keyboard_arrow_up
Accepted Papers
An Explainable Template-based Question Answering Approach Over Knowledge Graphs

Enayat Rajabi1, Geethu Susan Ebby2 and Somayeh Kafaie3, 1, 2Cape Breton University, Sydney, NS, Canada, 3Saint Mary’s University, Halifax, NS, Canada

ABSTRACT

With the emergence of knowledge graph technologies, the Semantic Web community encourages users to leverage the knowledge graphs and their underlying reasoning power. Question-answering systems can be built over the knowledge graphs and provide end-users an environment to ask questions in natural language instead of SPARQL or Cypher queries. This work adopts a template-based strategy to match the natural language to Neo4j Cypher queries by setting it apart from conventional automated translation practices. In this approach, we use a language model (BERT) to identify the context of questions and then match it with a set of templates created by domain experts. It also uses a knowledge graph API (Neo4j) to pose the query over a disease knowledge graph and visualize the results to users with an explanation.

KEYWORDS

Knowledge graph, Question answering, Explainability, Neo4j, Disease database.


Multi-cloud 6g Networks Orchestration-based Throughput Maximization on Uav-assisted Wireless Sensor Networks

Betalo Mesfin Leranso, Supeng Leng, and Wondemu Woyo Ganebo, University of Electronic Science and Technology of China (UESTC), China

ABSTRACT

The management and coordination of unmanned aerial vehicles (UAVs) in wireless sensor networks (WSNs) is achieved through the use of multiple cloud computing platforms. To achieve redundancy, scalability, and efficient resource allocation, multiple cloud providers are used and the potential of 6G networks is exploited to enable high-speed, low-latency communication between UAVs and WSN nodes. However, UAVs have limited energy and poor quality of service (QoS), consume a large amount of energy, and cause network latency problems when collecting data from multiple WSNs. This paper considers a multi-agent distributed deep deterministic policy gradient (MAD3PG)-enabled UAV-supported WSN for agricultural monitoring in rural areas. The UAVs are equipped with energy harvesting (EH) techniques to provide sustainable power to the UAVs and SNs. In particular, we formulate the data collection from the SNs and the EH from the environment as joint optimization problems to maximize the energy-efficient data throughput (EEDT) network performance and quality of service (QoS) in a large-scale WSN. To solve the dimensional constraints and complex optimization problems, we propose a MAD3PG algorithm to find the optimal data collection and EH decisions using multiple UAVs’ sensor scheduling techniques. The simulation results show that our proposed scheme outperforms the benchmark schemes in terms of the average number of UAVs, offloading capacity, and improving system reliability.

KEYWORDS

Unmanned aerial vehicles, multi-cloud 6G network orchestration, deep reinforcement learning, and data throughput.


Real-time Monitoring and Visualisation of the Water Quality of Rivers in Ghana Using Wireless Sensor Networks (Wsns)

Martin Adane1 and Laurene Adjei2, 1Department of Mathematics and ICT Education, University of Cape Coast, Cape Coast, Ghana, 2Ho Technical University, Ho, Ghana

ABSTRACT

There has been a public concern and outcry over the impact of artisanal mining on rivers in Ghana which serve as a constant source of water supply for domestic, industrial, and irrigation purposes. Monitoring the quality of water in the affected river basins is mainly carried out using traditional sampling techniques which do pose a significant challenge to water regulatory bodies. The water quality parameters that need to be monitored are spatiotemporal in nature, hence, the need for the use of a more efficient approach that will perform real-time water quality monitoring is relevant. The focus of this work is to use existing opensource tools, to demonstrate how the visualisation capabilities of commercially available Wireless Sensor Networks (WSNs) may be extended allowing them to be embedded in custom-made applications. The data collected is analysed and presented in real-time thus providing timely information to aid in efficient analysis, decision-making, and planning.

KEYWORDS

Artisanal mining, spatiotemporal, real-time water quality monitoring, visualisation, wireless sensors network.


Combining Discrete Wavelet and Cosine Transforms for Efficient Sentence Embedding

Rana Salama1, 3, Abdou Youssef1, and Mona Diab2, 1School of Engineering and Applied Science, George Washington University, 2Language Technologies Institute, Carnegie Mellon University, 3Faculty of Computers and Artificial Intelligence, Cairo University

ABSTRACT

Wavelets have emerged as a cutting edge technology in a number of fields. Concrete results of their application in Image and Signal processing suggest that wavelets can be effectively applied to Natural Language Processing (NLP) tasks that capture a variety of linguistic properties. In this paper, we leverage the power of applying Discrete Wavelet Transforms (DWT) to word and sentence embeddings. We first evaluate, intrinsically and extrinsically, how wavelets can effectively be used to consolidate important information in a word vector while reducing its dimensionality. We further combine DWT with Discrete Cosine Transform (DCT) to propose a non-parameterized model that compresses a sentence with a dense amount of information in a fixed size vector based on locally varying word features. We show the efficacy of the proposed paradigm on downstream applications models yielding comparable and even superior (in some tasks) results to original embeddings.

KEYWORDS

Natural Language Processing, Sentence Embedding, Discrete Wavelets Transform, Discrete Cosine Transform.


Summarizing Arabic Articles Using Large Language Models

Bader Alshemaimri, Ibrahim Alrayes, Turki Alothman, Fahad Almalik, Mohammed Almotlaq, Department of Software Engineering, King Saud University, Saudi Arabia

ABSTRACT

This paper explores abstractive and extractive Arabic text summarization using AI, employing finetuning and unsupervised machine learning techniques. We investigate the adaptation of pre-trained language models such as AraT5 through fine-tuning. Additionally, we explore unsupervised methods leveraging unlabeled Arabic text for generating concise and coherent summaries by utilizing different vectorizers and algorithms. The proposed models are rigorously evaluated using text-centric metrics like ROUGE Lin (2004). The research contributes to the development of robust Arabic summarization systems, offering culturally sensitive and contextually aware solutions. By bridging the gap between advanced AI techniques and Arabic language processing, this work fosters scalable and effective summarization in the Arabic domain.

KEYWORDS

Arabic Text Summarization, Abstractive Summarization, Extractive Summarization, Natural Language Processing (NLP), Fine-Tuning Language Models, Unsupervised Machine Learning, Vectorization Techniques, Pre-trained Language Models, Language Model Adaptation.


Universal Dependencies Annotation of Old English Texts With Spacy and Mobilebert. Evaluation and Perspectives

Javier Martín Arista1, Ana Elvira Ojanguren López1, Sara Domínguez Barragán1 and Mario Barcala Rodríguez2, 1Department of Modern Languages, University of La Rioja, Spain, 2NLPgo, Santiago de Compostela, Spain

ABSTRACT

The aim of this paper is to assess three training corpora of Old English and three configurations and training procedures as to the performance of the task of automatic annotation of Universal Dependencies (UD, Nivre et al., 2016 [1]). The method was aimed to deciding to what extent the size of the corpus improved results and which configuration turned out the best metrics. The training methods included a pipeline with default configuration, pre-training of tok2vec step and a model of language based on transformers. For all training methods, three training corpora with different sizes are tested: 1,000, 5,000, 10,000, and 20,000 words. The training and the evaluation corpora are based on ParCorOEv2 (Martín Arista et al., 2021 [2]). The results can be summarised as follows. The larger training corpora result in improved performance in all the stages of the pipeline, especially in POS tagging and dependency parsing. Pre-training the tok2vec stage yields better results than the default pipeline. It can be concluded that the performance could improve with more training data or by fine-tuning the models. However, even with the limited training data selected for this study, satisfactory results have been obtained for the task of automatically annotating Old English with UD.

KEYWORDS

Natural Language Processing, Universal Dependencies, syntactic annotation, NPL library, Transformers.


Amosl: Adaptive Modality-wise Structure Learning in Multi-view Graph Neural Networks for Enhanced Unified Representation

Peiyu Liang, Hongchang Gao, and Xubin He, Computer and Information Science, Temple University, Philadelphia, PA, USA, 19122

ABSTRACT

While MVGNNs excel at leveraging diverse modalities for learning object representation, existing methods assume identical local topology structures across modalities that overlook real-world discrepancies. This leads MVGNNs straggles in modality fusion and representations denoising. To address these issues, we propose adaptive modality-wise structure learning (AMoSL). AMoSL captures node correspondences between modalities via optimal transport, and jointly learning with graph embedding. To enable efficient end-to-end training, we employ an efficient solution for the resulting complex bilevel optimization problem. Furthermore, AMoSL adapts to downstream tasks through unsupervised learning on inter-modality distances. The effectiveness of AMoSL is demonstrated by its ability to train more accurate graph classifiers on six benchmark datasets.

KEYWORDS

Multi-view Graph Neural Network, Graph Classification, Graph Mining, Optimal.


menu
Reach Us

emailicaita@icaita2024.org


emailicaitaconf@yahoo.com

close