This article will teach you on how to use and configure our AI integration, along with useful information about this topic.
Here, you will find a video showcasing the capabilities of these features.
1. Introduction
1.1. Where does its information come from?
1.2. Data privacy
2. Activating or deactivating the AI integration
2.1. Account level
2.2. User/editor level
3. How to use the AI in the BWX editor
3.1. Ask AI to translate
3.2. AI Review
3.3. Translation Smells
1. Introduction
The BWX AI can also be called BWX Generative Language Engine and its stands as the cornerstone of our suite, providing unrivaled power and efficiency. By seamlessly combining Translation Memories, Machine Translations, and Glossaries into a unified, smart, and context-aware feed, it transforms the way you work, learn, and interact.

1.1. Where does its information come from?
Our team of machine learning engineers has successfully harnessed the power of large language models and advanced natural language processing techniques to create a unified feed that intelligently adapts to your preferences. This dynamic feed seamlessly incorporates your machine translation, translation memory, and glossary preferences in real-time during runtime operations. The result is an efficient and responsive system that caters precisely to your language processing needs.
Of course, apart from your TMs, glossaries and MT output, our platform AI is powered by Chat GPT (Open-AI).
1.2. Data privacy
BWX has SOC2 Type 2 certification, and we follow well-documented best practices to ensure data privacy, safety, and access controls. Client data may be stored in the system, like billing contacts or POs for projects, but it is up to you how you want to configure these features.
All data is encrypted at rest. All-access to BWX is logged, and we enforce multi-factor authentication. Finally, we provide a fine-grained permission module that allows you to assign responsibilities in the form of user roles (RBAC) and even groups of individual permissions.
To know more, visit https://www.bureauworks.com/security.
Specifically about the AI integration, according to Open AI and Microsoft Azure your data will not be used for training purposes. It will be stored for 30 days with encryption at rest and then permanently deleted.
Their full privacy policies are here for your reference:
More information can be found here.
2. Activating or deactivating the AI integration
2.1. Account level
In the lateral menu, you may click on Settings and My account. From the Account Settings tab, you will see the Enable AI Translation/Review option.
2.2. User/editor level
In the editor, you have the option to enable or disable AI-powered features according to your preferences. This functionality is particularly helpful when you want to use certain AI features while opting out of others. You can tailor the AI usage to suit your specific needs and make the most out of the tools available to you.
To manage our AI features, navigate to the editor menu and click on the Document tab. From there, you can easily enable or disable the AI functionalities according to your requirements.
- Show Translation Smells
- Enable AI Review
- Enable AI Translation
Each of these items will be elaborated upon in the upcoming section of this article.
3. How to use the AI in the BWX editor
3.1. Ask AI to translate
When we ask the AI to translate, it will take into account the glossaries, TMs and MT's outputs. So it would be an translation based on your choices and preferences.
The level of personalization in AI outputs is directly correlated with the currency of your glossaries and translation memories (TMs). Thus, it is imperative to diligently maintain and update these resources as they serve as the nourishment for the underlying mechanism.
With this AI option enabled, you may simply click on an empty segment and the translation will be begin.
If the segment is already fullfilled, you may click on it and then on Ask AI to translate:
The platform will show the result right under the AI section. In order to apply the suggested translation, you simply need to click on Apply this translation.
To regenerate the translation, the 🔄️ icon can be used:
If there are any changes to your input, such as glossaries, translation memories (TMs), or machine translations (MT), after the initial generation, it's recommended to regenerate your translation. By doing so, you ensure that the AI is working with the most up-to-date data, allowing for more accurate and reliable results.
3.2. AI Review
When initiating the AI review process, the system will solely consider the content within the source segment. Your glossaries, TMs and MTs outputs will not be taken into account. This approach leads to a more direct review, presenting a raw assessment with less contextual information.
Subsequently, the platform will present a window displaying the AI review result, as illustrated below:
3.3 Translation Smells
Our Translation Smells refer to any symptom within a translation that could potentially suggest the presence of an underlying problem. They encompass a range of problems or recurring patterns that indicate a translation might not be suitable for its intended purpose or may be inaccurate or unclear.
Examples of translation smells may include: inconsistent terminology, awkward phrasing, or inaccurate translations of idiomatic expressions.
It is important to note that this is a tool that is still in beta version and is intended to support linguists. This warnings can indicate discrepancies or improvements, taking into account commonly used terms and expressions by the target language. Since it doesn't have access to glossaries and TMs, the final translation must always be done by the linguist. The linguists may even disregard those warnings.
Note: Translation Smells can show the presence of potential erros, but not ensure their absence.
⚠️ This is a BETA feature. AI Translation is a machine learning solution only intended to assist and provide additional information, and the results should always be verified by human experts before making any decision or taking any action. It's the user's responsibility to be aware of the potential biases and limitations of the model and take that into consideration when interpreting its output.
If any questions were not answered in this article, please don't hesitate to contact our support team at help@bwx.io.
Comments
0 comments
Please sign in to leave a comment.