Telecom Review Africa conducted an exclusive interview with Salma Jamoussi, a researcher and associate professor at the University of Sfax, Tunisia (with a Ph.D. in Computer Science and Artificial Intelligence from the University of Lorraine, France), and delved into her experience and expertise in the field of AI. During the interview, she shed light on her intention to utilize Natural Language Processing techniques in her projects and also addressed the challenges that women may encounter when specializing in this field.
Can you tell us about your experience and expertise in the exciting field of Artificial Intelligence?
I am currently a member of the research laboratory MIRACL (Multimedia, Information Systems and Advanced Computing Laboratory). My research interests focus on artificial intelligence, data science, machine learning and natural language processing methods.
In fact, since my first research activities, I have been interested in artificial intelligence and machine learning, especially neural networks. I began my research career with robotics, and I did my PhD on natural language processing. I was among the first researchers to use machine learning methods to understand speech. To reach this goal, I employed neural networks and Bayesian networks. Afterwards, I decided to get deeper into the fundamental aspects of this field. Therefore, I was first interested in clustering methods and used meta-heuristics to enhance clustering results in many applications. I continued to work as well on Bayesian networks and especially on structure learning using bio-inspired methods like genetic algorithms and ant colonies. As my favorite application field is always text and social media data, the curse of dimensionality was one of my primary interests. I proposed many methods on this topic (filter, wrapper and embedded methods) for both supervised and unsupervised learning. All these research interests led me to focus more now on deep learning methods and conduct my research activities on recent advances in data mining, which are the Volume and Velocity of data in the era of Big Data.
How have you already utilized or how do you intend to utilize Natural Language Processing techniques in your projects or research?
The main focus of my projects is to use NLP techniques in order to gain valuable insights into users' characteristics, preferences, and behaviors from their user-generated content. These insights can be used for diverse applications ranging from user profiling to personalized recommendations and targeted marketing. In my case, I concentrate more on the early detection and prevention of mental health disorders. I mainly use sentiment analysis and NLP techniques to understand the emotional states of users and detect signs of these disorders, which can lead in some cases to distress, depression and suicidal ideations. This information can be used to provide personalized support and [initiate] early interventions.
What are some ethical considerations and potential biases that should be taken into account when developing AI models?
This is a very important question. Indeed, collecting and using personal data — when profiling users, for instance — is a very sensitive issue and must adhere to privacy laws and regulations. Sensitive information should be handled with care, and data anonymization techniques should be employed. In addition, user consent is needed in many situations.
Moreover, the designed AI models, which are used by experts and stakeholders in general, have to be explainable and interpretable. Indeed, when using AI models, users and stakeholders should have access to clear explanations of the model’s outputs and results to assess its behavior and its potential biases. When models can provide explanations, it becomes easier to hold them accountable for their actions and address any potential issues or concerns.
Can you share an example of a real-world application where Deep Learning has made a significant impact?
One prominent example of a real-world application where deep learning has made a significant impact in the context of NLP is in the field of question-answering systems. Deep learning models, particularly with models like BERT and GPT-3, have significantly improved the accuracy, fluency, and contextual understanding of question-answering systems, opening up new possibilities for intelligent human-computer interaction and knowledge retrieval.
ChatGPT, for instance, has revolutionized the AI field by significantly enhancing the capabilities of natural language understanding and generation. It can understand and respond to complex queries in a manner that closely resembles human-like understanding.
From your experience, what are some real difficulties and challenges that women might face when wanting to specialize in such a field?
According to my experience and my own situation, the main difficulty that could [hinder] a woman when working in the AI research field is related to the work-life balance issue. In fact, balancing work and personal responsibilities can be particularly challenging for women in AI, as the field often demands long work hours, intense workloads, and frequent upskilling. Striking a balance between career aspirations and personal commitments can be a significant hurdle, especially when we are mothers and living in an Arab country.
Moreover, another significant issue that women can face in such fields, is the underrepresentation problem, especially in leadership and responsibility roles. The main matter here is the underestimation of women's abilities and capabilities in research and academia. I think that research institutions and universities have to support gender diversity and give women the opportunity to take on leadership roles and responsibilities, harnessing the full potential of women's talents and contributions.
Can you discuss any ongoing or future projects that you are working on related to the focus areas we mentioned?
Actually, I have been working on many social media mining-related projects. Namely, the user profiling issue has been the focus of my research interests since the Tunisian revolution, where social networks played a prominent role. Currently, I am working on more advanced issues related to this topic, where I focus on the early detection of mental health disorders and suicidal intentions of social network users by analyzing their generated content. Apart from the use of NLP and sentiment analysis methods, I am now interested in online social network profile building by taking into account the multilingual nature of the user-generated content and the fake news and profiles that could distort our models. The consideration of these aspects will allow for a more accurate and more complete user profiling, making it possible to decide what are the right steps to take in order to properly support users and help them overcome their mental health problems. The ultimate objective of this project is to build a chatbot to interact with users in a conversational manner and offer them mental health support. Such a conversational application can supplement existing mental health services and provide accessible and convenient support to a wider population.