Prof. Dr. Ayad Tareq Imam (also Known as Ayad Al-Zobaydi) received his PhD degree in Computer Science / Artificial Intelligence from De Montfort University, Leicester, the UK in 2010, M.Sc. in Computer Science / Natural Language Processing from AlNahrain University in 1994, and B.Sc. in Computer Science from the University of Mosul in 1989. Currently, Dr. Ayad is a Proffesor at Al-Isra University / Amman / Jordan. Dr. Ayad has several published papers on various computer science and Software Engineering topics. Dr. Ayad is a reviewer in several journals and conferences on Computer and Information related areas.
In addition to using AI approaches and techniques to solve sorts of computing problems, my Interest in research includes the development of Intelligent CASE (I-CASE) tools, especially the use of Natural Language Processing (NLP) in the requirement engineering.
Proceedings of the 1st International Conference on Computing and Emerging Sciences - ICCES, 2020
This paper aims to articulate the current approaches to handle the uncertainty problem in data mi... more This paper aims to articulate the current approaches to handle the uncertainty problem in data mining (DM). The difficulties in DM are given to show the impact of the uncertainty problem in DM's various applications. As it has been stated that is no common DM method to handle the uncertainty problem in DM and based upon the literature of cognitive science, this paper highlights a new classification approach to overcome the uncertainty problem in DM that is the Relative-Fuzzy (RF) approach and its ML/RFL-Based Net software tool.
A mobile agent is a mobile object, whose movement requires a wise determination of the best path,... more A mobile agent is a mobile object, whose movement requires a wise determination of the best path, as this path is applied to migrate a mobile agent from a source node to a destination node in an interconnected network of computers. Usually, the choice of the best path is made using optimization algorithms, which use the minimum time as a key measure for choosing the best path. This work proposes a new complex approach, which is called The Genetic Algorithm with Node Compression-based Search Algorithm (GANCSA). This approach utilizes a mathematical process and an optimization technique to achieve the finding of the best migration path in minimal time (sequential nodes in a time frame) for migrating a mobile agent in an interconnected network of computers. GANCSA encompasses the Genetic Algorithm (GA) as an optimization technique, and the Node Compression-based Search Algorithm (NCSA) to minimize the number of nodes. The testing of the proposed GANCSA shows that the combination of GA and NCSA reduces the time of selecting the best path from 336.448 ms to 286.29 ms after 10 iterations, believing that more time reduction can be achieved by increasing the iterations.
Elicitation of the elements of Unified Modelling Language (UML) analysis and design models from s... more Elicitation of the elements of Unified Modelling Language (UML) analysis and design models from sentences written in scripted English is essential in the production of analysis and design models. The correct elicitation of these elements depends on the intuitive, manually defined set of linguistic heuristics, which is used to map a word in the sentence to its correct semantics in the domain of UML analysis and design models. This paper proposes a Genetic Algorithm-based classification rule discovery approach and a developed Enhanced Intuitive Linguistic Heuristics (EILH) dataset to automate the definition of the intuitive linguistic heuristics set to elicit five elements of UML analysis and design models from English sentences. These elements are the use case, the actor, the sender, the receiver, and the message. The automatically defined intuitive linguistic heuristics set was evaluated by developing an Artificial Neural Network (ANN) to recognize the elements of the UML analysis and design models using both manually defined and automatically defined sets. This comparison shows the superiority of the automatically defined set over the manually defined one.
This study aims to explore how empowerment is enabled in Web-based project teams. It also aims to... more This study aims to explore how empowerment is enabled in Web-based project teams. It also aims to identify differences in empowering practices and levels of individual empowerment in different type...
Journal of King Saud University - Computer and Information Sciences, 2022
In Software Engineering (SE), metrics are used for detecting software design problems (bad smells... more In Software Engineering (SE), metrics are used for detecting software design problems (bad smells) like the large-class bad smell, where a lot of different metrics were defined to find out the existence of this problem in the design of a class. Examples of these metrics are size metrics, cohesion metrics, and coupling metrics. Selecting the right metrics to detect the large-class bad smell is a common problem, and it is usually accomplished manually. The questions remain: Can a module with the best combination of two metrics, for detecting the problem of large-class bad smell, be formed automatically rather than manually? And how is this double valued threshold determined to be used to infer the existence of this problem? This paper proposes the Hybrid Approach to detect Large Class Bad Smell (HA-LCBS). This approach utilizes the Genetic Algorithm (GA) to automate the composing of a detecting module that consists of a cohesion metric type and a coupling metric type and passes its resulting paired value to a deep learning approach to automate the detection of the large class bad smell. The accuracy that has been gained from using this approach reached 94.21%.
The research is an attempt to design a Triligual (Arabic, English, and Kurdish) computer system b... more The research is an attempt to design a Triligual (Arabic, English, and Kurdish) computer system by changing three I/O device drivers simultaneously; i.e. a user can change the computer language between these three languages by pressing the appropriate hotkeys. Such a system is very useful with application programs such as word processors. The system has been implemented on an IBM-PC compatible using Turbo Pascal ver. 6.0, assembly macros and a tool program called MFont.
International Review on Computers and Software, Aug 20, 2014
In this paper, we report the development of an Arabic natural language interface to robot (ANLI2R... more In this paper, we report the development of an Arabic natural language interface to robot (ANLI2ROBOT). The imperative sentence in Arabic language is used to command a robot lifting arm to grasp small metal objects and move them from one place to another. Since this interface is a natural language processing (NLP) application, lexicon, syntactical, and morphological subsystems for the Arabic language interrogative sentence were created as components of this interface. A set of simple familiar Arabic command words, which are related to the operations of the robot lifting arm, have been defined in the lexicon of the interface, and different forms of these words have been created by using the morphological subsystem. ANLI2ROBOT is implemented using Prolog programming language since it gives a straightforward conversion from Definite Clause Grammar (DCG) to Prolog predicates. This work, which is classified as human–robot interaction (HRI), contributes the using of Arabic language to communicate with robot. ANLI2ROBOT is also considered as a written dialogue system (WDS) that uses Arabic written (not spoken) language. In addition to the success in handling the different forms of the imperative sentence, the work revealed other significant findings regarding the use of the Arabic natural language to instruct the robot (and similar machines).
Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has ... more Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has been gaining wide interest in recent years as a tool for processing data. Different ANN architectures have been defined for various applications. Yet, a number of difficulties existed when building an ANN like training time, over-training problem, retraining of the net for new types of data. This paper presents a new ANN architecture, which we called "MS/Ayad-Marwan Network" (MS stands for Multi-Stage), that compose of supernet and subnets ("Stage" architecture). The "Computer-Assisted" approach is used to develop this NN. Also, in this paper, a new learning algorithm has been proposed, which is required for adjusting weighting values for MS/Ayad-Marwan neural network. We called this algorithm a "hybrid unsupervised/supervised learning paradigm" This learning algorithm is used for training both supernet and subnets. Inheritance property of OOP plays an important role in the reasoning approach of MS/Ayad-Marwan Network, in that it deduces the final solution by gathering the property of the super network with the properties of its sub(s), to reach the ultimate goal. Thus, the results become more accurate. We called his reasoning approach " Deductive auto-associative recalling". MS/Ayad-Marwan ANN has been applied to an Automatic Speech Recognition (ASR) application. This new network architecture is being able to recognize phonemes with >91% accuracy. We note through the results a promising advancement has been reached and hence a Multi-Stage strategy is useful in applications of "high changeable data" property, like ASR. "MS/Ayad-Marwan" Neural Network has been implemented and tested using MATLAB, as a part of the ASR system.
Sequence diagram is used to illustrate the exchanging of information between users and system'... more Sequence diagram is used to illustrate the exchanging of information between users and system's components. Sequence diagram helps in leveling up the requirement description. The analyst is the role who is manually performing the developing of sequence diagram. The skill of defining objects (or roles) and their exchanged messages are highly depending on the user's expert to recognize sequence diagram units from different sources of information. It is very hard to find UML's reference that describes enough direct rules (specially for many trainee developers) to develop sequence diagram. This paper contributes a definition of a systematic a semi-automated algorithmic approach for generating sequence diagram. Flow of events included in a use case model is used as an input. Semantic Role Labeling (SRL) is the process used to discover the thematic role, which is used to determine the elements of sequence diagram. The extracted elements of sequence diagram are to be fed to a ready-made software tool for sketching sequence diagram. This algorithmic approach provides a direct starting, clear linguistic process, and easy-to-follow practice for developing sequence diagram. The semi-automated approach is viewed as a step towards developing an Intelligent Computer Aided Software Engineering (I-CASE) tool for sketching sequence diagram from flow of events of a use case model.
Proceedings of the International Conference on Advances in Image Processing, 2017
Sequence diagram is used to illustrate the exchanging of information between users and system'... more Sequence diagram is used to illustrate the exchanging of information between users and system's components. Sequence diagram helps in leveling up the requirement description. The analyst is the role who is manually performing the developing of sequence diagram. The skill of defining objects (or roles) and their exchanged messages are highly depending on the user's expert to recognize sequence diagram units from different sources of information. It is very hard to find UML's reference that describes enough direct rules (specially for many trainee developers) to develop sequence diagram. This paper contributes a definition of a systematic a semi-automated algorithmic approach for generating sequence diagram. Flow of events included in a use case model is used as an input. Semantic Role Labeling (SRL) is the process used to discover the thematic role, which is used to determine the elements of sequence diagram. The extracted elements of sequence diagram are to be fed to a ready-made software tool for sketching sequence diagram. This algorithmic approach provides a direct starting, clear linguistic process, and easy-to-follow practice for developing sequence diagram. The semi-automated approach is viewed as a step towards developing an Intelligent Computer Aided Software Engineering (I-CASE) tool for sketching sequence diagram from flow of events of a use case model.
Proceedings of the International Conference on Geoinformatics and Data Analysis, 2018
Use case is a model delivered by requirements engineering phase, which is considered as an input ... more Use case is a model delivered by requirements engineering phase, which is considered as an input to the forthcoming design phase and test phase. A use case model is a simplest representation of an actor's interactions with the system in which the user is involved. The development of a use case model requires the finding out the use case itself and the actor that uses this use case to interact with the system. These two tasks are achieved manually via analyst's experience, who starts with different sources of data to develop use case model. User requirements document is a common source of data that may be started with to develop use case model. The extracting of actors and their actions (use cases) is subjected to the linguistic properties of each on. The aim of this paper is to define a new algorithmic approach for extracting actors and their use cases by using thematic role technique. This algorithmic approach had been manually tested using known examples, and shown its validity. The success of this technique will lead to develop an Intelligent Computer Aided Software Engineering (I-CASE) tool that automatically extracts actions and actors of use case model from functional requirements by using Semantic Role Labelling (SRL) of Natural Language Processing (NLP) approach.
2014 5th International Conference on Information and Communication Systems (ICICS), 2014
ABSTRACT This paper aims to demonstrate the development of an expert code generator using rule-ba... more ABSTRACT This paper aims to demonstrate the development of an expert code generator using rule-based and frames knowledge representation techniques (ECG-RF). The ECG-RF system presented in this paper is a passive code generator that carries out the task of automatic code generation in fixed-structure software. To develop an ECG-RF system, the artificial intelligence (AI) of rule-based system and frames knowledge representation techniques was applied to a code generation task. ECG-RF fills a predefined frame of a certain fixed-structure program with code chunks retrieved from ECG-RF's knowledge base. The filling operation is achieved by ECG-RF's inference engine and is guided by the information collected from the user via a graphic user interface (GUI). In this paper, an ECG-RF system for generating a device driver program is presented and implemented with VBasic software. The results show that the ECG-RF design concept is reasonably reliable.
Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has ... more Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has been gaining wide interest in recent years as a tool for processing data. Different ANN architectures have been defined for various applications. Yet, a number of difficulties existed when building an ANN like training time, over-training problem, retraining of the net for new types of data. This paper presents a new ANN architecture, which we called "MS/Ayad-Marwan Network" (MS stands for Multi-Stage), that compose of supernet and subnets ("Stage" architecture). The "Computer-Assisted" approach is used to develop this NN. Also, in this paper, a new learning algorithm has been proposed, which is required for adjusting weighting values for MS/Ayad-Marwan neural network. We called this algorithm a "hybrid unsupervised/supervised learning paradigm" This learning algorithm is used for training both supernet and subnets. Inheritance property of OOP plays an important role in the reasoning approach of MS/Ayad-Marwan Network, in that it deduces the final solution by gathering the property of the super network with the properties of its sub(s), to reach the ultimate goal. Thus, the results become more accurate. We called his reasoning approach " Deductive auto-associative recalling". MS/Ayad-Marwan ANN has been applied to an Automatic Speech Recognition (ASR) application. This new network architecture is being able to recognize phonemes with >91% accuracy. We note through the results a promising advancement has been reached and hence a Multi-Stage strategy is useful in applications of "high changeable data" property, like ASR. "MS/Ayad-Marwan" Neural Network has been implemented and tested using MATLAB, as a part of the ASR system.
The paper is not for recognizing normal formed speech but for distorted speech via examining the ... more The paper is not for recognizing normal formed speech but for distorted speech via examining the ability of feed forward Artificial Neural Networks (ANN) to recognize speech flaws. In this paper we take the Arabic /r/ phoneme distortion that is somewhat common among native speakers as a case study.To do this, r-Distype program is developed as a script written using Praat speech processing software tool. r-Distype program automatically develops a feed forward ANN that tests the spoken word (which includes /r/ phoneme) to detect any possible type of distortion. Multiple feed forward ANNs of different architectures have been developed and their achievements reported. Training data and testing data of the developed ANNs are sets of spoken Arabic words that contain /r/ phoneme in different positions so they cover all distortion types of Arabic /r/ phoneme. These sets of words were produced by different genders and different ages.The results obtained from developed ANNs were used to draw ...
Proceedings of the 1st International Conference on Computing and Emerging Sciences - ICCES, 2020
This paper aims to articulate the current approaches to handle the uncertainty problem in data mi... more This paper aims to articulate the current approaches to handle the uncertainty problem in data mining (DM). The difficulties in DM are given to show the impact of the uncertainty problem in DM's various applications. As it has been stated that is no common DM method to handle the uncertainty problem in DM and based upon the literature of cognitive science, this paper highlights a new classification approach to overcome the uncertainty problem in DM that is the Relative-Fuzzy (RF) approach and its ML/RFL-Based Net software tool.
A mobile agent is a mobile object, whose movement requires a wise determination of the best path,... more A mobile agent is a mobile object, whose movement requires a wise determination of the best path, as this path is applied to migrate a mobile agent from a source node to a destination node in an interconnected network of computers. Usually, the choice of the best path is made using optimization algorithms, which use the minimum time as a key measure for choosing the best path. This work proposes a new complex approach, which is called The Genetic Algorithm with Node Compression-based Search Algorithm (GANCSA). This approach utilizes a mathematical process and an optimization technique to achieve the finding of the best migration path in minimal time (sequential nodes in a time frame) for migrating a mobile agent in an interconnected network of computers. GANCSA encompasses the Genetic Algorithm (GA) as an optimization technique, and the Node Compression-based Search Algorithm (NCSA) to minimize the number of nodes. The testing of the proposed GANCSA shows that the combination of GA and NCSA reduces the time of selecting the best path from 336.448 ms to 286.29 ms after 10 iterations, believing that more time reduction can be achieved by increasing the iterations.
Elicitation of the elements of Unified Modelling Language (UML) analysis and design models from s... more Elicitation of the elements of Unified Modelling Language (UML) analysis and design models from sentences written in scripted English is essential in the production of analysis and design models. The correct elicitation of these elements depends on the intuitive, manually defined set of linguistic heuristics, which is used to map a word in the sentence to its correct semantics in the domain of UML analysis and design models. This paper proposes a Genetic Algorithm-based classification rule discovery approach and a developed Enhanced Intuitive Linguistic Heuristics (EILH) dataset to automate the definition of the intuitive linguistic heuristics set to elicit five elements of UML analysis and design models from English sentences. These elements are the use case, the actor, the sender, the receiver, and the message. The automatically defined intuitive linguistic heuristics set was evaluated by developing an Artificial Neural Network (ANN) to recognize the elements of the UML analysis and design models using both manually defined and automatically defined sets. This comparison shows the superiority of the automatically defined set over the manually defined one.
This study aims to explore how empowerment is enabled in Web-based project teams. It also aims to... more This study aims to explore how empowerment is enabled in Web-based project teams. It also aims to identify differences in empowering practices and levels of individual empowerment in different type...
Journal of King Saud University - Computer and Information Sciences, 2022
In Software Engineering (SE), metrics are used for detecting software design problems (bad smells... more In Software Engineering (SE), metrics are used for detecting software design problems (bad smells) like the large-class bad smell, where a lot of different metrics were defined to find out the existence of this problem in the design of a class. Examples of these metrics are size metrics, cohesion metrics, and coupling metrics. Selecting the right metrics to detect the large-class bad smell is a common problem, and it is usually accomplished manually. The questions remain: Can a module with the best combination of two metrics, for detecting the problem of large-class bad smell, be formed automatically rather than manually? And how is this double valued threshold determined to be used to infer the existence of this problem? This paper proposes the Hybrid Approach to detect Large Class Bad Smell (HA-LCBS). This approach utilizes the Genetic Algorithm (GA) to automate the composing of a detecting module that consists of a cohesion metric type and a coupling metric type and passes its resulting paired value to a deep learning approach to automate the detection of the large class bad smell. The accuracy that has been gained from using this approach reached 94.21%.
The research is an attempt to design a Triligual (Arabic, English, and Kurdish) computer system b... more The research is an attempt to design a Triligual (Arabic, English, and Kurdish) computer system by changing three I/O device drivers simultaneously; i.e. a user can change the computer language between these three languages by pressing the appropriate hotkeys. Such a system is very useful with application programs such as word processors. The system has been implemented on an IBM-PC compatible using Turbo Pascal ver. 6.0, assembly macros and a tool program called MFont.
International Review on Computers and Software, Aug 20, 2014
In this paper, we report the development of an Arabic natural language interface to robot (ANLI2R... more In this paper, we report the development of an Arabic natural language interface to robot (ANLI2ROBOT). The imperative sentence in Arabic language is used to command a robot lifting arm to grasp small metal objects and move them from one place to another. Since this interface is a natural language processing (NLP) application, lexicon, syntactical, and morphological subsystems for the Arabic language interrogative sentence were created as components of this interface. A set of simple familiar Arabic command words, which are related to the operations of the robot lifting arm, have been defined in the lexicon of the interface, and different forms of these words have been created by using the morphological subsystem. ANLI2ROBOT is implemented using Prolog programming language since it gives a straightforward conversion from Definite Clause Grammar (DCG) to Prolog predicates. This work, which is classified as human–robot interaction (HRI), contributes the using of Arabic language to communicate with robot. ANLI2ROBOT is also considered as a written dialogue system (WDS) that uses Arabic written (not spoken) language. In addition to the success in handling the different forms of the imperative sentence, the work revealed other significant findings regarding the use of the Arabic natural language to instruct the robot (and similar machines).
Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has ... more Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has been gaining wide interest in recent years as a tool for processing data. Different ANN architectures have been defined for various applications. Yet, a number of difficulties existed when building an ANN like training time, over-training problem, retraining of the net for new types of data. This paper presents a new ANN architecture, which we called "MS/Ayad-Marwan Network" (MS stands for Multi-Stage), that compose of supernet and subnets ("Stage" architecture). The "Computer-Assisted" approach is used to develop this NN. Also, in this paper, a new learning algorithm has been proposed, which is required for adjusting weighting values for MS/Ayad-Marwan neural network. We called this algorithm a "hybrid unsupervised/supervised learning paradigm" This learning algorithm is used for training both supernet and subnets. Inheritance property of OOP plays an important role in the reasoning approach of MS/Ayad-Marwan Network, in that it deduces the final solution by gathering the property of the super network with the properties of its sub(s), to reach the ultimate goal. Thus, the results become more accurate. We called his reasoning approach " Deductive auto-associative recalling". MS/Ayad-Marwan ANN has been applied to an Automatic Speech Recognition (ASR) application. This new network architecture is being able to recognize phonemes with >91% accuracy. We note through the results a promising advancement has been reached and hence a Multi-Stage strategy is useful in applications of "high changeable data" property, like ASR. "MS/Ayad-Marwan" Neural Network has been implemented and tested using MATLAB, as a part of the ASR system.
Sequence diagram is used to illustrate the exchanging of information between users and system'... more Sequence diagram is used to illustrate the exchanging of information between users and system's components. Sequence diagram helps in leveling up the requirement description. The analyst is the role who is manually performing the developing of sequence diagram. The skill of defining objects (or roles) and their exchanged messages are highly depending on the user's expert to recognize sequence diagram units from different sources of information. It is very hard to find UML's reference that describes enough direct rules (specially for many trainee developers) to develop sequence diagram. This paper contributes a definition of a systematic a semi-automated algorithmic approach for generating sequence diagram. Flow of events included in a use case model is used as an input. Semantic Role Labeling (SRL) is the process used to discover the thematic role, which is used to determine the elements of sequence diagram. The extracted elements of sequence diagram are to be fed to a ready-made software tool for sketching sequence diagram. This algorithmic approach provides a direct starting, clear linguistic process, and easy-to-follow practice for developing sequence diagram. The semi-automated approach is viewed as a step towards developing an Intelligent Computer Aided Software Engineering (I-CASE) tool for sketching sequence diagram from flow of events of a use case model.
Proceedings of the International Conference on Advances in Image Processing, 2017
Sequence diagram is used to illustrate the exchanging of information between users and system'... more Sequence diagram is used to illustrate the exchanging of information between users and system's components. Sequence diagram helps in leveling up the requirement description. The analyst is the role who is manually performing the developing of sequence diagram. The skill of defining objects (or roles) and their exchanged messages are highly depending on the user's expert to recognize sequence diagram units from different sources of information. It is very hard to find UML's reference that describes enough direct rules (specially for many trainee developers) to develop sequence diagram. This paper contributes a definition of a systematic a semi-automated algorithmic approach for generating sequence diagram. Flow of events included in a use case model is used as an input. Semantic Role Labeling (SRL) is the process used to discover the thematic role, which is used to determine the elements of sequence diagram. The extracted elements of sequence diagram are to be fed to a ready-made software tool for sketching sequence diagram. This algorithmic approach provides a direct starting, clear linguistic process, and easy-to-follow practice for developing sequence diagram. The semi-automated approach is viewed as a step towards developing an Intelligent Computer Aided Software Engineering (I-CASE) tool for sketching sequence diagram from flow of events of a use case model.
Proceedings of the International Conference on Geoinformatics and Data Analysis, 2018
Use case is a model delivered by requirements engineering phase, which is considered as an input ... more Use case is a model delivered by requirements engineering phase, which is considered as an input to the forthcoming design phase and test phase. A use case model is a simplest representation of an actor's interactions with the system in which the user is involved. The development of a use case model requires the finding out the use case itself and the actor that uses this use case to interact with the system. These two tasks are achieved manually via analyst's experience, who starts with different sources of data to develop use case model. User requirements document is a common source of data that may be started with to develop use case model. The extracting of actors and their actions (use cases) is subjected to the linguistic properties of each on. The aim of this paper is to define a new algorithmic approach for extracting actors and their use cases by using thematic role technique. This algorithmic approach had been manually tested using known examples, and shown its validity. The success of this technique will lead to develop an Intelligent Computer Aided Software Engineering (I-CASE) tool that automatically extracts actions and actors of use case model from functional requirements by using Semantic Role Labelling (SRL) of Natural Language Processing (NLP) approach.
2014 5th International Conference on Information and Communication Systems (ICICS), 2014
ABSTRACT This paper aims to demonstrate the development of an expert code generator using rule-ba... more ABSTRACT This paper aims to demonstrate the development of an expert code generator using rule-based and frames knowledge representation techniques (ECG-RF). The ECG-RF system presented in this paper is a passive code generator that carries out the task of automatic code generation in fixed-structure software. To develop an ECG-RF system, the artificial intelligence (AI) of rule-based system and frames knowledge representation techniques was applied to a code generation task. ECG-RF fills a predefined frame of a certain fixed-structure program with code chunks retrieved from ECG-RF's knowledge base. The filling operation is achieved by ECG-RF's inference engine and is guided by the information collected from the user via a graphic user interface (GUI). In this paper, an ECG-RF system for generating a device driver program is presented and implemented with VBasic software. The results show that the ECG-RF design concept is reasonably reliable.
Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has ... more Due to its parallelism property in processing the data, the Artificial Neural Networks (ANN) has been gaining wide interest in recent years as a tool for processing data. Different ANN architectures have been defined for various applications. Yet, a number of difficulties existed when building an ANN like training time, over-training problem, retraining of the net for new types of data. This paper presents a new ANN architecture, which we called "MS/Ayad-Marwan Network" (MS stands for Multi-Stage), that compose of supernet and subnets ("Stage" architecture). The "Computer-Assisted" approach is used to develop this NN. Also, in this paper, a new learning algorithm has been proposed, which is required for adjusting weighting values for MS/Ayad-Marwan neural network. We called this algorithm a "hybrid unsupervised/supervised learning paradigm" This learning algorithm is used for training both supernet and subnets. Inheritance property of OOP plays an important role in the reasoning approach of MS/Ayad-Marwan Network, in that it deduces the final solution by gathering the property of the super network with the properties of its sub(s), to reach the ultimate goal. Thus, the results become more accurate. We called his reasoning approach " Deductive auto-associative recalling". MS/Ayad-Marwan ANN has been applied to an Automatic Speech Recognition (ASR) application. This new network architecture is being able to recognize phonemes with >91% accuracy. We note through the results a promising advancement has been reached and hence a Multi-Stage strategy is useful in applications of "high changeable data" property, like ASR. "MS/Ayad-Marwan" Neural Network has been implemented and tested using MATLAB, as a part of the ASR system.
The paper is not for recognizing normal formed speech but for distorted speech via examining the ... more The paper is not for recognizing normal formed speech but for distorted speech via examining the ability of feed forward Artificial Neural Networks (ANN) to recognize speech flaws. In this paper we take the Arabic /r/ phoneme distortion that is somewhat common among native speakers as a case study.To do this, r-Distype program is developed as a script written using Praat speech processing software tool. r-Distype program automatically develops a feed forward ANN that tests the spoken word (which includes /r/ phoneme) to detect any possible type of distortion. Multiple feed forward ANNs of different architectures have been developed and their achievements reported. Training data and testing data of the developed ANNs are sets of spoken Arabic words that contain /r/ phoneme in different positions so they cover all distortion types of Arabic /r/ phoneme. These sets of words were produced by different genders and different ages.The results obtained from developed ANNs were used to draw ...
Uploads
Papers by Ayad T . Imam
The system has been implemented on an IBM-PC compatible using Turbo Pascal ver. 6.0, assembly macros and a tool program called MFont.
The system has been implemented on an IBM-PC compatible using Turbo Pascal ver. 6.0, assembly macros and a tool program called MFont.