Robot Waiter - Speech Interpereter
At The University of Lincoln my group was taked with a hard task of creating three working parts to a robot waiter of which my part was allocated within our group to work on the Speech Interpereter section. This involved planning out how we would tackle this project and starting early, to test the viability of various ideas.
To begin with, the plan was to utilize the recently released local version of Deepseek AI within a web GUI for ease of acess, however after a painstaking process of trying to get the smaller, easier to run models working properly (without halucinating/trailing off) we realized we would have to cut our losses as the AI just wasn't ready given the combination of models available, hardware available for us to use, and time left. We also didn't want to utilize non-local models as we didn't want to spend any money. This early version of the project is still on my github.
After the setbacks involved with trying to get the AI version of the speech interpereter working we realised that we needed to take a step back and re-evaluate how to tackle the problem. We started back from scratch with a command line version of the interpereter that simply understood when specific items were mentioned. From here we could work slowly and begin adding other features we needed, such as the ability to save orders, a way for the waiter to give the robot other menu items, and understanding of multiples of items on the menu (e.g. two pizzas, five chicken). This allowed us time to add further quality of life features such as a GUI using Tkinter, a way for the waiter to set the number of tables and a page for the customer to select their table, a visual way for the user to edit their order, and a printed reciept for the waiter to see the orders as well as the table number and the time of order.
Allough we wanted to, we didn't have enought time left, nor time with the robot itself to fully implement the seperate modules together (allthough this wasn't an issue for the project as integration was never required within the project), however we did theorise some ways, which made its way into the report.
Speech Interpereter Repo Full Robot Waiter RepoPHP Chat Application
This is a PHP based chat application using SQLite that runs using a backend service, created as part of a work project as a proof of concept.
This application allows the user to create and join groups, and send messages within the groups.
PHP Chat App Github