Regarding the application of LINE's digital assistant service Clova, LINE's Chief Technology Officer Park Yi-bin revealed in an interview at the developer conference held in Tokyo that there are indeed plans to expand Clova's support for more languages in the future, but the most important consideration is the degree of integration of local data. He also emphasized that Clova will play a key role in the future of LINE services.
According to Park Yi-bin, Clova currently only supports interactions in three major languages: Japanese, Korean, and English. The main reason is that there is still a shortage of data, including voice-related data. However, they are indeed preparing to develop projects to support other languages.
The main reason why Clova still cannot support more languages is the lack of sufficient data
Microsoft has previously explained similarly that its localized Cortana digital assistant service, XiaoIce, which supports simplified Chinese and is now available in China, was primarily due to a lack of data on localized terms. Therefore, manufacturers like Apple and Google, which have deep integration and analysis of user data, can generally launch digital assistant services that interact through natural voice in a very short period of time.
Taking the service model of Apple or Google as an example, from the very beginning, Siri on iOS devices or Google Now service on Android devices actually converts voice into text first, then analyzes and compares the text, obtains the corresponding answer content, and then reads the text out through synthesized speech. In fact, it is still more like the original text search process, but presented in a more realistic interactive way.
Therefore, Siri, which was later used in HomePod, had to rebuild its supported languages, and Google Assistant also had to be retrained. It was only recently that it began to support natural interaction with Chinese voice. The reason behind this is that the amount of data, including sound content, available for reference and learning is still insufficient, making it more difficult for the device to understand the user's natural spoken content.
In fact, many services have already used Clova technology resources
Park Yi-bin said that LINE's services, including news and tourist attraction recommendation mechanisms, have long supported text analysis and comparison operations, and are compatible with language content such as Chinese and Thai. However, in order to enable natural interaction with users through spoken language in smart speakers, the current amount of data in languages other than Japanese, Korean, and English is clearly insufficient for the system to fully learn.
However, in areas with high LINE usage, such as Taiwan and Thailand, it is expected to be open to users in the future.More Clova integration servicesIn fact, many of the technical applications behind LINE services have been imported into the Clova artificial intelligence platform resources, including content analysis and recommendation, image recognition, and computer vision applications. Even the chatbots currently common in LINE are mostly related to Clova technology. In the future, they will continue to use the AutoML learning framework,Clova ExtensionsAnd other resources allow Clova to learn more skills and applications.
As for Clova's current development position within the LINE service, Park Yi-bin believes that it is primarily aimed at making LINE services more convenient while also making them smarter. For example, a significant proportion of the spam message blocking behind the system is currently done through artificial intelligence analysis and judgment. Although some messages still slip through the net during the analysis and filtering, it is actually able to block the vast majority of spam or phishing messages. This is why LINE hopes to continue to improve its artificial intelligence technology in the future.
Judging from the current development of the Clova artificial intelligence platform, in addition to providing front-end interaction in the form of a smart speaker and serving as a medium for future human-computer interaction, it is more important to connect and vertically integrate LINE's services and enable cross-platform and cross-device connections, thereby achieving LINE's goal of connecting with more people and things.
However, as far as LINE's current position is concerned, it will still comply with the privacy regulations of various regions and will not arbitrarily access user data for analysis and learning. Therefore, compared with other technology companies that also have a lot of data, the development speed of artificial intelligence technology applications will be relatively slower. However, it is expected that as the service gradually becomes more popular, LINE will be able to further develop its artificial intelligence technology with more data.
