Graphs. However, there are serious challenges when we deal with such systems having a huge number of interconnected elements (e.g. technical objects and their groups) that change over time. Thus, up-to-date there is no adequate solution for not only telecommunication networks but for any complex dynamic systems where inductive and deductive synthesis of large Knowledge Graph based models that are easily reconfigurable and scalable is required. We state and solve the problem of building such models for one of the most common types of objects where models can be represented as hierarchical re-configurable structures. This representation enables recent advances in multilevel inductive–deductive synthesis for model building. From a methodological viewpoint, we propose a novel complex approach to multilevel synthesis for objects with dynamic hierarchical structure based on modified methods for inductive and deductive synthesis of Knowledge Graphs. From a practical perspective we present a real case-study on an interactive service for digital cable TV networks – which is especially interesting for data engineers and scientists – where various problems ranging from network health monitoring to channel advertising can be solved with the same hierarchical model. We release an openly available domain benchmark, which features two realistic datasets (namely, for SPARQL querying performance analysis, and for our case study on dynamic network monitoring). Last but not least, our experiments with recent state-of-the-art approaches to knowledge graph querying Abdelaziz et al. (2017) show that the developed models of multilevel synthesis reduce the time complexity up to 73% on practice compared to the baselines, and are lossless and able to beat their competitors based on parallel knowledge graph processing from 4% to 91% in terms of computational time (depending on the query type). Further parallelisation of our multilevel models is even more efficient (the reduction of query processing time is about 40%–45%) and opens promising prospects for the creation and exploitation of dynamic Knowledge Graphs in practice.