The Main Interface

This is the main interface.

There are 3 main working tabs you can switch between. These are:

• The Translation Center: this is where you will do translation, proofreading and LQA.
• The Image Analyzer: this is where you can perform image-to-text OCR.
• The Utilities: this is where you can access the WYSISYG Editor, the Translation Memory Converter and the Chinese Converter.

Translation Center

In the Translation Center, you can either perform Translation, Proofreading or LQA.

IF YOU ARE NOT FAMILIAR WITH CAT TOOLS: When you use the Translation Center, you are expected to “import content”. The whole point of Raiverb’ Translation Center is to enable you to import whatever content you wish to translate – what we call “source“, but also any contextual element that will help in translating your content more accurately (third language translation, extra comment, expected character limit, etc). If you wish to do proofreading or LQA, you will also need to provide your translation – what we call “target“.
THEREFORE, the easiest and most common type of source document you can provide is an XLSX spreadsheet, where each column is set as a one of the items listed above. Other special formats are specifically dedicated to this type of content, for example PO/POT. Json and XML custom formats are also an option (although they come with their own caveats, detailed below in this manual). The point being, you need to prepare adequately your source content before importing it into Raiverb to get the best results out of the AI components of the program. It is also strongly advised to comment adequately any string that may be subject to confusion or interpretation.

Modes

Translation Mode

With the Translation Mode, Raiverb will translate your content into the target language you chose. Raiverb will use all the contextual information available to improve the accuracy of its output.

Along with translating, it will also try to extract any potential glossary candidate and add it automatically to its dynamic glossary.

Raiverb will also remember each string it translates and turn it into its own internal Translation Memory. This will ensure consistency across the whole job and project.

Proofreading Mode

In Proofreading Mode, Raiverb evaluate the quality of a translation on a scale of 1 to 10, and also propose an improvement if applicable. You will have the choice of the items you wish to review.

Proofreading mode requires you to have imported a target language. If the target is empty, it will be ignored.

LQA Mode

In this mode, Raiverb will automatically analyze your source and target, and will produce a report on the issues it finds with your text.

Proofreading mode requires you to have imported a target language. If the target is empty, it will be ignored.

Importing Content

To get content into the software, you can either use New Entry, Import File or Paste Clipboard. Or just drag and drop any supported file.

＋Custom entry will directly add an editable Translation Unit. This is not the preferred method of importing content, but this is an option you have nonetheless.
If you chose Paste Clipboard, you’ll access the text content you have copied in your clipboard, either from a spreadsheet or from pure text.
Import File allows you to import a supported file into Raiverb.

Raiverb supports the following formats for import: .xlsx .csv .docx .pdf* .xml* .pot .po

Here is the full List of supported formats

Once you have picked your file or used the clipboard paste function, you will see the Import Options interface. Regardless of the file format you have imported, the objective of the Import Interface is for you to tell Raiverb what part of your file is what type of content: either the source text, the translated text (target), or an alternative language or any of the supported contextual content types as described below.
Depending on the format, the content is organized in columns. You must pick the relevant use for each column.

You can chose between:
– Source, the text to be translated or proofread. This is the only non-optional element you need to form a valid Translation Unit.
– Target, if you are intending to proofread your content.
– Alternate translation: if there’s already a translation in another language available, you can chose to feed in this translation to help the engine. You can add as many translations as you want. You will be asked to input the “locale” for each column you add as Alternate translations. In most cases, Raiverb will automatically detect the language and locale, but if not, remember to correct it manually.
– Character limit, if any.
– String ID: String IDs may help in certain cases where they contain some context for the potential use case.
– Extra comment: Extra comment will be fed to the engine in the exact same manner as as the Extra comment in the main interface of the program. The extra content will be included in addition to any comment you input in the main interface. you can, however, change this behavior afterward.

Once imported, you will see Translation Units being deployed. Some pre-processing is necessary so you’ll need to wait for a little while.

Importing a Translation Memory

A Translation Memory is a special type of file, holding translation records, widespread in CAT tools to provide context for human professionals to review or translate.

The main two formats widely used for Translation Memories are: TMX, and XLIFF. They have their differences but are mostly the same thing for the same purpose as far as Raiverb is concerned. You can use either. Raiverb also provides a simple converter allowing you to turn spreadsheets into Translation Memories.

If import was successful, you will see the number of entries on display.

Importing Glossary

A Glossary, sometimes referred to as Term Base, is a list of terms to be used within a given project. The difference with Translation Memories is that Glossaries are composed of a couple of words only. In Raiverb, you can import a Glossary simply through a dual-column spreadsheet file.

Importing a Glossary will open the same window as the file importer, allowing you to customize which column will serve as the source, and which column will serve as the target.

Extra Comments

Here, you are free to provide any extra set of instructions you want so the AI can know more about your project.

This is a very important step that should not be overlooked. You may specify here the industry you’re working with so that the translation may adapt accordingly.

Depending on your task, you may want to direct the AI to use a specific style or advise it to add punctuation if your source file misses it. For example, you may want to strictly capitalize all glossary terms, or to imitate a writing pattern. It is up to you to decide what is relevant to your project, or not.

Here is a few advice that tend to yield good results :

– For stylized translations: try to find a known/classic author with the style you’re trying to aim for, or any other influence in the target language of your task and ask the AI to mimic this style. Remember though, AI has its limits regarding creative tasks.
– Precise your industry and the type of project you’re working on.
– Request the AI to comment on the translation. For example, if you are translating highly specific legal text that’s not guaranteed to have a perfectly translatable equivalent, you may ask for extra explanations as comments.
– Advise the AI of the specific job you are running. If this is a new game project, you could use:
“This is a casual party game, please use a casual, relaxed tone when applicable”.
– After you started running a job, pay attention to the first set of translations returned. Observe what recurring mistakes are made, then if needed, stop or pause the job to update your prompt.

Remember that what you write here will be propagated to all Translation Units. When importing documents, you always have the possibility to add extra comments.

Getting Ready

If you are running any sort of important or large job, preparing the task is an important step, and you should make sure every bit of data you want to feed in has been imported and set properly.

Translation Units

Once a document is imported, you will see Translation Units appearing in the interface. Translation Units are each segments of text you have imported previously.

Translation Units can be paragraphs or pages, or cells. What defines or separates a Translation Unit is up to you to decide and to set up on import. This step is called segmentation. It depends mostly on what type of data you want to use along with your source text.

Each Translation Unit has a control box with several buttons:

• ❌Delete

• 🖼See with/without color tags

• 💬Translation Unit Info

• 🔇Ignore/Restore

Translation Unit Info Interface

This is the most important interface as you can here check all information relevant to the unit and check/rectify eventual errors before starting to process the text.

All the information you see will be submitted to the final prompt.

You can also see the matches to the Translation Memory and the Glossary. It is advised to check the relevance of matched content for errors.

Note: Raiverb is designed to create its own Translation Memory and Glossary entries when running the job. Self-generated matches will be marked accordingly.
Note2: External TM and glossary matches will always have the priority over self-generated content.

Other options

Translation options

In Translation mode, you can choose to show the past and previous Translation Units to the AI so it can know more about the context of what it is translating. This is useful for split content, such as dialogues or subtitles. This tends to increase the relevance of translations, but also increases tokens consumption. Make a reasonable evaluation of the impact of this option on your project.

Proofreading options

Proofreading has a lot of different options.

– Skip TM Matches: Allows you to not review 100% matches. Use this option if you trust your Translation Memory to save time and words.
– Check glossary: Will check if entries from the glossary are consistent with the translation. Ineffective without a Glossary loaded.
– Check TM consistency: Will check if the translation is coherent with the TM entry or entries. Ineffective without a TM loaded.
– Check typos & grammar.
– Code, number, tags: Will check if tags, variables, and other strips of code are correctly assigned in the translation.
– Style & quality: Will check for overall style and quality. It is strongly advised to use the Extra Comment dialog to direct the AI about what you expect.
– Character limit respect: Will check if the limit set was respected and reformulate if not. Ineffective without a limit set.

Processing your Task

Log-in & Server choice

Don’t forget to log-in. Without an account, you won’t be able to run a translation or proofreading job.

Once you are logged-in, you will be able to pick a server. Raiverb proposes Open servers, marked as such, available freely but subject to speed and character limitations. Pro servers will require a paid subscription, with increased speed and no limitations.

If you want to try what your job will run like before, you can choose “Sandbox”. The Sandbox mode will output the data that would have been sent to the server without actually processing it or using Tokens.

Start

Once you’re good with your setup, hit the Start button.

⚠ It is strongly advised to monitor how the software is doing on the first few entries, to determine early on if there are any issues with the current setup and how the AI is doing. You can pause or stop processing at any time, but it won’t be possible to course-correct or modify any setup once the job is started. You will have to start over. So it’s crucial to make sure everything is set and working as intended early-on.

Raiverb will process your document one Translation Unit at a time. All repeated content will be auto-filled.

Managing Jobs

Glossary

Once the translation job is started, the AI will start to pick up glossary entries as it goes.
You can view these new entries via the button “Manage” in the Glossary panel.

It will be possible to export these glossary entries to your final file. While not all entries might be relevant to the final user, it is important for the software to collect them to be able to provide potentially accurate results.

Archives

Once your job is running, each Translation Unit will be recorded and saved automatically, so you don’t lose data, and are able to do editing once the translation is complete.

Toolbox

• 🖼 Visual toggle will attempt to interpret color tags and end of lines correctly, for ease of display.
• 💯 Score filter is only used in Proofreading mode, and will discard any result scored above the threshold you will have set in Saving options. This will allow you to ignore entries with no issue.
• 📈 Guidance: The Guidance is a special feature which evaluates and report on the part of content generated by the machine VS the content fed to the machine via the Translation and Glossary. It helps the user evaluate the part of autonomy of the machine in the translation, thus an indication of the reliability of the result.
• Rec: Enables/disables automatic recording of translations into the Archiver.

Read more on Guidance

Saving documents

List of supported formats for export

The simple way

If you don’t feel like going into in-depth setup, hitting “Paste” will make a quick copy of all your content.

The more involved way

However, if you need more options, you can use the Saving Interface.

The Saving Interface proposes you more export options.
– Raiverb Spreadsheet is the main way to export content. It is technically an XLSX spreadsheet that is entirely customizable using “Customize Output”. You’ll have the choice to get your source, target, glossary, comments, score and much more in one place.
– Same as Source will attempt to export your file in the same format it was imported. PDF export isn’t supported yet. If you have imported images, they will be exported along with a text file with the source and translation.
– Translation Memory will export your job into TMX or XLIFF formats. Or both.
– Glossary will simply export the extracted glossary from a job, in a spreadsheet format.
– More features are being in development, including the future option to export your job in JSON formats with online callbacks. Stay tuned! And please support Raiverb!

Customize Output is an interface for more in-depth options regarding file exports for each format.

Archives Manager

The Archives Manager is the place where you can review and edit every translation passing through the Translation Center.

You’ll need to have the REC button enabled so the recording can happen. REC is enabled by default.

The Archive Manager contains all Translation Units recorded. These units are recorded on your local machine and available at any time. The main purpose is to enable you to post-edit a translation even after a job is finishes, and to allow you to either:
– Export your edited content as a separate file/Translation Memory.
– Reuse it as a Translation Memory for new jobs without having to juggle through external files.

The Archive Manager can organize and filter your content by Source Language, Target Language, Batch ID and Projects.

To import content from the Archive Manager into the Translation Center, simply press the “Archive” button in the Translation Memory section. You will then see how many Archives have been imported.
Archives behave effectively the same as Translation Memories you import.

Image Analyzer

In Raiverb, you can import image files, turn them into text and send them to the Translation Center. You can also input a URL pointing to an image or to a webpage. If you open a webpage, Raiverb will extract all the images it can find and grab the text from them.

List of supported formats for import

To import pictures, you can use Import Local Files.
You can also import a whole folder with Import Local Folder.
You can also download your pictures from the Internet. If you URL points directly to an image, a single image will be imported. If your URL points to a webpage, Raiverb will attempt to extract all the image content it can find.

From Clipboard is still under development and will be coming at a later date. Optimization options are also under development.

Once images are imported, if the text recognition was successful, you can export your content to the Translation Center.

Sensitivity

If you see your text has a lot of undesired content, unwanted letters or symbols, you can try the Sensitivity slider. The Sensitivity will decrease the tolerance of the exporter and may remove false positives.

Efforts are being made to improve the output of the Image Analyzer. Please note that handwritten text or rotated text is very unlikely to be extracted properly in the current version.

Utilities

TM Converter

Raiverb only accepts .tmx or .xliff files for Translation Memories. But sometimes, if the only source you have at hand is multilingual spreadsheet, turning these into proper Translation Memories can be challenging if you don’t have a dedicated tool.

Raiverb integrates a simple converter that can turn your multilingual spreadsheets into a .tmx or a .xliff easily. It also can append your content into existing Translation Memories, or even merge Translation Memories. You can then use these Translation Memories, either within Raiverb, or within any CAT tool or software supporting these formats.

Chinese Converter

This is a simple tool to convert Traditional Chinese to Simplified Chinese and vice versa.

WYSIWYG Editor

The WYSIWYG Editor allows you to get some text A dedicated page is available here.

https://www.bouncerock.com/hub/raiverb/wysiwyg-editor