Knowledge Base

Getting Started

  • How to create a glossary to customise the translation engine

    A glossary allows you to specify translations for words and short phrases. Similar to a standard dictionary, there is no need to enter plural forms, gender forms, or inflected verbal forms. The glossary supports the following languages listed here.

    At KUDO, we will take care to customise your personal translation engine with your company glossary. Our team will guide you on the best practices to collect the relevant information and will make sure that your AI engine is using your terminology. Please contact your KUDO point of contact to get in touch with our linguistic expert. 

    In the following, you will find information to understand the data we will need to customize your translation experience. 

    1. The original term in the speaker's language must be an exact match 

    • The original term in the speaker's language must match exactly what the speaker says. However, as previously mentioned, the glossary supports different plural, gender, or inflected verb forms. For example, if we add the following term to the glossary:

     

    English Spanish
    smartphone application aplicación para móvil

     

    If the speaker says "smartphone app", it will not be translated correctly, as "app" is a shortened form of "application," which is not an exact match to the glossary entry. However, if the speaker uses the plural form "smartphone applications", it will be correctly translated as "aplicaciones para móviles". Therefore, to ensure accurate translations, the glossary term must be an exact match to the speaker's words.

     

    2. Avoid Punctuation Marks (no commas)

    • In the original term (speaker's language): Refrain from adding commas
    • In the translated terms: Commas should also be avoided. However, other punctuation marks such as "?", "!", ":", and "." are allowed.

    3. Focus on Short and Simple Terms

    • The glossary is designed for individual words, compound words, or very short phrases; not entire sentences.
    • If sentences must be included, follow these guidelines to ensure accuracy and consistency:
      • Split sentences into smaller, meaningful segments, with a maximum of seven words per segment. Here it is an example:
        English Spanish
        So I say Por eso les digo:
        let the Holy Spirit guide your lives dejen que el Espíritu Santo los guíe en la vida
        Then you won’t be doing Entonces no se dejarán llevar
        what your sinful nature craves por los impulsos de la naturaleza pecaminosa
        The sinful nature wants to do evil La naturaleza pecaminosa desea hacer el mal
        which is just the opposite of que es precisamente lo contrario de
        what the Spirit wants. lo que quiere el Espíritu.

    Splitting sentences in the glossary will not have any negative impact on how sentences are actually split, in this case, the previous example will look like this:

    Screenshot 2025-01-07 at 12.56.12.png

    4. Use One Translation per Term

    • Provide only one possible translation per term.

    Incorrect:

    French English
    Zone Area/Zone

    Correct:

    French English
    Zone Area

     

    This ensures the glossary entry has a single, clear translation for each term.

     

    5. Be Careful with the use of Parentheses

    • Instead of adding terms with parentheses, separate the term and abbreviation into distinct entries.

    Incorrect:

    English Spanish
    Environmental Protection Agency (EPA) Agencia de Protección Ambiental (APA)

    Remember that the term in the speaker's language must be an exact match of what the speaker says, so if the speaker says "Environmental Protection Agency" without "EPA", this will not respected the target term included in the glossary. In the same way, if the speaker says only "EPA", this will not be translated as "APA". To approach this issue, it is recommended to separate the term from its abbreviation or any other information within parenthesis, as in the following table:

    Correct:

    English Spanish
    Environmental Protection Agency Agencia de Protección Ambiental
    EPA APA

     

    6. Avoid Symbols or Trademarks

    • Do not include symbols like ®, ™, or ℠ in glossary terms.

    7. Check for Typos and Spelling Errors

    • Ensure terms are free from orthographic errors, such as missing accents or misspelled words. Typos in the glossary will not be corrected by the AI, negatively affecting captions and the AI voice.

    By following these guidelines, you can enhance the quality and consistency of your translations. If you have any further questions, our sales team will connect you with one of our linguistics experts.

     

     

     

  • Customizing Translation: Advanced Linguistic Features

     

     

    Our Speech translation system offers several advanced features tailored specifically for our One-Way Translation type (see this article to understand our translation types; note that these features are not available in the Interactive Translation Type). These advanced functionalities are designed to improve translation quality, ensuring accuracy, and maintaining consistency.

    1. Glossary
    2. Do-not-translate
    3. Sounds like 
    4. AI voice pronunciation control

    At KUDO, we will take care to customise your personal translation engine with your company data. Our team will guide you on the best practices to collect the relevant information and will make sure that your AI engine is using your terminology. Please contact your KUDO point of contact to get in touch with our linguistic expert. Below, you’ll find a detailed overview of how each linguistic feature can benefit your translation process.

     

    1. Glossary

    The Glossary allows to define specific translations for terms in the listeners’ languages, ensuring consistency and precision. This feature supports different variations of the same word. Therefore, there is no need to enter plural forms or inflected verbal forms. Similarly, this feature is not case sensitive, which means that capitalization is not important. See example below:

     

    Original sentence:  “Hoy vamos a hablar sobre la AV que es un tema muy relevante.”

     

                       with glossary                                                        without glossary

               

    In a corporate setting, an acronym like "AV" in Spanish may have a default translation in English as "VA", but by using a glossary, it can be translated more accurately into its complete version "Visual Acuity" to reflect the complete meaning in context.

     

    Max. size of glossary:

    There is no limitation on the number of terms that can be added to the glossary.

    Where to access this feature:

    This is a backend feature; therefore, a sales representative will connect you with our linguist specialist to ensure glossary is customized and optimized for your specific translation needs.

    Supported Languages:

    The Glossary feature is available in the following 16 languages:

    • Chinese
    • Danish
    • Dutch
    • English
    • French
    • German
    • Italian
    • Japanese
    • Korean
    • Norwegian
    • Polish
    • Portuguese
    • Romanian
    • Russian
    • Spanish
    • Swedish

    See article on best practices to create a glossary.

    2. Do-Not-Translate

    The Do-Not-Translate (DNT) feature ensures that specific words or terms are not translated. This is especially helpful for preserving product names, proper nouns, brand names, and acronyms in their original form. Adding the speakers’ names to the DNT list is also recommended, as it prevents the name from being translated and improves its recognition during speech processing. This feature is case sensitive. Refer to the example below:

     

    Original sentence: “Remember to have at your disposal the office package.”

                              ✔ with DNT                                                              without DNT

         

    For example, the product name "Office Package" could be incorrectly translated as “paquete de oficina” into Spanish, which loses the specific reference to Microsoft’s product suite. However, by adding the word "Office" to the DNT list, we ensure that it remains untranslated, resulting in the correct translation of “paquete Office.” This maintains brand consistency and preserves the intended meaning.

     

    Max size of DNT:

    The number of words that can be added to the DNT list is limited to 500 words.

    Where to access this feature:

    The DNT feature can be edited autonomously under the Translation Setting tab. The list of terms added can be saved for upcoming meetings by clicking on the bottom option “Save these words for future meetings”.

     

    Supported Languages:

    This feature is available in all 39 supported speaker languages

    • Arabic
    • Bulgarian
    • Cantonese
    • Catalan
    • Chinese
    • Croatian
    • Czech
    • Danish
    • Dutch
    • English (US, UK)
    • Estonian
    • Filipino
    • Finnish
    • French (France, Canada)
    • German
    • Greek
    • Hebrew
    • Hindi
    • Hungarian
    • Indonesian
    • Italian
    • Japanese
    • Korean
    • Latvian
    • Lithuanian
    • Malay
    • Norwegian
    • Polish
    • Persian (Farsi)
    • Portuguese (Portugal, Brazil)
    • Romanian
    • Russian
    • Slovak
    • Slovenian
    • Spanish (Spain, LatAm)
    • Swedish
    • Thai
    • Turkish
    • Ukrainian
    • Vietnamese

    Recommendations:

    • When adding acronyms, avoid including periods between letters (e.g.,  "ABC" instead of "A.B.C."). 
    • Refrain from using symbols such as ®, ™, or © 
    • Note that this feature does not account for variations or extensions of the word. For example, if you add "car" to the system, related forms such as "cars" will not be automatically recognized. Ensure all necessary variations are explicitly added.

    3. Sounds Like 

    The Sounds-Like feature improves the recognition of complex terms pronounced in unexpected ways. This is essential for words with similar pronunciations, acronyms, abbreviations, and terms where phonetics play a crucial role. See the example provided below:

     

    Original sentence: “SSHRC es uno de los tres principales organismos federales de subvención” 

                                                    Translated sentences

                with sounds like                                            without sounds like

              

    For instance, a term like "SSHRC" may be pronounced differently depending on the speaker's accent or region, but with the Sounds-Like feature, the AI will recognize the intended pronunciation ("ʃɜːk") and avoid mistaking it for other similar-sounding words like "shirk".

    Capacity:

    The Sounds Like feature is limited to 500 words.

    Where to access this feature:

    This is a backend feature; therefore, a sales representative will connect you with our linguist specialist to ensure the sounds like feature is customized and optimized for your specific translation needs.

    Supported Languages:

    This feature is supported in all 39 speaker languages.

     

    • Arabic
    • Bulgarian
    • Cantonese
    • Catalan
    • Chinese
    • Croatian
    • Czech
    • Danish
    • Dutch
    • English (US, UK)
    • Estonian
    • Filipino
    • Finnish
    • French (France, Canada)
    • German
    • Greek
    • Hebrew
    • Hindi
    • Hungarian
    • Indonesian
    • Italian
    • Japanese
    • Korean
    • Latvian
    • Lithuanian
    • Malay
    • Norwegian
    • Polish
    • Persian (Farsi)
    • Portuguese (Portugal, Brazil)
    • Romanian
    • Russian
    • Slovak
    • Slovenian
    • Spanish (Spain, LatAm)
    • Swedish
    • Thai
    • Turkish
    • Ukrainian
    • Vietnamese

     

    4. AI Voice Pronunciation Control

    The AI Voice Pronunciation Control feature ensures the AI voice pronounces complex terms or phrases in a specific, predefined way. This is critical for ensuring clarity and maintaining professionalism in synthesized speech. Refer to the example below:

     

    Original sentence: “SSHRC es uno de los tres principales organismos federales de subvención” 

                                                          Translated sentences

    with AI voice pronunciation control            without AI voice pronunciation control     

           

              SSHRC is pronounced  ("ʃɜːk")                SSHRC is pronounced  ("es-es-eich-ar-si")

    Capacity:

    Unlimited number of terms.

    Where to access this feature:

    This is a backend feature; therefore, a sales representative will connect you with our linguist specialist to ensure the AI voice Pronunciation Control feature is customized and optimized for your specific translation needs.

    Supported Languages:

    21 supported listener’s languages, including:

    • Bulgarian
    • Catalan
    • Czech
    • German
    • Greek
    • English (US, UK)
    • Spanish (Spain, US)
    • Filipino
    • Finnish
    • French (France, Canada)
    • Indonesian
    • Italian
    • Korean
    • Dutch
    • Polish
    • Romanian
    • Russian
    • Slovak
    • Slovenian
    • Thai
    • Turkish

     

    Conclusion

    KUDO AI’s advanced features are integral to enhancing the overall user experience in real-time translation systems, providing a more tailored and accurate translation. These capabilities allow for greater customization, particularly in the recognition, translation and pronunciation of complex terms. With KUDO AI, clients can fine-tune the pronunciation of specific terms, ensuring that even the most intricate technical jargon is recognized and articulated with precision. This level of personalization not only boosts translation accuracy but also significantly improves communication in multilingual settings, leading to higher user satisfaction and a more seamless, professional experience.

     

     

     

     

     

  • Translation Types: One-Way and Interactive Communication

    KUDO AI Speech translation offers two distinct modalities, One-Way and Interactive Communication, each of them designed to cater to different use cases and user needs. You can select the desired translation modality while scheduling a session. You cannot change between modalities once a session has been scheduled.

    One-Way Communication 

    Use Case: Suitable for presentations with one speaker at a time. Ideal for formal presentations, lectures, or scenarios where translation accuracy and clarity are critical. 

    Output: translated audio and/or captions. 

    Quality: Highest quality with very accurate translations. 

    Advanced Functionalities:

    • Glossaries: Customize how specific words are translated. 
    • Do-Not-Translate Words: Specify words that should remain untranslated (e.g. product names). 
    • Sounds_Like Feature: Improve recognition of complex terms pronounced in unexpected ways (e.g., recognise “SSHRC” when the speaker pronounces it as “ʃɜːk”). 
    • AI Voice Pronunciation Control: Force the AI voice to pronounce complex terms in a specific way (e.g., pronounce “SSHRC” as “ʃɜːk”). 
     

    Interactive Communication 

     

    Use Case: back-and-forth conversations. 

    Output: Captions only; no AI voice. 

    Quality: Lower quality but faster translations. 

    Advanced Functionalities: Does not support customization tools like glossaries, do-not-translate words, or pronunciation adjustments. 

     

    Both modalities support the same number of source and target languages. 

  • Testing KUDO AI

    You have got access to KUDO AI and want to run a test to see how it works. There are various ways to evaluate KUDO AI's transcription and translation capabilities. Below, we provide an easy setup for testing. This method allows you to create a session on your computer and speak into the application while a second device, such as a mobile phone, acts as the receiver for the translation. For optimal testing, we recommend involving two people:

    • Speaker: Uses the computer to host the session and speak.
    • Audience: Uses the mobile device to receive and review the translation.

    Setup Instructions:

    1. Create a New Session: Set up a session with two languages: the language you will speak and the target language for translation. (Refer to this guide for detailed instructions on creating a session.)
    2. Prepare the receiver (your mobile): Locate the QR code created for your session in the Session Overview Page. Scan the QR code with your mobile device to join the session as a listener. Your mobile device will receive the translation, both in captions and in audio.
    3. Prepare the sender (your computer):

      • On your computer, in the Session Overview Page locate the Participant Link and click on it to begin the session.
      • A new tab in your browser will open and the KUDO Web Conferencing interface will appear.
      • Enter the required details (typically your email and name) and click Join.
      • Select your speaking language to proceed.
    4. Begin Speaking: Once inside the live session, turn on your microphone and start speaking.

    5. Open the Mobile Device:

      • On your mobile device, select the language you wish to receive as the translation.
      • Click Join to enter the session as a listener.

    Live Testing:

    You are now in a live session. The speaker (on the computer) can talk, and the listener (on the mobile device) can hear and read the real-time translation.

    Testing Tips:

    • Monodirectional Testing: This setup is designed for one-way communication, where one person speaks and the other listens.
    • Avoid Audio Interference: Ensure the speaker and audience are in separate rooms or maintain enough distance to prevent audio feedback or interference between the two devices. You can also use headset to prevent audio repetitions.

     

     

  • Start/End a Session and Session Links

    Once you create a new session, an overview page will be displayed, showing all the session details. This page can be accessed at any time from the list of upcoming and past meetings available in your account.

    Screen Shot 2024-12-10 at 16.59.26.png

    In addition to details about the meeting schedule, languages, and other essential information, you have the following options available:

    • Start, modify, and end the session using the three icons located in the top-right corner of the page.
    • Access and share session links with participants as needed.

    Session Links

    The links generated for each session are categorized as follows:

    1. KUDO Platform Links
      These links provide access to a meeting hosted on the KUDO Web Conferencing platform:

      • Participant: Grants access to users who can both listen and speak during the meeting.
      • Viewer-Only: Grants access to users who can only listen; they cannot activate their microphone or speak.
      • Operator: Grants access to a special interface designed to support the meeting. The user entering this link has some special privileges to support technically the meeting. The privileges are comparable to the role of a host, for example they can restart the translation service in case of issues. 
    2. External Links
      These links are intended for integration with other platforms and access via mobile devices:

      • Integration Link: Embeddable code to integrate KUDO with third-party applications.
      • Mobile Link: Provides a URL and a downloadable QR code for easy access to the meeting via any mobile device. No app download is required, making it ideal for on-site meetings.
        Tip: Print and display the QR code at the entrance of the venue to simplify participant access.

    This structure ensures seamless access for all participants, whether they are joining via desktop, mobile, or third-party platforms, both online and on-site.

     

  • Advanced Settings for KUDO AI

    Note that some of the options refer to the updated version that will be released at the end of December.

    When scheduling a session, you have the option to configure advanced settings. All these settings have been pre-configured with the optimal defaults suitable for the vast majority of meetings. Therefore, in most cases, you do not need to modify them. However, to enhance the experience for specific needs, you may need to customize them. Below is a list of all the settings you can configure when creating a new session:

    Mobile Link and Integrations

    This option enables the automatic creation of a link for accessing KUDO AI on any mobile device through a dedicated URL or a QR code. The QR code will open a page in the browser. Additionally, this option provides an embed code to integrate KUDO AI into third-party platforms. This option is enabled by default.

    This option includes the following two settings:

    • Enable Autofloor: When activated, this setting allows participants to hear the original (speaker's) audio within the translation channel if the speaker is speaking the same language as that particular channel.
    • Enable Horizontal Screen Projection: When activated, this setting adjusts the appearance of the Mobile Link, allowing it to be fully resizable. This feature is especially useful for projecting captions on large screens during in-person events, where space is limited due to other visuals being projected on the screen.

    Your Logo

    This option allows you to upload your company's logo, which will be displayed in place of the KUDO logo on the Mobile Link. This option is useful if you want to customize the experience and display your own brand when using translation services.

    Speaking Mode

    This option applies to sessions held via KUDO Web Conferencing. The default setting is Open Mic, but you can customize this based on your needs. The available modes are:

    • Open Mic: Participants can independently turn their microphones on and off. You can also limit the maximum number of microphones that can be active simultaneously by adjusting the "Number of Active Mics" setting.
    • Request to Speak: In this mode, participants must request permission to speak by clicking a dedicated icon within the KUDO Web Conferencing application. The session host will then enable their microphone.

    Recordings

    This option allows you to record the audio and video of your meeting. Recordings will be available for 30 days. By default, this option is set to OFF, ensuring that no recording of your meeting will occur without your explicit consent.

    Audio Manager

    Allow the Audio Manager role in order to ingest and route audio for hybrid or in-person meeting. (this option cannot be changed after scheduling the meeting). This option is useful when a technician needs to ingest into KUDO AI the audio source from a meeting.

    PIN protection

    Protect your meeting with a PIN code. Participants will be required to enter this code to join the meeting. By default, this option is set to OFF.

    Preparation time

    Allow the meeting to start before the scheduled time. This is useful for technical setup, such as checking if connections are established and ensuring that audio is properly captured by the device.

    Grant consent for the KUDO team to access your meeting

    Allow members of the KUDO team to join your meeting when necessary, for example, for debugging purposes.

See all 9 articles