Web Of Knowledge Research Articles

Leveraging NLP and web knowledge graphs to harmonize locations: A case study on US patent transactions

In the present study, we introduce a novel methodology for the harmonization and standardization of locations associated with patent transactions recorded at the USPTO from 2005 to 2022. Using natural language processing (NLP) techniques in conjunction with search engine-based web knowledge graphs, our method comprises four phases: data pre-processing, semantic clustering, exploitation of web-knowledge graphs, and API-driven harmonization. Initiating our analysis with a dataset of 63,838 unique locations, our methodology effectively reduces this number by more than 50 %. This approach exhibits an accuracy rate of approximately 92 %. The resulting geolocated dataset of companies’ patent transactions offers a valuable resource for fine-grained geographical analyses of the markets for technology; in particular, we provide examples of relevant economic insights which can be learned from looking at the geographical patterns of those transactions.

PheMIME: an interactive web app and knowledge base for phenome-wide, multi-institutional multimorbidity analysis.

To address the need for interactive visualization tools and databases in characterizing multimorbidity patterns across different populations, we developed the Phenome-wide Multi-Institutional Multimorbidity Explorer (PheMIME). This tool leverages three large-scale EHR systems to facilitate efficient analysis and visualization of disease multimorbidity, aiming to reveal both robust and novel disease associations that are consistent across different systems and to provide insight for enhancing personalized healthcare strategies. PheMIME integrates summary statistics from phenome-wide analyses of disease multimorbidities, utilizing data from Vanderbilt University Medical Center, Mass General Brigham, and the UK Biobank. It offers interactive and multifaceted visualizations for exploring multimorbidity. Incorporating an enhanced version of associationSubgraphs, PheMIME also enables dynamic analysis and inference of disease clusters, promoting the discovery of complex multimorbidity patterns. A case study on schizophrenia demonstrates its capability for generating interactive visualizations of multimorbidity networks within and across multiple systems. Additionally, PheMIME supports diverse multimorbidity-based discoveries, detailed further in online case studies. The PheMIME is accessible at https://prod.tbilab.org/PheMIME/. A comprehensive tutorial and multiple case studies for demonstration are available at https://prod.tbilab.org/PheMIME_supplementary_materials/. The source code can be downloaded from https://github.com/tbilab/PheMIME. PheMIME represents a significant advancement in medical informatics, offering an efficient solution for accessing, analyzing, and interpreting the complex and noisy real-world patient data in electronic health records. PheMIME provides an extensive multimorbidity knowledge base that consolidates data from three EHR systems, and it is a novel interactive tool designed to analyze and visualize multimorbidities across multiple EHR datasets. It stands out as the first of its kind to offer extensive multimorbidity knowledge integration with substantial support for efficient online analysis and interactive visualization.

FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design

Recent advancements in Large Language Models (LLMs) have exhibited notable efficacy in question-answering (QA) tasks across diverse domains. Their prowess in integrating extensive web knowledge has fueled interest in developing LLM-based autonomous agents. While LLMs are efficient in decoding human instructions and deriving solutions by holistically processing historical inputs, transitioning to purpose-driven agents requires a supplementary rational architecture to process multi-source information, establish reasoning chains, and prioritize critical tasks. Addressing this, we introduce FinMem, a novel LLM-based agent framework devised for financial decision-making. It encompasses three core modules: Profiling, to customize the agent's characteristics; Memory, with layered message processing, to aid the agent in assimilating hierarchical financial data; and Decision-making, to convert insights gained from memories into investment decisions. Notably, FinMem's memory module aligns closely with the cognitive structure of human traders, offering robust interpretability and real-time tuning. Its adjustable cognitive span allows for the retention of critical information beyond human perceptual limits, thereby enhancing trading outcomes. This framework enables the agent to self-evolve its professional knowledge, react agilely to new investment cues, and continuously refine trading decisions in the volatile financial environment. We first compare FinMem with various algorithmic agents on a scalable real-world financial dataset, underscoring its leading trading performance in stocks. We then fine-tuned the agent's perceptual span and character setting to achieve a significantly enhanced trading performance. Collectively, FinMem presents a cutting-edge LLM agent framework for automated trading, boosting cumulative investment returns.

Increasing marine trophic web knowledge through DNA analyses of fish stomach content: a step towards an ecosystem-based approach to fisheries research.

Multispecies and ecosystem models, which are key for the implementation of ecosystem-based approaches to fisheries management, require extensive data on the trophic interactions between marine organisms, including changes over time. DNA metabarcoding, by allowing the simultaneous taxonomic identification of the community present in hundreds of samples, could be used for speeding up large-scale stomach content data collection. Yet, for DNA metabarcoding to be routinely implemented, technical challenges should be addressed, such as the potentially complicated sampling logistics, the detection of a high proportion of predator DNA, and the inability to provide reliable abundance estimations. Here, we present a DNA metabarcoding assay developed to examine the diet of five commercially important fish, which can be feasibly incorporated into routinary samplings. The method is devised to speed up the analysis process by avoiding the stomach dissection and content extraction steps, while preventing the amplification of predator DNA by using blocking primers. Tested in mock samples and in real stomach samples, the method has proven effective and shows great effectiveness discerning diet variations due to predator ecology or prey availability. Additionally, by applying our protocol to mackerel stomachs previously analyzed by visual inspection, we showcase how DNA metabarcoding could complement visually based data by detecting overlooked prey by the visual approach. We finally discuss how DNA metabarcoding-based data can contribute to trophic data collection. Our work reinforces the potential of DNA metabarcoding for the study and monitoring of fish trophic interactions and provides a basis for its incorporation into routine monitoring programs, which will be critical for the implementation of ecosystem-based approaches to fisheries management.

The dark web: A hidden menace or a tool for privacy protection

The dark web has long been a subject of debate, as it raises questions about its covert threats and privacy safeguarding capabilities. It's an internet realm requiring specific software or authorization, notorious for illegal activities and untraceable transactions. Researchers and law enforcement agencies use the dark web for intelligence gathering on cybercriminals, making it crucial in understanding the evolution of the internet and emerging criminal activities. Additionally, studying the dark web helps identify cybersecurity threats and vulnerabilities. This research explores the relationship between the dark web and mental health, investigating the public's perception of the dark web as both a hidden menace and a tool for privacy protection. The primary aim is to examine the correlation between public perception and engagement in dark web services, while also exploring potential mental health ramifications. To collect data, we conducted a survey among university students and regular internet users. The research process involved identifying 18 research questions covering topics such as dark web knowledge, usage frequency, security measures, awareness of government regulations, ability to differentiate legal and illegal content, transaction methods, and understanding the dark web's role in promoting freedom of speech. We collected 158 responses from different departments of Chandigarh University, removing duplicates to retain 156 responses meeting our criteria. Data was compiled in an Excel file for further statistical evaluation. While the dark web is not widely accessed, users may experience psychological consequences due to its association with cybercrime, disturbing content, and radicalization. This connection leads to heightened anxiety, fear, and distress among users. Concerns about privacy, security, and associated risks further contribute to these apprehensions. Motivations for dark web access, such as curiosity and anonymity, may stem from underlying psychological factors. However, navigating illegal and potentially harmful content can result in emotional turmoil, moral dilemmas, guilt, or shame. These findings highlight the need for public awareness campaigns, educational initiatives, and targeted interventions promoting mental well-being and safe internet practices. Enhancing individuals' knowledge about the dark web's risks and potential psychological consequences empowers them to make informed decisions and protect their mental health in the digital age.

CROA-based feature selection with BERT model for detecting the offensive speech in Twitter data

<p>Online hate speech has flourished on social networking sites due to the widespread availability of mobile computers and other Web knowledge. Extensive research has shown that online exposure to hate speech has real-world effects on marginalized communities. Research into methods of automatically identifying hate speech has garnered significant attention. Hate speech can affect any demographic, while some populations are more vulnerable than others. Relying solely on progressive learning is insufficient for achieving the goal of automatic hate speech identification. It need access to large amounts of labelled data to train a model. Inaccurate statistics on hate speech and preconceived notions have been the biggest obstacles in the field of hate speech research for a long time. This research provides a novel strategy for meeting these needs by combining a transfer-learning attitude-based BERT (Bidirectional Encoder Representations from Transformers) with a coral reef optimization-based approach (CROA). A feature selection (FC) optimization strategy for coral reefs, a coral reefs optimization method mimics coral behaviours for reef location and development. We might think of each potential answer to the problem as a coral trying to establish itself in the reefs. The results are refined at each stage by applying specialized operators from the coral reefs optimization algorithm. When everything is said and done, the optimal solution is chosen. We also use a cutting-edge fine-tuning method based on transfer learning to assess BERT’s ability to recognize hostile contexts in social media communications. The paper evaluates the proposed approach using Twitter datasets tagged for racist, sexist, homophobic, or otherwise offensive content. The numbers show that our strategy achieves 5%–10% higher precision and recall compared to other approaches.</p>

Social Network Analysis Uses and Contributions to Innovation Initiatives in Rural Areas: A Review

This paper identifies the main contributions of Social Network Analysis (SNA) use in the study of innovations in rural areas with an emphasis on agriculture and forestry. The bibliographic analysis was carried out on the Web of Knowledge (WoK) and Scopus platforms. Sixty-eight studies were found in which SNA was used as the main research tool in innovation processes in rural areas. The main fields of the SNA contribution were as follows: (i) social capital; (ii) social learning: information and knowledge flow for the adoption of innovations produced through existing social structures; (iii) the implementation and evaluation of innovations for local and territorial dynamization. The study contributes to summarizing the existing knowledge on SNA use in the study of innovations in rural areas and to informing future research. Understanding social networks is essential to strengthen and enhance the existing social capital and to promote social learning related to innovations in rural areas.

A Reference Architecture for Supporting Multidimensional Big Data Analytics over Big Web Knowledge Bases: Definitions, Implementation, Case Studies

This paper focuses attention on the problem of supporting multidimensional big data analytics tools over Big Web Knowledge Bases. These big data repositories are generated from web systems built on top of big datasets, as in several real-life examples such as web interfaces to social networks, e-government systems, smart city tools (e.g., for weather forecasting), and so forth. The problem of extracting actionable knowledge insights from big web knowledge bases is a relevant challenge now. In line with this emerging research field, the paper proposes models, issues and research trends in the context of multidimensional big data analytics over big web knowledge bases, along with a reference architecture that impersonates this paradigm.

The Facets of Intangible Heritage in Southern Chinese Martial Arts: Applying a Knowledge-driven Cultural Contact Detection Approach

Investigating the intangible nature of a cultural domain can take multiple forms, addressing, for example, the aesthetic, epistemic, and social dimensions of its phenomenology. The context of Southern Chinese martial arts is of particular significance, as it carries immaterial components of all these aspects: The technical and stylistic framework of a martial art system; the imagery associated to movements; and the transmission of knowledge orally, practically, or through influence, are but examples of intangible characteristics that can and should be captured, not unlike cultural artifacts. The latter case—the one of formalizing cultural influence through its various forms of evidenceis emblematic as well as largely untrodden ground. A previous attempt at detecting cultural influence computationally was made in the context of Roman archaeology, though the binding of that early effort with the domain model was tight; also, there has hardly been any prior dedicated effort to model the martial arts domain through ontologies. In this article, we present the realization of the full cycle of a computational approach to investigating cultural contact in Southern Chinese martial arts. The entire approach is predicated upon the usage of standards and techniques of the Semantic Web and formal knowledge. Starting from a modular domain ontology, which models martial arts independently of the goal of capturing cultural influence, we perform knowledge extraction from archival material from the Hong Kong Martial Arts Living Archive and generate a dataset of the results modeled after said ontology. Then, we combine the resulting knowledge base with a rule model that represents ways to infer knowledge of potential contact between cultures based on the evidence present in the knowledge base. The results offer an insight into how an inference-based computational model can be applied to detect interesting facts even in the as-yet underexplored domain of intangible cultural heritage. The implemented workflow shows that the full-cycle employment of semantic technologies can offer the ground truth required for largely different approaches, such as statistical and machine learning ones, to operate.

SocioCoast: Design and Implementation of a Data-Driven Platform for Citizen Science in Coastal Areas

Due to the rapid urbanization and globalization that has occurred in recent decades, cities have undergone significant changes and transformations. The smart city concept can lead to improvements in the life quality of citizens and contribute to the city’s sustainability by collecting humans’ real-time observations. More specifically in smart coastal areas, citizen’s science data can be utilized not only to increase the effectiveness of its services like smart tourism and smart transit but also to raise awareness about environmental issues. Emphasizing on coastal areas under the Blue Flag program and the Natura 2000 network, in this work, we introduce a system that aims to collect crowdsourcing data from citizens and tourists and combine or compare them with existing knowledge for these areas. We describe the design and implementation: the back-end, its data sources and also the front-end outputs, a web knowledge platform, and a crowdsourcing mobile application. Geospatial information as well as historical or real-time data collected will be provided to users for further exploitation and creation of their own applications. A demonstration of the system is also provided to showcase its usefulness in increasing attractiveness and raising awareness for coastal areas by offering users rich and updated information.

Towards improving prediction accuracy and user-level explainability using deep learning and knowledge graphs: A study on cassava disease

Food security is currently a major concern due to the growing global population, the exponential increase in food demand, the deterioration of soil quality, the occurrence of numerous diseases, and the effects of climate change on crop yield. Sustainable agriculture is necessary to solve this food security challenge. Disruptive technologies, such as of artificial intelligence, especially, deep learning techniques can contribute to agricultural sustainability. For example, applying deep learning techniques for early disease classification allows us to take timely action, thereby helping to increase the yield without inflicting unnecessary environmental damage, such as excessive use of fertilisers or pesticides. Several studies have been conducted on agricultural sustainability using deep learning techniques and also semantic web technologies such as ontologies and knowledge graphs. However, the three major challenges remain: (i) the lack of explainability of deep learning-based systems (e.g. disease information), especially to non-experts like farmers; (ii) a lack of contextual information (e.g. soil or plant information) and domain-expert knowledge in deep learning-based systems; and (iii) the lack of pattern learning ability of systems based on the semantic web, despite their ability to incorporate domain knowledge. Therefore, this paper presents the work on disease classification, addressing the challenges as mentioned earlier by combining deep learning and semantic web technologies, namely ontologies and knowledge graphs. The findings are: (i) 0.905 (90.5%) prediction accuracy on large noisy dataset; (ii) ability to generate user-level explanations about disease and incorporate contextual and domain knowledge; (iii) the average prediction latency of 3.8514 s on 5268 samples; (iv) 95% of users finding the explanation of the proposed method useful; and (v) 85% of users being able to understand generated explanations easily—show that the proposed method is superior to the state-of-the-art in terms of performance and explainability and is also suitable for real-world scenarios.

Systematic review of current academic production of the concepts of BIM, thermal performance, energy efficiency and sustainability

This paper aims to discuss the current applications of Building Information Modelling along with Thermal Performance, Energy Efficiency, and Sustainability. It proposes one integrative matrix of concepts, containing information about the practical applications of high-impact papers within the academic knowledge, exposing the discussion to other stakeholders such as constructors, students, and managers. With the integrated matrix the study aims to present a state-of-the-art synthesis in terms of BIM applied, along with energy and thermal efficiency and sustainability, which may aid BIM practitioners in developing BIM applications in the context presented. This paper will perform a critical review of the existing literature, using the science web knowledge base, analysing works published between 2018 and 2021, with a higher impact factor. This research has a focus on works that perform practical applications of BIM tools to perform thermal performance and energy efficiency analysis, presenting the sustainability factor as a topic of discussion. After the systematic review, the works are selected to compound the integrated matrix will be analysed in order to clarify its practical applications. After selecting 29 (20% of the most relevant) high-impact academic papers, they will be schematized in an integrated matrix that summarizes the discussion about the applications of each paper. The paper provides qualitative data about the current applications of BIM, summarizing high-impact applications into an integrated matrix, allowing BIM practitioners and academics to use these results to find new applications, authors, discussions, and data more efficiently.

Perceived Influence of Digital Competence on Knowledge Sharing Behaviour of College Librarians in the South-East and South-South Geo-Political Zones of Nigeria

Knowledge and the competence to share it in the digital age have become potent tools for the attainment of competitive advantage and sustainable development in organizations and institutions of learning. This awareness has necessitated the need to embark on a study with the main objective of determining how digital competence variables influence the knowledge sharing behaviour of librarians in college libraries in the South–East and South–South of Nigeria. The research was supported by the Theory of Planned Behaviour (TPB) and Technology Acceptance Model (TAM). Survey research method was used to investigate a sample of 264 librarians. A validated and structured questionnaire with a correlation coefficient of 0.92 was used to collect data for the research through the census method. A response rate of 82% (217) was received and analysed. Findings reveal that librarians in college libraries possess the ability to use and manipulate a broad range of digital devices like smart phones, tablets, lap/desk tops; have a working knowledge of the World Wide Web (www); use social media platforms to maintain workplace collaboration; and use popular library application software packages. The study concludes that librarians in the institutions studied have shown significant skills in the use of ICT in their knowledge sharing behaviour and therefore should be sustained. The research recommends amongst others regular ICT training, workshops and conferences to enhance the knowledge sharing behaviour of the librarians and the creation of digital platforms and infrastructure for knowledge creation and sharing for professional and institutional ranking and visibility. Keywords: Digital Competence, Knowledge Sharing Behaviour, College Librarians, Nigeria DOI: 10.7176/IKM/13-3-03 Publication date: March 31 st 2023

Technological Prospects of Cloud Computing in Web Mining: Recent Trends and Opportunities

Web has immense potentials to grow with new technologies and flourish with new opportunities in almost every sphere of human life. The internet grew by WWW and explored with e-business and social network. The Web is known and participating in mining of various kinds of data ranging from users views and patterns to Bitcoin applications. Web mining includes different set and types of data and extract useful information and from various sources of web and gain knowledge using applicable data mining technique on dataset. Web mining types broadly categorised in three main areas, web uses mining, web content mining, and web structured mining. Each category of web mining is involved to handle issues of heterogeneous behaviour of web data. All three technique and applications are highly required of high end architectures which can facilitate infrastructures and support for all required criteria. Cloud computing is an emerging technology and blowing with intensive support for varieties of applications. The techniques and applications of Web usage mining are extremely demanded in cloud computing. Cloud computing allow these technique to retrieve relevant and useful data through virtually integrated mode of data warehouse. It helps the users to reduce cost and infrastructure for implementation. This paper presents methodologies of web mining using Cloud Computing technology and its prospects.

Individualistic Versus Collaborative Learning in an eHealth Literacy Intervention for Older Adults: Quasi-Experimental Study.

Older adults tend to have insufficient health literacy, which includes eHealth literacy-the ability to access, assess, and use digital health information. Interventions using methods such as collaborative learning (CL) and individualistic learning (IL) may be effective in addressing older adults' low eHealth literacy, but little is known about the short- and long-term effects of CL versus IL on older adults' eHealth literacy. The objective of this study was to use a 3 × 2 × 3 mixed factorial design to examine older adults' learning with CL versus IL for eHealth literacy. Older adults (N=466; mean age 70.5, SD 7.2; range 60-96 years) from diverse racial and ethnic groups were randomly assigned to either the CL or IL group (233/466, 50% in each). The intervention consisted of 4 weeks of training in 2-hour sessions held twice a week. Using ANOVA and multiple regression, we focused on the main effects of learning condition and interaction between learning condition and previous computer experience. Learning method (CL or IL) and previous computer experience (experienced, new, or mixed) were between-subject variables, and time of measurement (pretest measurement, posttest measurement, and 6-month follow-up) was the within-subject variable. Primary outcome variables were eHealth literacy efficacy, computer and web knowledge, basic computer and web operation skills, information-seeking skills, and website evaluation skills. Control variables were age, sex, education, health status, race and ethnicity, income, primary language, and previous health literacy. eHealth literacy efficacy, computer and web knowledge, basic computer and web operation skills, information-seeking skills, and website evaluation skills improved significantly (P<.001 in all cases) from before to after the intervention. From postintervention measurement to 6-month follow-up, there was a significant interaction between learning condition and previous computer experience based on 1 outcome measure, computer and web operation skills (F2,55=3.69; P=.03). To maintain computer and web operation skills 6 months after the intervention, it was more effective for people with little to no previous computer experience to learn individually, whereas for people with more previous computer experience, it was more effective to learn collaboratively. From postintervention measurement to 6-month follow-up, statistically significant decreases were found in 3 of the 5 outcome measures: eHealth literacy efficacy, computer and web knowledge, and basic computer and web operation skills (P<.001 for all 3 cases). Older adults' eHealth literacy can be improved through effective intervention, and the IL or CL condition may have little effect on short-term outcomes. However, to maintain long-term benefits, it may be best to learn collaboratively with others who have similar previous computer experience. eHealth literacy is multidimensional, with some components retained better over time. Findings suggest a need for resources to provide continuous training or periodic boosting to maintain intervention gains.

ETwinning And Activities of Daily Living İn Special Education

ITCs are considered necessary and positive, but in special education it is necessary to take a step forward that allows students to communicate with other students from different cities, show their work and at the same time learn together with the help of their teachers. This study aims to improve communication between schools and special education students and to know the effect it generates on the learning of life skills through the eTwinning platform. In addition, an initial and final survey of parents and teachers working with the group is conducted. A total of 80 teachers from special education schools in different cities of Turkey and from a Spanish university participate together with their primary education teachers. Working with the Etwinning platform has been found to improve communication between schools and special education students. In this research, students with special educational needs have overcome communication barriers by interacting with others outside of their everyday environments by demonstrating that they can achieve anything, when given the opportunity to create, learn, know and develop digital skills. There is a greater knowledge of web 2.0 tools, such as canva, wordart, ingpss learning, KineMaster, Pixiz, vivavideo, meitu, phinsh to publicize the work done, assuming responsibility for their own learning, in addition, parents of students consider that the activities carried out within the framework of the eTwinning program have had an impact on their children's daily life skills.

Table interpretation of the temporal description logic LTLALC

Description logics are widely used to describe and represent knowledge in the Semantic Web. This is a modern and powerful mechanism that provides the possibility of extracting knowledge from already existing ones. Thanks to this, conceptual of subject areas modeling has become one of the fields of application of descriptive logics, taking into account the use of inference mechanisms. Conceptual modeling is used to create databases and knowledge bases. A key issue of the subject area modeling is the ability to monitor the dynamics of changes in the state of the subject area over time. It is necessary to describe not only the current actual state of the database (knowledge bases), but also the background. Temporal descriptive logics are used to solve this problem. They have the same set of algorithmic problems that are presented in conventional descriptive logics, but to them are added questions related to the description of knowledge in time. This refers to the form of time (continuous or discrete), time structure (moments of time, intervals, chains of intervals), time linearity (linear or branched), domain (present, past, future), the concept of “now”, the method of measurement, etc. An urgent task today is to create an algorithm for the temporal interpretation of conventional descriptive logics. That is, to show a way in which temporal descriptive logic can be applied to ordinary descriptive logic. The paper presents an algorithm for temporal interpretation of LTL into ALC. Linear, unbranched time is chosen for the description goal. It is presented in the form of a whole temporal axis with a given linear order on it. Only the future tense is considered. The algorithm contains graphic notations of LTL application in ALC: concepts, concept constructors, roles, role constructors, TBox and ABox. Numerous examples are used to illustrate the application of the algorithm.

Taylor Horse Herd Optimized Deep Fuzzy clustering and Laplace based K-nearest neighbor for web page recommendation

Background: Web-page recommendation serves a significant part in intelligent web frameworks and discovery of meaningful information from web and sufficient knowledge illustration for efficient web-page recommendations are quintessential and highly demanding issue. The conventional key matching and statistical models for web page recommendation are insufficient as they recommend web pages to the user mostly irrelevant to the query. It is essential to satisfy individual user separately according to the available preference data. In contrast, graph framework can deliver suitable illustration of user-item preference data and by employing this graph structure-based web page recommendation system is highly desirable and efficient. Methods: This research proposes a web page recommendation model using Taylor Horse Herd Optimization (THHO)-based Deep Fuzzy Clustering (DFC). Here, the interesting sub graphs from web log database are retrieved using Weighted-Gaston (W-Gaston) algorithm. DFC is exploited to cluster the sub graphs and DFC is effectively trained using THHO algorithm. Moreover, THHO is devised by the incorporation of Taylor series with Horse Herd Optimization Algorithm (HOA). Finally, suitable web pages are recommended to the user based on their query using Laplace correction-based K-Nearest neighbor (LKNN) model. Result: The proposed THHO_DFC approach has obtained maximum precision of 0.950, recall of 0.897, F-measure of 0.922, and accuracy of 0.926 while analyzing the system based on 90% data using MSNBC dataset. Conclusion: The proposed model has delivered better insights in terms of efficiency and provides high performance in recommending suitable web pages.

Learning and Reasoning for Cultural Metadata Quality: Coupling Symbolic AI and Machine Learning over a Semantic Web Knowledge Graph to Support Museum Curators in Improving the Quality of Cultural Metadata and Information Retrieval

This work combines semantic reasoning and machine learning to create tools that allow curators of the visual art collections to identify and correct the annotations of the artwork as well as to improve the relevance of the content-based search results in these collections. The research is based on the Joconde database maintained by the French Ministry of Culture that contains illustrated artwork records from main French public and private museums representing archeological objects, decorative arts, fine arts, historical and scientific documents, etc. The Joconde database includes semantic metadata that describes properties of the artworks and their content. The developed methods create a data pipeline that processes metadata, trains a Convolutional Neural Network image classification model, makes prediction for the entire collection and expands the metadata to be the base for the SPARQL search queries. We developed a set of such queries to identify noise and silence in the human annotations and to search image content with results ranked according to the relevance of the objects quantified by the prediction score provided by the deep learning model. We also developed methods to discover new contextual relationships between the concepts in the metadata by analyzing the contrast between the concepts’ similarities in the Joconde's semantic model and other vocabularies and we tried to improve the model prediction scores based on the semantic relations. Our results show that cross-fertilization between symbolic AI and machine learning can indeed provide the tools to address the challenges of the museum curators work describing the artwork pieces and searching for the relevant images.

An Open-Ended Web Knowledge Retrieval Framework for the Household Domain With Explanation and Learning Through Argumentation

The authors present a knowledge retrieval framework for the household domain enhanced with external knowledge sources that can argue over the information that it returns and learn new knowledge through an argumentation dialogue. The framework provides access to commonsense knowledge about household environments and performs semantic matching between entities from the web knowledge graph ConceptNet, using semantic knowledge from DBpedia and WordNet, with the ones existing in the knowledge graph. They offer a set of predefined SPARQL templates that directly address the ontology on which their knowledge retrieval framework is built and querying through SPARQL. The framework also features an argumentation component, where the user can argue against the answers of the knowledge retrieval component of the framework under two different scenarios: the missing knowledge scenario, where an entity should be in the answers, and the wrong knowledge scenario, where an entity should not be in the answers. This argumentation dialogue can end up in learning a new piece of knowledge when the user wins the dialogue.