Yasser Salem
I am a visiting research fellow in computer science in the School of Electronics, Electrical Engineering and Computer Science at Queen's University Belfast. I am a member of the Knowledge and Data Engineering research cluster within the school.
I received my PhD degree from the Queen's University Belfast in 2016, my PhD thesis entitled, "An Experience-Based Critiquing Approach to Conversational Recommendation" under the guidance of Dr Jun Hong and Prof. Weiru Liu.
I completed a B.Sc. in Computer Science, with First Class honours at the Institute of Technology, Blanchardstown (ITB) in 2007. My BSc project achieved a prize for the best software project in May 2007.
I then completed an M.Sc. by research in the area of computational linguistics and machine translation, in 2009, also at ITB, with a thesis entitled “A Generic Framework for Arabic to English Machine translation of Simplex Sentences using the Role and Reference Grammar Linguistic Model”. This research was the first contribution using the Role and Reference Grammar (RRG) model as a basis for machine translation. My MSc thesis is available on the official Role and Reference Grammar website.
While working on my MSc, I published 6 papers and a book chapter.
I was a reviewer for the International Arab Conference on Information Technology (ACIT 2008), and external reviewer for the 15th Asia-Pacific Web Conference (APWeb 2013). I also delivered an invited talk at Dublin City University (DCU) in July 2008 entitled "UniArab: a universal machine translator system for Arabic based on Role and Reference Grammar".
I received my PhD degree from the Queen's University Belfast in 2016, my PhD thesis entitled, "An Experience-Based Critiquing Approach to Conversational Recommendation" under the guidance of Dr Jun Hong and Prof. Weiru Liu.
I completed a B.Sc. in Computer Science, with First Class honours at the Institute of Technology, Blanchardstown (ITB) in 2007. My BSc project achieved a prize for the best software project in May 2007.
I then completed an M.Sc. by research in the area of computational linguistics and machine translation, in 2009, also at ITB, with a thesis entitled “A Generic Framework for Arabic to English Machine translation of Simplex Sentences using the Role and Reference Grammar Linguistic Model”. This research was the first contribution using the Role and Reference Grammar (RRG) model as a basis for machine translation. My MSc thesis is available on the official Role and Reference Grammar website.
While working on my MSc, I published 6 papers and a book chapter.
I was a reviewer for the International Arab Conference on Information Technology (ACIT 2008), and external reviewer for the 15th Asia-Pacific Web Conference (APWeb 2013). I also delivered an invited talk at Dublin City University (DCU) in July 2008 entitled "UniArab: a universal machine translator system for Arabic based on Role and Reference Grammar".
less
InterestsView All (13)
Uploads
Papers by Yasser Salem
The Arabic language is written from right to left and it has complex, language-specific grammar rules and a relatively free word order. These distinguishing features pose a major challenge in processing Arabic text for linguistic analysis. Our framework demonstrates that RRG is a feasible foundation for building multi-language machine translation [MT] systems. Arabic is a Semitic language originating in the area presently known as the Arabian Peninsula. The Arabic language is one of six major world languages, and one of the six official languages of the United Nations. The version of Arabic we consider in this work is Modern Standard Arabic [MSA]. When we mention Arabic throughout this paper, we mean MSA, which is a distinct, modernized form of Classical Arabic (Alosh 2005). MSA is the universal written language of the Arabic-speaking population, printed in most books, newspapers, magazines, official documents, and reading primers for children. Most of the oral Arabic spoken today is more divergent than the written Arabic language, because of dialectal interference. However, MSA is the literary and standard variety of Arabic used in writing and formal speeches today (Schulz 2005).
In this paper we discuss the RRG UniArab MT research project and the Interlingua model of Arabic MT that we designed and built using Java and XML. With this we discuss the challenges inherent within Arabic MT and the part that RRG played in helping to overcome many of the challenges. The architecture of the lexicon and its design and implementation in XML is discussed, along with a presentation of the results produced by the UniArab software evaluation
Thesis Chapters by Yasser Salem
We focus on critiquing-based recommenders, which allow users to tweak the features of recommended products to refine their needs and preferences. Critiquing-based recommender systems have proven to be an effective approach to conversational recommendation. However they have a tendency to produce protracted recommendation sessions, due to limited feedback that critiques can provide.
We describe a novel approach to reusing past critiquing sessions in order to improve overall recommendation efficiency.
This new critiquing-based approach is capable of using the current user's critiques so far to predict the most likely product recommendations and therefore short-cut sometimes protracted recommendation sessions in standard critiquing approaches. We demonstrate the potential for using this new technique to improve upon a number of state of the art critiquing techniques.
Our approach has the capability of improving both efficiency and quality of recommendation.
Recommending users for a new social network user to follow is a topic of interest at present. The existing approaches rely on using various types of information about the new user to determine recommended users who have similar interests to the new user. However, this presents a cold start problem when a new user joins a social network, who is yet to have any interaction on the social network. We present a particular type of conversational recommendation approach, critiquing-based recommendation, to solve the cold start problem.
A traditional critiquing-based recommendation system allows a user to critique a feature of a recommended item at a time and gradually leads the user to the target recommendation. However this may require a lengthy recommendation session. Our approaches aim to reduce the session length by taking a case-based reasoning approach. It selects relevant recommendation sessions of past users that match the recommendation session of the current user to short-cut the current recommendation session. It selects relevant recommendation sessions from a case base that contains the successful recommendation sessions of past users. A past recommendation session can be selected if it contains recommended items and critiques that sufficiently overlap with the ones in the current session.
Our new techniques show a significant improvement in the interactions between users and recommendation systems. We also show that the new techniques enable satisfactory recommendations made at an earlier stage in the session.
The Arabic language is written from right to left and it has complex, language-specific grammar rules and a relatively free word order. These distinguishing features pose a major challenge in processing Arabic text for linguistic analysis. Our framework demonstrates that RRG is a feasible foundation for building multi-language machine translation [MT] systems. Arabic is a Semitic language originating in the area presently known as the Arabian Peninsula. The Arabic language is one of six major world languages, and one of the six official languages of the United Nations. The version of Arabic we consider in this work is Modern Standard Arabic [MSA]. When we mention Arabic throughout this paper, we mean MSA, which is a distinct, modernized form of Classical Arabic (Alosh 2005). MSA is the universal written language of the Arabic-speaking population, printed in most books, newspapers, magazines, official documents, and reading primers for children. Most of the oral Arabic spoken today is more divergent than the written Arabic language, because of dialectal interference. However, MSA is the literary and standard variety of Arabic used in writing and formal speeches today (Schulz 2005).
In this paper we discuss the RRG UniArab MT research project and the Interlingua model of Arabic MT that we designed and built using Java and XML. With this we discuss the challenges inherent within Arabic MT and the part that RRG played in helping to overcome many of the challenges. The architecture of the lexicon and its design and implementation in XML is discussed, along with a presentation of the results produced by the UniArab software evaluation
We focus on critiquing-based recommenders, which allow users to tweak the features of recommended products to refine their needs and preferences. Critiquing-based recommender systems have proven to be an effective approach to conversational recommendation. However they have a tendency to produce protracted recommendation sessions, due to limited feedback that critiques can provide.
We describe a novel approach to reusing past critiquing sessions in order to improve overall recommendation efficiency.
This new critiquing-based approach is capable of using the current user's critiques so far to predict the most likely product recommendations and therefore short-cut sometimes protracted recommendation sessions in standard critiquing approaches. We demonstrate the potential for using this new technique to improve upon a number of state of the art critiquing techniques.
Our approach has the capability of improving both efficiency and quality of recommendation.
Recommending users for a new social network user to follow is a topic of interest at present. The existing approaches rely on using various types of information about the new user to determine recommended users who have similar interests to the new user. However, this presents a cold start problem when a new user joins a social network, who is yet to have any interaction on the social network. We present a particular type of conversational recommendation approach, critiquing-based recommendation, to solve the cold start problem.
A traditional critiquing-based recommendation system allows a user to critique a feature of a recommended item at a time and gradually leads the user to the target recommendation. However this may require a lengthy recommendation session. Our approaches aim to reduce the session length by taking a case-based reasoning approach. It selects relevant recommendation sessions of past users that match the recommendation session of the current user to short-cut the current recommendation session. It selects relevant recommendation sessions from a case base that contains the successful recommendation sessions of past users. A past recommendation session can be selected if it contains recommended items and critiques that sufficiently overlap with the ones in the current session.
Our new techniques show a significant improvement in the interactions between users and recommendation systems. We also show that the new techniques enable satisfactory recommendations made at an earlier stage in the session.