Techniques are disclosed herein for transliterating contact names and/or other data from a source language to one or more target languages. The contact names and/or other data are received from a client device (e.g., a feature phone) having an automated assistant. The data in the target language(s) can then be utilized in determining that further input, received from the client device in one of the target language(s), corresponds to certain data in the source language. In one aspect, a method includes receiving, over a network and at one or more remote servers, contact names that are in a source language and that are transmitted from a client device. The method can further include transliterating each of the contact names from the source language to one or more corresponding transliterated contact names in one or more target languages, and mapping each transliterated contact name in the target language(s) to a corresponding contact name in the source language and/or to a corresponding contact entry. The method can further include storing the mapping, along with an identifier of the client device, such as a temporary identifier of the client device. Thereafter, in response to further input received from the client device (e.g., spoken input) in one of the target language(s), the spoken input can be matched to a transliterated contact name, and the mapping utilized to identify a corresponding contact entry and/or contact name in the source language. Thus, techniques disclosed herein can enable a user to create, at a client device, a contact name in a first alphabet (e.g., Latin). The contact name can then be provided to remote server(s), which transliterates the contact name to one or more additional alphabets (e.g., Devanāgarī). Subsequently, the user can provide, at the client device, spoken input that indicates a desire to contact (e.g., call, message, etc.) the contact name, but that references the contact name in the Hindi language (which corresponds to the Devanāgarī alphabet). The spoken input can be provided to one or more of the servers, and the server(s) can match the contact name of the spoken input to a transliteration, of the contact name, that is in the Devanāgarī alphabet. Thus, the contact name can be contacted responsive to the spoken input.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Accame, Diego; Lee, Abraham; Ji, Feng; and Cheung, Ka Hei, "Transliteration of Contact Names and/or Other Data Using an Automated Assistant", Technical Disclosure Commons, (April 03, 2019)