Introduction
This project showcases a broad range of our services but also highlights our ability to adapt to the client’s growing needs. When Lexorama asked us for a cross-platform desktop app, we didn't hesitate to broaden the project's scope. Throughout the development, we wrote the Neural Desktop mostly in C++ and Qt, implemented multiple features like language model customization, and conducted a robust series of testing.
Interested in details? Here is how this project went.
In this article:
- Introduction & Client's Background Go to text
- Research & Collaboration Go to text
- Design & Development Go to text
- Conclusions Go to text
Introduction & Client's Background
Lexorama is a Cyprus-based company located in Larnaka.
Established in 2017 by Panagiotis Kanavos, Laratech Ltd., a subsidiary, focuses mainly on professional translation and automation.
When Lexorama contacted us to build their neural machine translation software, they already had their project’s scope predefined at the outset, which included core features like multi-language support, integration with CAT tools, and advanced AI-driven translation.
We worked collaboratively to detail each aspect of the project, ensuring that all technical and functional requirements were documented. This was also the starting point for our extensive research phase.
Research & Collaboration
During our research, we established that Machine Translation (MT) systems tend to translate sentence by sentence, largely skipping the wider context.
In addition, many professional translators perceive their job as challenging and time-consuming – especially in technology-driven industries such as CAD, AEC, BIM, etc.
That is why our software had to be not only precise and user-friendly but also cross-platform ready.
After an additional round of consultations and market research we began to work on the Neural Desktop.
Project Requirements:
- A specialized translation tool in the form of a desktop app.
- Cross-platform compatibility.
- Multi-language support.
- CAT tools integrity.
- Advanced AI-driven translation capabilities.
Design & Development
Our team developed NeuralDesktop using mostly the C++ and the Qt frameworks, covering both FrontEnd and BackEnd processes.
The initial scope of the project was clearly defined; however, as the work progressed, the client expanded the requirements.
Handling all tasks internally often allows us to establish long-lasting business partnerships. Through continuous collaboration and open communication, we adapted to these changes seamlessly and without outsourcing.
Main project’s key features included:
- context-aware output,
- neural fuzzy engine,
- dictionary integration,
- seamless CAT tool plugins.
From the very beginning, we aimed NeuralDesktop to be user-friendly, highly customizable, and efficient. After all, Lexorama’s software had to meet the varied needs of freelance translators and small to medium-sized translation offices.
One of our goals was to grant Neural Desktop users full control and ultimate privacy. That is why we ensured that it runs locally as a system service without unwanted interaction or data exchange with any third parties.
“Context” was one of the unique features we wrote for this project.
Whether it’s through file translation, Clipboard translation, or even through plugins in CAT tools like Trados and memoQ – the user could always translate his texts thanks to the AI-driven Neural Machine Translation.
We also ensured that both the engine and its language models were able to incorporate all the latest advancements – offering accurate, fluent, and adaptable translations.
Additionally, our team implemented custom dictionaries to help the system adapt to users’ terminology. We even allowed users to test on the fly every setting they changed via “Test Area” in the configuration dialog box.
Services provided:
- Desktop application development,
- Front-end development,
- Back-end development,
- C++ and Qt development,
- Cross-platform development.
NeuralDesktop significantly boosts productivity by generating high-quality, context-aware results and integrating smoothly with popular CAT tools. Our app is compatible with WindowsOS and macOS and supports multiple language pairs.
Thanks to our expertise in C++ and Qt development, Neural Desktop users can also easily attach translation memories and glossaries and fine-tune the engine itself to control the output.
Additionally, our team implemented a real-time fine-tuning feature to allow users to see how every change affects the translation results. Moreover, we ensured that all models were multi-client and multi-domain so that users have absolute control over the terminology and style.
In essence:
We made our client’s product so simple and practical that it can be used by anyone who wants to save time and boost their productivity.
It is worth mentioning that all work was conducted remotely, demonstrating our ability to effectively manage international projects.
At this point it is worth adding that Neural Desktop is continuously updated with new models and features.
Conclusions
This successful delivery exemplifies our capability to meet and exceed client expectations.
The Lexorama project demonstrates our proficiency in handling complex software development tasks remotely as well as robust resource management. Both of those enabled us to deliver a high-quality product that met the client's needs and enhanced their operational efficiency.
Would you like to finally turn your vision into a successful product?
