As expected, the Google translate app, like the other machine translation services, is still far from the quality of human translation. However, its current state can be improved so it can be effectively used for general purposes. Increasing the translation quality by using more effective deep learning methods can thereby increase the degree of confidence in the translation. I became annoyed while trying out the app when it kept shifting from one output to another as my phone was moved a bit. That is because the app was generating outputs with different degrees of success each time it recognizes a text in the scene. Therefore, a higher degree of success in translation would also give it more stability.
In addition, the app translates almost every word to its corresponding meaning in the preferred language whereas in the real world not everything can be translated word for word. For example, company names, trademarks, names of historical places, etc. mostly have the same pronunciation across many languages. As a regular user I don’t expect to see the name of Apple Company at the back of my phone to be shown and read as (Tofaha) which means (apple, the fruit) in Arabic. Although the Google translate app can handle that for many cases, it is necessary to incorporate the named-entity recognition task so as to effectively classify named entities according to their categories such as organizations, locations, names of people, etc. The app can also add the capability of providing the user with optional additional information about locations or organizations. What if I was browsing a manuscript in a foreign language at the library? Do I need the app to translate pages of texts or to give a summary of what I’m looking at? It’s still something that I can think of while using such an application. Why not as all the technology is here.
The service, however, in its current state is useful especially for tourists who like to travel to different places. It would help to some degree to figure out instructions, restaurant menus, etc. in an easy and fast way. This is especially because it can translate between large numbers of languages. In this way, users do not have to resort to web-based translators or ask the native speakers about things that can be successfully translated, such as ingredients or menu contents.
Having the same technology integrated into augmented reality glasses or lenses would also provide easy, fast and real-time translation service. To be more effective, the above-mentioned issues need to be addressed first. The user needs to feel comfortable while using such glasses. The technology has to have a high degree of translation success, stability, and accurate mapping of the synthetic. There are some pros and cons of walking around wearing such a technology in its current state:
Pros
Cons
It would be beneficial for the user to have some control over enhancing their experience. It would be better if users were allowed to set a threshold degree of confidence to their required results, so that only results with a degree of success above a predefined threshold value are shown. In less critical cases, users may specify lowed threshold values. Since the app tries to translate everything it sees, users might use hand gestures or voice commands to control synthetics in the scene, such as omitting unrequired ones. Lastly, users might contribute to improve the service quality as in the web based Google translate. They can, for example, provide suggestions by editing the results using gestures and voice commands in this case.
In addition, the app translates almost every word to its corresponding meaning in the preferred language whereas in the real world not everything can be translated word for word. For example, company names, trademarks, names of historical places, etc. mostly have the same pronunciation across many languages. As a regular user I don’t expect to see the name of Apple Company at the back of my phone to be shown and read as (Tofaha) which means (apple, the fruit) in Arabic. Although the Google translate app can handle that for many cases, it is necessary to incorporate the named-entity recognition task so as to effectively classify named entities according to their categories such as organizations, locations, names of people, etc. The app can also add the capability of providing the user with optional additional information about locations or organizations. What if I was browsing a manuscript in a foreign language at the library? Do I need the app to translate pages of texts or to give a summary of what I’m looking at? It’s still something that I can think of while using such an application. Why not as all the technology is here.
The service, however, in its current state is useful especially for tourists who like to travel to different places. It would help to some degree to figure out instructions, restaurant menus, etc. in an easy and fast way. This is especially because it can translate between large numbers of languages. In this way, users do not have to resort to web-based translators or ask the native speakers about things that can be successfully translated, such as ingredients or menu contents.
Having the same technology integrated into augmented reality glasses or lenses would also provide easy, fast and real-time translation service. To be more effective, the above-mentioned issues need to be addressed first. The user needs to feel comfortable while using such glasses. The technology has to have a high degree of translation success, stability, and accurate mapping of the synthetic. There are some pros and cons of walking around wearing such a technology in its current state:
Pros
- Using the technology running in augmented reality glasses is seamless. Users can put their glasses on and take them off easily.
- It provides the user with fast, free and real-time translation.
- Google translate is a good choice among other machine translation services in terms of accuracy.
- It can translate between large numbers of languages.
- Although the quality of the service cannot be used commercially, it’s sufficient for casual and general use.
Cons
- Its current state suffers from instability of synthetic mapping.
- It outputs results in different degrees of success and some have failed.
- The accuracy varies from language to language. Translation from Japanese to English, for example, is more accurate than translation from English to Arabic.
- No feedback about the degree of confidence. If you translate from a language that you’re not familiar with like Japanese, there is no way to tell which result is successful and which is a failure. That is, which result that best matches or represents the original text.
- The longer the contents, the lower the accuracy.
- Not everything needs to be translated word for word.
- The user has no control over the contents to be translated in real time, as it almost translate everything it sees.
- Using the app for a long time is annoying and can cause headaches.
It would be beneficial for the user to have some control over enhancing their experience. It would be better if users were allowed to set a threshold degree of confidence to their required results, so that only results with a degree of success above a predefined threshold value are shown. In less critical cases, users may specify lowed threshold values. Since the app tries to translate everything it sees, users might use hand gestures or voice commands to control synthetics in the scene, such as omitting unrequired ones. Lastly, users might contribute to improve the service quality as in the web based Google translate. They can, for example, provide suggestions by editing the results using gestures and voice commands in this case.