Sarra is an expert in image processing, machine learning and decision support. On Thursday 28 September, she presented her SIRS-Bot project to some twenty IT, HR and Business experts as part of our Innov'Rencontres. SIRS-Bot is a visual chatbot, designed for the e-commerce market, that processes user data in order to best meet their purchasing needs. Sarra shares her expertise on the development of her project.
Sarra, can you tell us about your SIRS-Bot project and what it consists of ?
SIRS-Bot (Smart Information Retrieval System – Robot) is an intelligent visual chatbot that allows it to analyze complex data. Developed for the e-commerce sector, SIRS-Bot leverages its image processing skills to extract user preferences, and its natural language processing capabilities to understand user needs.
What are the objectives of this project ?
The chatbot is able to provide concise and accurate answers to users' questions. Current chatbots are either textual or visual, but they do not take into account user preferences. Moreover, their operation relies on deep learning*, which makes them very demanding in terms of time and space.
Our objective was to explore other techniques that would reduce the temporal and spatial complexity of existing chatbots while maintaining satisfactory accuracy.
*Deep learning is a type of artificial intelligence derived from machine learning where the machine is able to learn on its own.
What is the technology used ?
On the one hand, we use NLP (Natural Language Processing) techniques, that is to say:
- Semantic and syntactic similarity metrics
- Web scraping: this involves extracting data from websites and saving it for analysis or use in any other way.
- Web mining: it consists of mining technical data on a very large scale to discover models of the Internet and any website.
And on the other hand, the techniques of image processing.
How is SIRS-Bot an innovative project ?
Even with the emergence of the new generation of visual chatbots such as ChatGPT and Google Bard, SIRSBot stands out for its ability to process amateur images. In the field of e-commerce, SIRSBot favors a unique response, that specifically sought by the user.
This project innovates by providing a concise answer to a visual question and reducing temporal and spatial complexity.
What are the next steps for this project ?
In order for this project to become a standalone application capable of searching for answers throughout the web, it is necessary to advance its skills. This involves optimizing the results as well as the models we have developed. Secondly, the objective is to offer it to our customers in the e-commerce and health sector. Indeed, SIRS-Bot is able to analyze any medical document and establish a correspondence between the patient’s state of health and the doctor most qualified to treat this condition.
Thanks to Sarra, from Audensiel’s Lab'Innov, for her contribution to this interview.
Qu'avez-vous pensé de cette interview ?
Intéressante
Constructive
Moyenne
Sans plus