Activities 2019
Sumary activity sub-project 1 2019
Sumary activity sub-project 2 2019
Sumary activity sub-project 3 2019
Sumary activity sub-project 4 2019
Sumary activity sub-project 5 2019
INTERMEDIATE SCIENTIFIC REPORT 2
Developing an emerging concept of new technologies developed in the field of ICT in online social networks
January – December 2019
The complex project "Empirical modelling and experimental development of tools associated with emerging technologies in the field of online social networks" (FutureSocialWeb) proposes the study of emerging technologies in the field of information technology and communications in relation to online social networks, with the degree and manner of adoption of potential users of these technologies, assessing the impact they will generate on the socio-economic environment.
The proposed project aims at conducting scientific studies and research in order to achieve empirical models and experimental developments to assess the impact of these new technologies, including complex analysis and recommendation systems (deep learning), Internet of Things, body Communication devices (Wearables), neural analysis (pupilometers, galvanometers or EEG), augmented reality and semantic searches in the field of online social networks. The project is operationalized through 5 component subprojects, which are achieved in three distinct stages over three years (2018-2020).
SUMMARY OF ACTIVITIES CARRIED OUT IN THE YEAR 2019
SUB-PROJECT 1: "Experimental development of emerging technologies in the field of deep learning on big data at the level of online social networks and the study of their impact at users ' level (AI Media)"
Activity 1-2-2: Conducting the testing and evaluation methodology of the Media application modules
In this activity, a test plan was conducted for the functions implemented in the Media application and the methodology for testing them was presented.
To assess the efficiency of recognition of the services implemented in the AIMedia application, the F-score measure was used to assess the accuracy of a test applied to an implemented recognition/classification system. The F-score calculation is based on two terms, precision P and recall R, in which p represents the number of correct results identified divided by the number of results deemed correct returned, and r represents the number of correctly identified results divided by the number of results that are correct in reality. For each of the images and videos uploaded to the AIMedia application, the recognition service is applied, monitoring the period between the moment the service is launched until the result is received. For this activity, a study was conducted and presented as an outcome, according to plan.
Activity 1-2-3: Conducting qualitative research to determine the functionality and utility problems of the platform
The increasing use of image location recognition and social context in social media posts offers companies intending to use the AI Media platform opportunities to post relevant content on each social network and to induce the customization of the distributed digital content.
The proposed and achieved aim through this research was to identify and analyze the feedback of potential users of the AI Media platform, in terms of its capabilities, focused on the recognition of geolocation and social context. The marketing specialists of the FutureWeb research team proposed different causal configurations of the capabilities of the AI Media platform and conducted a study they presented as an outcome.
Activity 1-2-4: Making specifications to improve the AI Media application
The purpose of this activity was to identify possible developments in the AI Media application, and the main proposed and achieved objective was to identify aspects related to logos recognition functions that don`t exist among the initial AI Media application functions. The activity was successfully accomplished, and an available study was presented as a result.
Activity 1-2-5: Development of AIMedia Application: Developing a functional model of the AIMedia app with 2 modules which purpose is the logo recognition of brands or companies in images posted on social networks and media videos.
In this activity, the performances of three types of neural networks convolutions were analyzed: R-CNN, Fast R-CNN and Faster R-CNN. For testing purposes, out of the 15 datasets identified, the following were selected: BelgaLogos, FlickrLogos-27, Logos-32plus, QMUL-OpenLogo, TopLogo-10. The Pretrained Networks identified in this study are the follwing: alexnet, squeezenet, cifar10Net, densenet201, googlenet, nceptionresnetv2, inceptionv3, mobilenetv2, nasnetlarge, nasnetmobile, resnet101, resnet18, resnet50, shufflenet, vgg16, vgg19, xception. For each of the three types of neural networks convolutions, R-CNN, Fast R-CNN and Faster R-CNN, the performance was tested and the results were presented in the table existing in Tests section of Study no. 4, attached to this report as an outcome of the research. The AIMedia app is available at: http://195.34.77.2:12181.
Activity 1-3-1: Developing and implementing options and services (subsystems) for the AIMedia application: Developing the Geolocation recognition service for the photos or media clips in which a brand or a company logo appears
The GeoEstimation method was used for the recognition of geolocation. Thus, recognizing the location in the photo using CNN convolutional networks addresses the division on geographic cells. The approach used in the study involves: a) Deep networks that are trained separately with images from distinct scene categories; b) A multi-task network trained with both geographic and scene labels. To assign image labels, the ResNet152 network is used, and the model is trained on more than 16 million images. To enter the scene information, individual recognition networks have been built. The data set MediaEval Placing Task 2016 was used, which contains about 5 million images of Geotag. The assessment of these approaches is made on two sets of public reference data for geolocation estimation: The Im2GPS dataset, which contains 237 photographs, and the Im2GPS3k dataset, containing 3,000 images. In the implementation section of the completed study (available) screenshots on the steps followed for the implementation and use of geolocation recognition within the AIMedia application are presented, as well as the main method, based on neural networks and the alternative methods mentioned above.
Activity 1-3-2: Developing and implementing options and services (subsystems) for the AI Media application level: Developing a recognition service in order to identify the social context where the photos or media clips in which the brand or company`s logo appears
In order to identify the social context in which the logos appear, four categories of interest were defined: family environment; business environment; institutional environment and recreational environment. As with the recognition of logos in images, in order to obtain a better accuracy, both the learning transfer technique and the pretrained GoogleNet network were used on the Places365 Dataset. GoogleNet is a pretrained convolutional neural network that has a depth of 22 layers, with 144 layers in total and 170 connections. In the Implementation Section of the Report no. 6 (available), screenshots related to the steps followed for the implementation and use of GoogleNet convolutional neural prescriptions within the AIMedia application are presented.
SUB-PROJECT 2: "The experimental development of the emerging technologies in the field of mobile communications at the level of online social networks and the study of their impact on users" (Integrated Mobile Social Networks)
By making prototypes based on the concepts of IoT (Internet of Things) and wearable (portable devices), the impact of these technologies on users of the FutureWeb online platform is evaluated, as well as the impact on the activity of the business environment, especially from the marketing activities perspective. This way, the improvement of the institutional performance is ensured, as well as the management of the research resources involved in this project and the exploitation of the results in the economic and social environment, as well as in the users’ individual activities. In 2019, the development of the module based on mobile technologies (IoT and wearable) was carried out in several stages / activities, following the planned activities:
1.The development of an experimental prototype of the mobile system within the FutureWeb platform
The first step was the proposal of simple or integrated mock-ups: Discount coupons, Gamification application, Fitness & sport, Integrated proposal - Biometrics360. The unique approach is to combine the 360-degree evaluation method used in the business environment and the biometric information (movement, sleep, pulse) collected through IoT technologies. The collected data are processed using a data flow capable of generating a personal and professional development profile using an ontology that allows the expression of concepts in this field. A personalized personal development plan can be built. Used tools: Performance - feedback form, Biometric data.
Integrated Proposal - Video Sentiment-based Segmentation. Creating an online social network where users will view different videous in order to improve them. This will be done with neuromarketing technologies (EEG and eyetracking) able to identify several elements such as: the key points of the video, the reaction to different video stimuli, elements regarding the emotional involvement/non-involvement in seeing the video. These videos can be an online course (eg. online marketing or Facebook Ads) or advertising material. INSTRUMENTS: Neuromarketing, IoT, Semantic Web
Monitoring the air quality in the course/training rooms and study on the level of students’ involvement
The data about the air quality from the study rooms (CO2, particles, humidity, etc.) will be collected and sent to the server. These data are useful for maintaining the air quality in the study rooms at an optimal level, whether they are individual study rooms or are the course rooms.
2.Testing the prototype by analyzing users’ behavior, based on the emerging concept of IoT (Internet of Things)
Research and innovation in IoT technologies play a vital role in the development of the economy. Investing in research is related to investments in people and is essentially an intellectual activity. Scientific research leads to progress and is vital for improving the quality of life. Usability testing is an efficient method of testing the IOT devices created, as well as the integration platforms. This needs to be done with the help of real users, in order to detect the deficiencies of these IoT devices, as well as the platform where the data from the IoT devices are integrated. For testing the prototype by analyzing the users’ behavior, based on the emerging conceptual basis of IoT (Internet of Things), three studies were conducted through combined research methods, namely case studies and semi-conducted interviews. We considered three categories of users: adult females, adult males and students (females / males).
3.Testing the prototype by analyzing the users’ behavior, based on the emerging concept of Wearable (portable devices)
For testing the prototype by analyzing the users’ behavior, based on the emerging conceptual basis of wearable (portable devices), a smart watch will be used - Samsung Gear S2 - having an application for monitoring the pulse and the number of steps. The watch notifications were intended to be prompt and clear - in our case they are sent every 10 seconds. The application works only if it is within 10 meters of the watch and the watch must be connected to the internet.
Six case studies were performed, using research methods appropriate to each situation. All the information obtained through research are presented and used in the following section: Creating specifications for system improvement and implementation of the necessary changes.
4. Creating specifications for system improvement and implementation of the necessary changes
During this stage were presented the needs expressed by the potential users in each case study. All the details regarding the improvement specifications have been taken over and used within the project.
5. Building the IoT section within the subsystem
Within the FutureWeb platform, graphic elements are implemented that can communicate the data coming from the IoT devices for air quality monitoring and from the application for monitoring the pulse and the number of steps. For this purpose a WordPress interface has been created where all the data can be accessed and where new users can be added. The dissemination of the studies’ results was achieved through published scientific articles: “Leverage IoT technologies for Customer Acquisition and Retention” and “Challenges to IoT Innovative Marketing Strategies.
SUB-PROJECT 3: "Experimental development of emerging technologies in the field of neuromarketing on the level of online social networks and the study of their impact on users" (NeuroMedia)
ACTIVITY 3 - 2 - 3. Development of an experimental model (prototype) of the subsystem of neuromarketing within the FutureWeb platform
The classic version proposed for implementation as mockup V0 is a very close alternative to Facebook social network. A first proposal starts from a combination of the text with pictures, and a four-quadrant layout of the platform. Another proposal starts from niche social networks, where the focus is on the images. The members of the network share their eyes in their activity as "photos" - small screenshots of the drawings they are currently working on. Other members then comment and give feedback on the work presented. The photo-based design and the grid model of the platform make it very easy to adapt with the help of eye-traking tools, so the second platform model we propose is based on Dribbdel inspiration. This type of mockup can receive at least two variations, depending on how the text is placed, central or left / right. These are the options that will be developed in the FutureWeb platform as alternatives to the classic mockup and that will be tested with eye-tracking and EEG tools in order to capture the users' reactions when interacting with each of them.
ACTIVITY 3 - 2 - 4. Testing the prototype by analyzing the unconscious responses to visual stimuli from online social networks, using Eye tracking technology
Within this activity were realized:
- Testing the general prototype using Eye tracking technology;
- Testing the V1 prototype using Eye tracking technology;
- Testing the V2 prototype using Eye tracking technology;
- Testing the V3 prototype using Eye tracking technology;
- Testing the V4 prototype using Eye tracking technology;
ACTIVITY 3 - 2 - 5. Testing the prototype by analyzing the unconscious responses to visual stimuli from online social networks, using brain scan technology (EEG)
Within this activity were realized:
- Testing the general prototype using EEG technology;
- Testing the V1 prototype using EEG technology;
- Testing the V2 prototype using EEG technology;
- Testing the V3 prototype using EEG technology;
- Testing the V4 prototype using EEG technology;
ACTIVITY 3 - 2 - 6. Realization of specifications for system improvement and implementation of needed modifications
Of all the analyzes, the classic mockup alternative is the one that arouses the interest the least. Mockup V2 and Mockup V4 are the ones that stimulate the most. The differences not being major between V1 and V2 or between V3 and V4, it seems that the position of the writing on the middle of the page is the one that stimulates the most the respondents’ interest. However, this analysis is not exhaustive because correlating the EEG results with those of Eye-Traking, there are respondents who on V4 have scanned the page the most complex. However, one conclusion would be that the changes in the positioning of images and text influence the perception of information as well as the feelings of users’ commitment.
ACTIVITY 3-3-1: Building the eye-tracking section within the subsystem
Within the Futureweb platform a user study module will be implemented and developed through eyetracking, technology that uses webcams to study location in which the user is looking at those times. In order to stimulate and visually test the users, an area populated with both text and attractive images was performed to study the response time, attention and immediate orientation of the user. The visual content area consists of four mockups. The first two content areas are populated at a standard level with an image and a photo, and the last two areas are populated with a minimum of four photos and a text module.
Content v3 !: consists of five modules, four are populated with images and the last module is populated with text. Content v4 !: consists of six modules, five are populated with images and the last module is populated with text.
ACTIVITY 3 - 4 - 3. Broadly dissemination of project results
Articles in scientific conferences:
1. Individuals' perspective regarding the ethic of neuromarketing techniques integration in online social networks
2. The ethic of using neuromarketing techniques in online social networks from a business perspective
3. An EEG Analysis on the Perception of the Consumers Regarding Video-Commercials from the Automotive Industry
4. Mercedes-Benz and Volkswagen video-commercials – A pluralistic research based on an eye-tracking experiment
5. Assessing the applicability of neuromarketing tools in online social networks from a business perspective
Articles published in scientific journals:
1. Urban green areas’ sustainable development for quality of life improvement. Arguing for increased citizen participation
SUB-PROJECT 4: "Experimental development of augmented reality tools on the level of online social networks and studying their impact on users (AR Media)"
Activity 4-2-1. Development of an experimental model (prototype) of the online social network AR Media using augmented reality of registered users based on the mobile devices cameras
In studying the process of designing the computer system, the starting point represented the notion of system. The proposed system represents a set of interdependent elements (components) between which a dynamic interaction is established, based on predetermined rules, in order to achieve the objective. The designed system accepts the input of information as inputs. It processes the information provided by them and then sends the results as outputs, in the environment in which the system evolves. The system communicates through interfaces with the environment. The system will interact with it through the data and information received.
Subsystem design is very important because, if the system as a whole is no longer functioning properly, fixing should be possible by simply remedying the subsystem that caused the failure. The limitations of the system refer to the constraints imposed by its parameters (for example algorithms of facial recognition or identification of a person's feelings) depending on which system has to work in order to achieve the purpose for which it was achieved.
The main functions and functionalities pursued in this stage are:
- the possibility of adding an unlimited number of members
- the possibility of adding the personal data of the members
- the possibility of sending personal messages between members
- the possibility of posting public messages with text and images to all members of the social network
- the possibility to evaluate the posts with LIKE
- the possibility to track the activity of a member
- the possibility of creating a group of friends
- the possibility to manage personal profile, friends, received messages, etc.
- identification of persons based on the information contained in the database with the members of the social network.
The open-source DNN platform, a web content management system and web application framework based on Microsoft .NET, was used. Being a social media platform, it was made possible to create an unlimited number of user accounts characterized by a series of personal information, introduced by each individual, such as first name, last name, username, address, contact details, profile photo.
Activity 4.2.2. Creating the methodology for testing and evaluating the functions and modules of the online social network with facial recognition using augmented reality - AR Media
For testing and evaluating the functions and modules of the online social network with facial recognition using augmented reality, the project specialists’ team chose qualitative (exploratory) research, namely a research based on the focus group method and the in-depth interview method.
For the focus group, the following methodological aspects were considered:
1. Design and instrumentation:
- studying a bibliography on the given topic;
- studying some materials about the characteristics of the populationconsidered for the research;
- designing the interview guide
2. Recruitment and selection of participants
- Recruitment is foreseen to be done on a voluntary basis;
- Students from different levels (bachelor and master) and teachers are considered;
- group size: 5-12 participants.
3. Selection of the moderator - the moderator selected by the coordinator is Alina Tecău, assoc.prof.PhD, member of the research team;
4. The logistic preparation: it is realized by the researchers engaged in the project and by the moderator (the space must offer the necessary conditions: silence, warmth, light, arrangement, round table)
For the in-depth interview, it was decided to design and conduct 8 interviews, among the specialists from various fields of activity, users of social networks, in order to know their opinions regarding augmented reality applications integrated in social networks. Consideration was given to a classification of in-depth interviews, given by the types of interviews that cover a wide range of possibilities, considering various descriptions, of which we mention: clinical interview, free interview, focused interview, non-direct interview, extended interview, structured interview etc. Generally, these interviews are in a category that can be described as less structured or more intensive, compared to a standardized questionnaire or a quantitative interview.
In order to achieve the objectives, specific to this research approach, it was considered appropriate to choose the method of semi-structured in-depth interview, so as to allow determining the interviewees' opinions to the key aspects pursued in this research in a more detailed manner.
Activity 4.2.3. Determining the functionality, usability and design problems of the online social network with facial recognition using augmented reality AR Media
In accordance with the presented methodology, a marketing study based on a qualitative research using group interview was conducted, which consisted in the organization and deployment of three focus groups among the users of social networks and a qualitative research based on the in-depth interview method.
Focus group 1 - 8 participants, young people aged 20 to 24. A number of 4 women and 4 men volunteered. This category of subjects was chosen due to the fact that their inclination towards the use of social networks is recognized. The selection criterion was subjective, based on the students' willingness to participate in the research. The participants in the focus-group were students in the third year or master's programs in the Faculty of Economic Sciences and Business Administration (study programs: Marketing, Accounting and Management Informatics, Business Administration in Tourism) and the Faculty of Mathematics and Computer Science, within Transilvania University of Brasov.
Focus Group 2 - 10 participants, young people aged 21 to 23. A number of 7 women and 3 men participated voluntarily, thus ensuring the heterogeneity of the participating groups. The participants in this study are students in the third year, from the Faculty of Economic Sciences and Business Administration, study program of Economy of Commerce, Tourism and Services. This specialization was chosen due to the large number of AR applications present in their field of study.
Focus Group 3 - 7 participants, aged 30 to 45, selected from the teaching staff of the Transilvania University of Brașov. The heterogeneity of the group was ensured by the fact that 6 women and 1 man participated voluntarily, as well as the fact that the participants are teachers from different fields, with different research interests. This category of subjects was chosen due to the fact that their inclination towards the use of social networks is recognized, but also because of the desire to research the opinion of professionals in the educational field regarding the aspects pursued as objectives in the present research.
For the research based on the structured interview method, the research approach started from the desire to know the opinion of the subjects regarding the following aspects:
1. Reasons, criteria and options for choosing social networks
2. The degree of knowledge of AR-based applications
3. Opportunity to integrate new AR technologies into social networks
4. Problems of functionality, usability and design of the online social network with facial recognition using the augmented reality application - AR Media, developed within the FUTUREWEB project
5. Proposals to improve the augmented reality application -AR Media, developed within the FUTUREWEB project.
These have also become the objectives of the proposed interview guide. The objectives were transformed into discussion topics as follows:
Theme 1 - Choosing social networks
Objective: To identify the reasons, criteria and options regarding the choice of social networks
Theme 2 - Knowledge of AR-based applications
Objective: To identify the degree of knowledge of AR applications
Theme 3 - Integration of new AR technologies in social networks
Objective: Identifying the opinions regarding the opportunity of integrating the new AR technologies in the social networks
Theme 4 - Aspects of improving the online social network with facial recognition using augmented reality application - AR Media, developed within the FUTUREWEB project
Objective: Determining the functionality, usability and design problems of the online social network with facial recognition using the augmented reality application - AR Media, developed within the FUTUREWEB project
Topic 5 - Areas of use of the AR Media application, developed within the FUTUREWEB project
Objective: To identify the areas of use for the presented AR application. The research highlights a series of proposals, recommendations made by the interview subjects on the functionality, usability and design issues of the online social network with facial recognition using the augmented reality application - AR Media, developed within the FUTUREWEB project, as well as ideas that could improve this app.
According to the research results, the importance of the efforts made to include AR modern technologies in the social networks is obvious. From the discussions with the respondents, we could identify the reasons, criteria and options regarding the choice of social networks, but also the degree of knowledge of AR-based applications and about the opportunity of integrating new AR-based technologies in social networks. The most important part of the discussions was devoted to discussing the functionality, usability and design issues of the online social network with facial recognition using augmented reality application - AR Media, developed within the FUTUREWEB project. The respondents freely expressed their opinion and proposed a large number of ideas that could contribute to the improvement of this application.
Activity 4.2.4. Realization of specifications for improving the online social media network AR Media and implementation of the necessary changes
Starting from the results of the activity 4-2-1, the prototype of the online social network AR Media and adding the suggestions for improvement resulted from conducting the qualitative research (3 focus groups and 8 in-depth interviews among specialists from different fields), the functional model of the social network with facial recognition was created.
As main functionalities, there were pursued:
1. the possibility of adding an unlimited number of members;
2. the possibility of adding the personal data of the members;
3. the possibility of sending personal messages between members;
4. the possibility of posting public messages with text and images to all members of the social network;
5. the possibility to evaluate posts with LIKE;
6. the possibility of tracking the activity of a member;
7. the possibility of creating a group of friends;
8. the possibility to manage personal profile, friends, messages received, etc.
9. identification of persons based on the information contained in the database with the members of the social network.
To this end, the open source DNN platform, a web content management system and web application framework based on Microsoft .NET, was used.
The functioning of the facial recognition process involves the following steps:
1. Capture the image
2. Identifying the face of the person
3. Feature extraction and model generation
4. Comparison of models
5. Declaration of identity
Functions used in the facial recognition application:
Within the facial recognition application we have the following categories of functions:
1. Basic functions required in the training and recognition processes
2. Secondary functions necessary for integration and operation within the computer system
1. The basic functions are:
- Creating the database with images used to recognize people;
- The training function by which the images in the database are associated with a person. For training, we use the OpenCV open-source module, specialized in real-time computer vision;
- The recognition function by which the associations in the database are used to sequentially browse the photographic files in which the recognition will be made.
2. The secondary functions required for integration and operation within the system are:
- Upload function, which allows uploading files to the database and associating with the persons’ names;
- The photo downloads function where the people recognized by the algorithms of the basic functions are marked;
- The function of downloading information related to recognized persons such as accuracy, name and location coordinates in the picture. This information will be integrated and used in the projected social media network.
The face recognition module is based on a series of mathematical algorithms of similarity between the profile picture of the members and the faces of the persons identified in the photographs. This recognition module was created in the Python environment and integrated into the created social media platform. Basically, for every face of a person identified in the pictures, the algorithm searches that person inside the database. Obviously, the search is done only among the members who uploaded their profile picture. If the person is not identified in the database, the text UNKNOWN appears instead of the name.
Activity 4-3-1. Development of the service of visualization of the emotional state according to the last posts and photos of the users of the online social network with facial recognition using augmented reality - AR Media
The function of visualization of the emotional state comes as a complement to the function of facial recognition. Basically, after the person in an image is recognized with a person in the database, the features of the face are analyzed. The feelings of joy, sadness, neutrality and nervousness are recognized based on mathematical algorithms that analyze over 180 points of the face. For each face it calculates percentages for these fundamental feelings which are then displayed.
Activity 4-3-2. Development of the service for viewing and sharing favorite photos and videous of users registered on the AR Media social network using visual techniques and effects specific to augmented reality
The process of improving the application was completed with the development of the service of viewing and sharing the favorite photos and videous of the users registered in the AR Media social network using techniques and visual effects specific to augmented reality. To view and share photos within the social network, a posting module has been created that allows the addition of image files as well as their distribution to the other network members.
SUB-PROJECT 5: Experimental development of a ontology specific to the Romanian language and testing the effects of searches based on the semantic web inside online social networks ”(Semantic Media)
Activity 5-2-1: Formulating the functional and non-functional requirements of the ontology specific to the Romanian language
A clear definition of functional and non-functional requirements is an essential condition for the successful development of any software product. For the FutureWeb ontology, six functional requirements and twelve non-functional requirements were defined. Functional requirements define the functionalities of a system or its components, each functionality being described by specifying the behavior of the system between outputs and inputs. Thus, in the process of developing the proposed ontology, the following functional requirements were considered: CF1. Ontology will allow to identify general concepts in relation to a certain concept; CF2. Ontology will allow to identify specific concepts for a particular concept; CF3. The ontology will allow the identification of the concepts related to a certain concept; CF4. Ontology will allow to identify concepts using different terms; CF5. The ontology will facilitate the identification of concepts using the usual abbreviations; CF6. Ontology will allow identification of concepts in the case of common writing errors.
Non-functional requirements take into account various types of constraints imposed on the system, orthogonal to the functional requirements. In the case of ontologies, the non-functional requirements concern the characteristics, qualities or general aspects of the ontology, such as ease of use, scalability and subsequent maintenance, which are not related to its actual content. For the proposed ontology, the following non-functional requirements were considered: CNF1. Use of concepts and properties from ontologies recognized or standardized by W3C; CNF2. The ontology should be able to be extended to other languages; CNF3. Using a modular structure; CNF4. Compliance with conventions for the name of classes and properties; CNF5. Ontology versioning; CNF6. Ontology description using Turtle syntax; CNF7. Use of prefixes for ontology and its modules; CNF8. Separation between TBox and ABox; CNF9. Declaration of classes as disjunct; CNF10. Declaration from inverse properties; CNF11. Including labels and descriptions for the defined classes and properties; CNF12. Avoiding inconsistencies in ontology.
Activity 5-2-2: Development of an experimental model (prototype) of the ontology specific to the Romanian language for an online social network
The development of the ontology prototype aimed at meeting the functional and non-functional requirements, taking into account the good practices for creating ontologies and the methodologies from the literature. The developed ontology aims to model on the one hand the concepts relevant to the FutureWeb social network, such as user accounts or published messages, and on the other hand the notions in the field of software development, which are of interest to students. The proposed ontology is named for the rest of the document as FutureWeb and uses the prefix fw as well as the Uniform Resource Identifier (URI). Taking into account the good practices in the field, we opted for the stepped development of the ontology: Step 1. Setting the domain and objectives of the ontology; Step 2. Analysis of the existing ontologies; Step 3. A list of terms that will appear in the ontology; Step 4. Definition of classes and class hierarchy; Step 5. Definition of class properties; Step 6. Defining the types of properties; Step 7. Creating instances.
During the development of the experimental ontology model, all the functional and non-functional requirements were verified.
Activity 5-2-3: Creating a testing and evaluation methodology of the developed prototype
In the framework of the methodology of testing and evaluating the developed ontology prototype, the following necessary steps have been identified: Step 1. Making a test file that includes questions regarding the correctness of the results returned by the platform by comparing these with the data from the ontology. The test sheet will include the explanation of the purpose of the platform, elements such as: the key concepts to be tested, the necessary steps to be followed, the expected results for each key concept and for each functionality, the way of recording the fulfillment of each tested element and the description of situations where the platform does not behave according to specifications. Step 2. Selecting a team of four people with skills in testing the web applications for testing the platform and completing the evaluation sheet. Step 3. Browsing with the test team members of the evaluation sheet and clarifying any concerns regarding its content. Step 4. Testing of the platform by the test team and recording the answers on the individual test sheets. Step 5. Collecting and analyzing the answers recorded for each item in the test file. Step 6. Make a report on the possible deficiencies found in the application and the ways to correct them. If inconsistencies between the expected results and those obtained for at least one of the elements considered in the test file are identified, this will be specified in the test report. Step 7. Remedy of any deficiencies found and testing again to validate their correction.
In order to test and validate the ontology prototype, a web application developed in the C# language was implemented, using the Microsoft ASP.NET Core platform, which implements the Model-View-Controller pattern. The server Apache Jena Fuseki was used to run SPARQL queries. The testing of any inconsistencies in declaring classes and properties has been carried out since the development stage by using an inference engine.
Activity 5-2-4: Experimental testing of searches based on the semantic web for the online social network after the development of the ontology specific to the Romanian language
For the experimental testing, an evaluation sheet organized into two main sections was made, a section aimed at testing and evaluating in case of searching for a single concept, and a section dedicated to testing and evaluating in case of searching for several concepts. For each test, the path steps were specified and the results obtained were recorded. The results obtained from the test using four users were included in a centralizing table. For the case of searching for a single concept, there were verified aspects such as: Testing the identification of concepts using the name; Testing the identification of concepts using the acronyms; Testing the identification of the concepts using different terms; Testing the identification of the concepts in the case of common errors; Testing the identification of the concepts using the singular/plural terms; Testing the identification of concepts using articulated terms; Testing the identification of concepts using plural articulated terms; Testing of display of concepts with a higher degree of generality; Testing of display of concepts with a lower degree of generality, Testing of display of similar concepts. For the case of searching for several concepts, there were verified aspects such as: Testing the identification of concepts using the name; Testing the identification of the concepts in case of using the connecting words; Testing the identification of the concepts using acronyms; Testing the display of the concepts with a higher degree of generality in case of searching for some similar concepts; Testing the display of concepts with a higher degree of generality if one of the concepts has a higher degree of generality compared to the other concepts; Testing the display of concepts with a higher degree of generality in the case of searching for several different concepts.
Activity 5-2-5: Conducting a qualitative research to determine the functionality and utility problems of the ontology for the online social network
To determine the functionality and utility problems of the ontology six studies were performed. The first study aimed at understanding the users' opinion about the deficiencies present in classic search engines, regarding the advantages offered by a search engine semantics, as well as highlighting the level of importance of the additional functionalities that a semantic search engine can provide them. The second study aimed to understand the users' opinion on the usefulness and the necessary improvements for the functionality of determining the concepts with a higher degree of generality, implemented in the developed prototype. The third study aimed at understanding the users' opinion regarding the usefulness and the necessary improvements for the functionality of determining the concepts with a lower degree of generality. The fourth study aimed at understanding the opinion of users regarding the functionality of identifying concepts similar to the concept sought by the user, implemented in the developed prototype. The fifth study aimed to understand the users' opinion regarding the functionality of identifying the concepts starting from the use of abbreviations. The sixth study aimed at understanding the opinion of users regarding the ability of the implemented prototype to offer functionalities such as identifying the searched concepts, identifying the concepts with a lower or higher degree of generality, identifying similar concepts, searching for acronyms. The aim was to obtain observations and suggestions for improvement from the participants.
Activity 5-2-6: Creating improvement specifications for the ontology based on the research results
Based on the studies carried out, it became clear the need to extend both the list of concepts and the lists of terms and abbreviations related to them. To remedy the shortcomings observed, two stages will be considered. In the first stage, consideration will be given to including in FutureWeb ontology the concepts and abbreviations identified through the studies: Step 1 - Expand the list of concepts by including the mentioned concepts by the respondents in the studies; Step 2 - Expand the list of abbreviations by including the mentioned abbreviations of respondents in the studies; Step 3 - Validate the updated version of the ontology. The second stage involves the identification of new concepts in the fields related to computer science and their inclusion in the FutureWeb ontology: Step 1 - Automatic processing of documents from the respective fields, such as teaching materials (course / seminar); Step 2 - The punctuation marks will then be removed; Step 3 - The text will be divided into individual words; Step 4 - The individual words will then be reduced to the basic form using a stemming algorithm; Step 5 - The technique called TF-IDF (TermFrequency-InverseDocumentFrequency) will then be used to differentiate the words specific to the domain of link words and general terms; Step 6 - Organizing the terms in a hierarchy with the help of experts in the field; Step 7 - Inclusion of new concepts in the developed ontology by creating the corresponding courts and the necessary relations; Step 8 - Validate the updated version of the ontology.
Activity 5-4-3: Broadly dissemination of project results
To disseminate the project results, all the available communication channels were used: Website, Social Media, communications within each partner faculty, as well as workshops and conferences. A scientific article with an impact factor 0.3 was published:
• C. Delcea, L.A. Cotfas, C. L. Trică, L. Crăciun, and A. G. Molanescu, “Modeling the Consumers Opinion Influence in Online Social Media in the Case of Eco-friendly Products”, Sustainability, vol. 11, no. 6, p. 1796, Jan. 2019.
Based on the participation in the conferences, the following articles were presented and published:
• Liviu-Adrian COTFAS, Mihai ORZAN, Camelia DELCEA, Chuanmin MI - Corporate Social Responsibility Evaluation on Social Media using Machine Learning and Semantic Web, 29th EBES Conference, Lisbon, 2019.
• Camelia DELCEA, Liviu-Adrian COTFAS, Rafal Mierzwiak, Mihai ORZAN – Consumers Contagion in Online Social Networks Regarding Recycling Habits, 29th EBES Conference, Lisbon, 2019.
• Liviu-Adrian COTFAS, Ionut Costinel NICA - Uncovering Social Media Users Emotions towards Companies using Semantic Web Technologies, 28th EBES Conference, Coventry, 2019.
• Camelia DELCEA, Liviu-Adrian COTFAS, Ionut Costinel NICA - Analyzing Customers’Opinions towards Product Characteristics using Social Media, 28th EBES Conference, Coventry, 2019.
• Liviu-Adrian COTFAS, Ioan ROXIN, Camelia DELCEA - SEMANTIC SEARCH IN SOCIAL MEDIA ANALYSIS. In: Proceedings of the 18th International Conference on Conference on Informatics in Economy (IE 2019), Bucharest University of Economic Studies Press, Bucharest, 2019
• Alexandra Cristina Dinu, Raluca Giorgiana Chivu, Alexandru Valentin Teodorov, Otilia Platon, Gheorghe ORZAN - The impact of semantic web in user-machine interaction, SCM 4 ECR Technology and Innovation in Supply Chain Management for Creating New Value for Consumers, Valahia University of Târgovişte, 2019
• Alexandra Cristina Dinu, Violeta Rădulescu, Anca – Francisca Cruceru & Mihai Orzan, Necessity for Semantic Web development in user interaction. Emerging Trends in Marketing and Management International Conference (ETIMM), 4, Bucuresti, 2019
• Alexandra Cristina Dinu, Violeta Rădulescu, Anca – Francisca Cruceru & Mihai Orzan, How Semantic Web Can be Used in Better Machine Decision Making, 34th IBIMA Conference: 13-14 November 2019, Madrid, Spain/