In order to bring you the best possible user experience, this site uses Javascript. If you are seeing this message, it is likely that the Javascript option in your browser is disabled. For optimal viewing of this site, please ensure that Javascript is enabled for your browser.
Newsroom_detail  

SEARCH BY KEYWORD

FILTER

QU graduate Omama Abdulrahman studies the use of neural network models | Qatar University

QU graduate Omama Abdulrahman studies the use of neural network models

2022-06-20
inforgraphic

to understand emotions in user statements and generate empathetic text

Omama Abdul Rahman Hamad, a Master of Computer Science graduate from Qatar University (QU) Department of Computer Science and Engineering, recently spoke about her graduation research thesis, which was titled: Effective attention-based models for empathetic text generation. Dr. Khaled Shaaban, Professor of Computer Science, oversaw the thesis.

The study that explores the usage of neural network models that replicate the human mind in the way of learning to understand and interpret human languages is the most important aspect of this research thesis. Preprocessing the text data and translating it to a numerical form so that these neural networks can understand and deal with it is required for these neural networks. As a result, understanding this numerical representation is the foundation for many implicit natural language processing activities that people perform, such as responding to human enquiries.

In her statement, Omama Abdul Rahman, a graduate of MSc Computer Science, said: "One of the tasks of this study is to find the best ways to create a human-like Chabot that engages in dialogue purposefully by interpreting feelings and responding to user questions empathetically. Emotions are divided into more than thirty categories, and each emotion has a different intensity, which makes some emotions close to each other. There are several challenges that this study addressed, first: There are few studies and methods concerned with understanding and responding to feelings correctly in the Arabic language is compared to the English language, so this study suggested a model that solves this, there is a fine line between the text belonging to different feelings due to the vernacular dialects and the different ways of expressing them.”

Omama added: "There are texts whose way of expressing them may be positive, for example, but they may carry different emotions, such as: happiness, surprise or fun. In addition, some feelings can introduce noise to the model, which can be expressed positively or negatively. Therefore, this study proposes a model called Sentiment and Emotion Experts (SEE), which is a comprehensive model for Chabot by understanding the feelings in general if the text belongs to one of the following categories: positive, negative or neutral and then analyzing deep feelings, such as: happiness, fun, surprise, sadness or resentment. Extensive experiments and various evaluations of the model's performance were conducted which demonstrated the effectiveness of the proposed approach by improving the accuracy of emotion detection and the relevance of the texts generated from this model."

Omama confirmed that this study is beneficial in several areas, such as: The presence of a mental health Chabot can play the role of responding with appropriate responses to the questions that the user may ask on the Chabot, as the resulting response has a fundamental effect on the interlocutor, such as making the interlocutor he feels better and thus improves his mental health.

The graduate also pointed out the most important ways to achieve ambitions, saying: "I first seek to increase my knowledge in my field of specialization, which is natural language processing and machine learning, by reading the latest research in this field, and then working on developing the model proposed in my master's thesis to address all Arabic dialects and publishing research in international journals and conferences to highlight its importance.”