Hybrid Query Expansion Model Based on Pseudo Relevance Feedback and Semantic Tree for Arabic IR

In this paper, the authors propose and readapt a new concept-based approach of query expansion in the context of Arabic information retrieval. The purpose is to represent the query by a set of weighted concepts in order to identify better the user's information need. Firstly, concepts are extracted from the initially retrieved documents by the Pseudo-Relevance Feedback method, and then they are integrated into a semantic weighted tree in order to detect more information contained in the related concepts connected by semantic relations to the primary concepts. The authors use the “Arabic WordNet” as a resource to extract, disambiguate concepts and build the semantic tree. Experimental results demonstrate that measure of MAP (Mean Average Precision) is about 10% of improvement using the open source Lucene as IR System on a collection formed from the Arabic BBC news.

Improving Text-to-Code Generation with Features of Code Graph on GPT-2

Code generation, as a very hot application area of deep learning models for text, consists of two different fields: code-to-code and text-to-code. A recent approach, GraphCodeBERT uses code graph, which is called data flow, and showed good performance improvement. The base model architecture of it is bidirectional encoder representations from transformers (BERT), which uses the encoder part of a transformer. On the other hand, generative pre-trained transformer (GPT)—another multiple transformer architecture—uses the decoder part and shows great performance in the multilayer perceptron model. In this study, we investigate the improvement of code graphs with several variances on GPT-2 to refer to the abstract semantic tree used to collect the features of variables in the code. Here, we mainly focus on GPT-2 with additional features of code graphs that allow the model to learn the effect of the data stream. The experimental phase is divided into two parts: fine-tuning of the existing GPT-2 model, and pre-training from scratch using code data. When we pre-train a new model from scratch, the model produces an outperformed result compared with using the code graph with enough data.

An Information-Centric Semantic Data Collection Tree for Wireless Sensor Networks

Data collection is an important application of wireless sensor networks (WSNs) and Internet of Things (IoT). Current routing and addressing operations in WSNs are based on IP addresses, while data collection and data queries are normally information-centric. The current IP-based approach incurs significant management overheads and is inefficient for semantic data collection and queries. To address the above issue, this paper proposes a semantic data collection tree (sDCT) construction scheme to build up a semantic data collection tree for wireless sensor networks. The semantic tree is rooted at the edge/sink and supports data collection tasks, queries, and configurations efficiently. We implement the sDCT in Contiki and evaluate the performance of the sDCT in comparison with the state-of-the-art scheme, 6LoWPAN/RPL and L2RMR, using telosb sensors under various scenarios. The obtained results show that the sDCT achieves a significant improvement in terms of the energy efficiency and the packet transmissions required for data collection or a query task compared to 6LoWPAN/RPL and L2RMR.

Method of information obtaining from ontology on the basis of a natural language phrase analysis

A method for phrases analyzing in natural languages of inflective type (Ukrainian and Russian) has been developed. The method allows one to outline main expressed ideas and groups of words in the text by which they are stated. The semantic trees of propositions formed in this way, each of which expresses one specific idea, are a convenient source material for constructing queries to the ontology in the SPARQL language. The analysis algorithm is based on the following sequence of basic steps: word tokenize, determining of marker words and phrases, identifying available type of proposition, identifying nouns groups, building a syntactic graph of a sentence, building semantic trees of propositions based on existing types of propositions, substituting parameters from semantic trees of propositions in the corresponding SPARQL query templates. The choice of an appropriate template depends on the type of proposition expressed by a given semantic tree of a proposition. The sets of concepts received as an answer are tied as corresponding answers to the previously defined semantic tree of proposition. In case of non-receipt of information from the ontology, the reduction of noun groups is carried out to express more general concepts and the building queries using them. This allows us to get some answer, although not as accurate as when we use the full noun group. The use of SPARQL query templates requires an a priori known ontology structure, which is also proposed in this paper. Such a system is applicable for dialogue using chat-bots or for automatically receiving answers to questions from the text.

An Efficient Mining for Recommendation System for Academics

At present time huge numbers of research articles are available on World Wide Web in any domain. The research scholar explores a research papers to get the appropriate information and it takes time and effort of the researcher. In this scenario, there is the need for a researcher to search a research based on its research article. In the present paper a method of Knowledge ablation from a collection of research articles, is presented to evolve a system research paper recommendation system (RPRS), which would generate the recommendations for research article based on researcher choice. The RPRS accumulate the knowledge ablated from the pertinent research articles in the form of semantic tree. It accumulates all the literal sub parts with their reckoning in nodes. These parts are arranged based on their types in such a way that the leaf nodes stores the words with its prospect, the higher layer gives details about dictum with its reckoning, next to it an abstract. A Bayesian network is applied to construct a verisimilitude model which would quotation the pertinent tidings from the knowledge tree to construct the recommendation and word would be scored through TF-IDF value