Digital Library[ Search Result ]
Knowledge Completion System using Neuro-Symbolic-based Rule Induction and Inference Engine
Won-Chul Shin, Hyun-Kyu Park, Young-Tack Park
http://doi.org/10.5626/JOK.2021.48.11.1202
Recently, there have been several studies on knowledge completion methods aimed to solve the incomplete knowledge graphs problem. Methods such as Neural Theorem Prover (NTP), which combines the advantages of deep learning methods and logic systems, have performed well over existing methods. However, NTP faces challenges in processing large-scale knowledge graphs because all the triples of the knowledge graph are involved in the computation to obtain prediction results for one input. In this paper, we propose an integrated system of deep learning and logic inference methods that can learn vector representations of symbols from improved models of computational complexity of NTP to rule induction, and perform knowledge inference from induced rules using inference engines. In this paper, for rule-induction performance verification of the rule generation model, we compared test data inference ability with NTP using induced rules on Nations, Kinship, and UMLS data set. Experiments with Kdata and WiseKB knowledge inference through inference engines resulted in a 30% increase in Kdata and a 95% increase in WiseKB compared to the knowledge graphs used in experiments.
A Knowledge Completion Approach using Rule Generation based on Neuro-Symbolic Method
Jea-Seung Roh, Won-Chul Shin, Hyun-Kyu Park, Young-Tack Park
http://doi.org/10.5626/JOK.2021.48.4.425
A knowledge graph is a structured representation of real-world knowledge and is designed by collecting information from various sources. These knowledge graphs are networks that represent relationships between data and are applied in various fields of artificial intelligence; however, there exists problems related to incomplete knowledge due to the omission of entities or omission links between the entities. To resolve the problems, research on automatic knowledge completion techniques is necessitated. Consequently, various studies have been examined including embedding techniques, deep learning or symbolic rule inference using ontology. Although automatic knowledge completion can be efficiently performed through the above-mentioned methods, deep learning methods require a large amount of learning data due to data-driven processing methods, and there exist problems related to the results that are hard to explain. Futhermore, ontology-based methods require ontology and rules that are defined by the experts. To overcome this limitation, in this study, we propose an automatic knowledge completion method by explicitly extracting the implicit rules from the data based on the Neuro-Symbolic method. For rule extraction, we have implemented a symbolic unification based embedding path and defined a cost function for it to automatically generate the rules. Compared with the approaches presented in previous embedding studies, the proposed method demonstrates the superiority of the Neuro-Symbolic method concerning speed and performance. To assess the performance of the proposed method, for datasets like Nations, UMLS, and Kinship, experiments were conducted in comparison with the approach of the state-of-the-art knowledge completion studies. Consequently, an immense reduction in the training time and 37.5%p increase in the average performance were observed.
An Approach for Recognition of Elderly Living Patterns Based on Event Calculus Using Percept Sequence
Hyun-Kyu Park, Young-Tack Park
http://doi.org/10.5626/JOK.2019.46.11.1149
This paper proposes a method for recognizing the intentions of human activity based on percept sequences that represent the activities of daily living (ADL) in a residential space. Based on the activity intention ontology, which represents actions and poses related to human activity intentions, the proposed method identifies the intention of a human activity by using event calculus when a percept sequence is entered. Based on the action intent identified, frequency and pattern analysis of the action intention is used to characterize the lifestyle patterns of the elderly. The intentions of everyday behavior occurring in an elderly living space are complex, and it is difficult to recognize the pattern of life through these intentions, which makes it difficult to recognize the intention of a complex occurrence. To solve these problems, this paper constructs an ontology of percept sequences expressed as daily behavioral information, and makes inferences to help recognize activity intent based on event calculus. When evaluating the techniques proposed in this paper, the results of the act intention cognition experiment based on the perceptual information recorded showed 84% precision and 85% recall.
A UAV Situational Awareness Method through the Threat-Related Relation Reasoning between UAV and Surrounding Objects
Seok-Hyun Bae, Myung-Joong Jeon, Hyun-Kyu Park, Young-Tack Park, Hyung-Sik Yoon, Yun-Geun Kim
http://doi.org/10.5626/JOK.2019.46.2.141
As the technological capabilities of UAV(Unmanned Aerial Vehicles) improves, studies are being carried out to intelligently analyze and understand the situation of UAV in order to gain access to the target area while recognizing and avoiding various risks. To achieve the mission of UAV, it is necessary to judge the situation accurately and quickly. To do this, this paper proposes ways to infer the threat-related relationship between an UAV and perceived surrounding objects through a 3 step approach and provide abstract information about the situation of UAV. The first step is to instantiate the object data recognized by UAV to be utilized for ontology and rule-based reasoning. The second step is to define the priority of instantiated object data and to infer the threat-related relationship between them. Finally, recognizing the situation through the relationship inference that takes into account the association between current and past inferred relationships. To evaluation the performance of the proposed method, a virtual UAV environment simulator was built and tested the data 1,000 times that were randomly generated through five sequential UAV moving point paths. Eight kinds of objects could be recognized in UAV path and ten kinds of relationships can be inferred. Overall performance of situation Awareness was an average of 91 percent.
Knowledge Completion Modeling using Knowledge Base Embedding
Hyun-Young Choi, Ji-Hun Hong, Wan-Gon Lee, Batselem Jagvaral, Myung-Joong Jeon, Hyun-Kyu Park, Young-Tack Park
http://doi.org/10.5626/JOK.2018.45.9.895
In recent years, a number of studies have been conducted for the purpose of automatically building a knowledge base that is based on web data. However, due to the incomplete nature of web data, there can be missing data or a lack of connections among the data entities that are present. In order to solve this problem, recent studies have proposed methods that train a model to predict this missing data through an artificial neural network based on natural language embedding, but there is a drawback to embedding entities. In practice, natural language corpus is not present in many knowledge bases. Therefore, in this paper, we propose a knowledge completion method that converts the knowledge base of RDF data into an RDF-sentence and uses embedding to create word vectors. We conducted a triple classification experiment in order to measure the performance of the proposed method. The proposed method was then compared with existing NTN models, and on average, 15% accuracy was obtained. In addition, we obtained 88%accuracy by applying the proposed method to the Korean knowledge base known as WiseKB.
An Approach to a Learning Prediction Model for Recognition of Daily Life Pattern based on Event Calculus
Seok-Hyun Bae, Sung-hyuk Bang, Hyun-Kyu Park, Myung-Joong Jeon, Je-Min Kim, Young-Tack Park
http://doi.org/10.5626/JOK.2018.45.5.466
Several studies have been conducted on data analysis and predicting results with the advance of machine learning algorithms. Still, there are many problems of cleaning the noise of the real-life dataset, which is disturbing a clear recognition on complex patterns of human intention. To overcome this limitation, this paper proposes an event calculus methodology with 3 additional steps for the recognition of human intention: intention reasoning, conflict resolution, and noise reduction. Intention reasoning identifies the intention of the living pattern time-series data. In conflict resolution, existing ongoing intentions and inferred intention are checked by a conflict graph, so that the intentions that can occur in parallel are inferred. Finally, for noise reduction, the inferred intention from the noise of living pattern data is filtered by the history of fluent. For the evaluation of the event calculus module, this paper also proposes data generation methodology based on a gaussian mixture model and heuristic rules. The performance estimation was conducted with 300 sequential instances with 5 intentions that were observed for 13 hours. An accuracy of 89.3% was achieved between the probabilistic model and event calculus module.
Approach for Learning Intention Prediction Model based on Recurrent Neural Network
Sung-hyuk Bang, Seok-Hyun Bae, Hyun-Kyu Park, Myung-Joong Jeon, Je-Min Kim, Young-Tack Park
http://doi.org/10.5626/JOK.2018.45.4.360
Several studies have been conducted on human intention prediction with the help of machine learning models. However, these studies have indicated a fundamental shortcoming of machine learning models since they are unable to reflect a long span of past information. To overcome this limitation, this paper proposes a human intention prediction model based on a recurrent neural network(RNN). For performing predictions, the RNN model classifies the patterns of time-series data by reflecting previous sequence patterns of the time-series data. For performing intention prediction using the proposed model, an RNN model was trained to classify predefined intentions by using attributes such as time, location, activity and detected objects in a house. Each RNN node is composed of a long short-term memory cell to solve the long term dependency problem. To evaluate the proposed intention prediction model, a data generator based on the weighted-graph structure has been developed for generating data on a daily basis. By incorporating 23,000 data instances for training and testing the proposed intention prediction model, a prediction accuracy value of 90.52% was achieved.
SWAT: A Study on the Efficient Integration of SWRL and ATMS based on a Distributed In-Memory System
Myung-Joong Jeon, Wan-Gon Lee, Batselem Jagvaral, Hyun-Kyu Park, Young-Tack Park
http://doi.org/10.5626/JOK.2018.45.2.113
Recently, with the advent of the Big Data era, we have gained the capability of acquiring vast amounts of knowledge from various fields. The collected knowledge is expressed by well-formed formula and in particular, OWL, a standard language of ontology, is a typical form of well-formed formula. The symbolic reasoning is actively being studied using large amounts of ontology data for extracting intrinsic information. However, most studies of this reasoning support the restricted rule expression based on Description Logic and they have limited applicability to the real world. Moreover, knowledge management for inaccurate information is required, since knowledge inferred from the wrong information will also generate more incorrect information based on the dependencies between the inference rules. Therefore, this paper suggests that the SWAT, knowledge management system should be combined with the SWRL (Semantic Web Rule Language) reasoning based on ATMS (Assumption-based Truth Maintenance System). Moreover, this system was constructed by combining with SWRL reasoning and ATMS for managing large ontology data based on the distributed In-memory framework. Based on this, the ATMS monitoring system allows users to easily detect and correct wrong knowledge. We used the LUBM (Lehigh University Benchmark) dataset for evaluating the suggested method which is managing the knowledge through the retraction of the wrong SWRL inference data on large data.
Extracting Rules from Neural Networks with Continuous Attributes
Batselem Jagvaral, Wan-Gon Lee, Myung-joong Jeon, Hyun-Kyu Park, Young-Tack Park
http://doi.org/10.5626/JOK.2018.45.1.22
Over the decades, neural networks have been successfully used in numerous applications from speech recognition to image classification. However, these neural networks cannot explain their results and one needs to know how and why a specific conclusion was drawn. Most studies focus on extracting binary rules from neural networks, which is often impractical to do, since data sets used for machine learning applications contain continuous values. To fill the gap, this paper presents an algorithm to extract logic rules from a trained neural network for data with continuous attributes. It uses hyperplane-based linear classifiers to extract rules with numeric values from trained weights between input and hidden layers and then combines these classifiers with binary rules learned from hidden and output layers to form non-linear classification rules. Experiments with different datasets show that the proposed approach can accurately extract logical rules for data with nonlinear continuous attributes.
Automated Modelling of Ontology Schema for Media Classification
Nam-Gee Lee, Hyun-Kyu Park, Young-Tack Park
With the personal-media development that has emerged through various means such as UCC and SNS, many media studies have been completed for the purposes of analysis and recognition, thereby improving the object-recognition level. The focus of these studies is a classification of media that is based on a recognition of the corresponding objects, rather than the use of the title, tag, and scripter information. The media-classification task, however, is intensive in terms of the consumption of time and energy because human experts need to model the underlying media ontology. This paper therefore proposes an automated approach for the modeling of the media-classification ontology schema; here, the OWL-DL Axiom that is based on the frequency of the recognized media-based objects is considered, and the automation of the ontology modeling is described. The authors conducted media-classification experiments across 15 YouTube-video categories, and the media-classification accuracy was measured through the application of the automated ontology-modeling approach. The promising experiment results show that 1500 actions were successfully classified from 15 media events with an 86 % accuracy.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr