mr-desilva/Multilingual-Dialogue-Summarization-for-Low-Resource-Enviornments

Final Year Research Project - This project focuses on the customer service problem domain where it engage with multinational participants. This proposed application is capable of summarizing dialogues between customer and the support agent in multiple languages where it dosn't require training data # MultiDialSum - Multilingual Dialogue Summary Generation System A written or spoken conversation between two or more people is known as a dialogue. In most of modern-day applications where the conversation happens, it generates lots of data either in textual format or audio format. Summarizing these data or dialogues is known as the dialogue summarization where it only focusses on the relevant information. Dialogue summarization aims summarize only the necessary information where a reader can quickly capture the highlights of the dialogue without reviewing the whole. Customer service is a domain where it generates lot of data daily. Conversation between a support agent and the customer can take up to a lengthy dialogue. At the end of the dialogue conversation, support agents should write a short description (a summary) of the dialogue which will help for later reference. This is a time consuming and requires human resources. MultiDialogSum is a dialogue summary generation system for customer services where it supports generating summaries for multiple languages. The use of recent transformers modals and machine translations techniques are utilized to achieve these capabilities. This research project aims to contribute following factors. - Use of existing text-summarization model for dialogue summarization task. - Use of machine translation techniques to address the data scarcity in dialogue summarization domain. - Releasing a format of dataset which can be used to fine-tuning the text-summarization modals for dialogue summarization task. ---------- ### Proposed System Design ### Model Architecture ### Sample of the web application