Country's patent office has granted an Indian patent to Mark Zuckerberg-led Facebook for an invention relating to a system that enables the social media giant provide quality and error-free translation of texts for its members.
Country’s patent office has granted an Indian patent to Mark Zuckerberg-led Facebook for an invention relating to a system that enables the social media giant provide quality and error-free translation of texts for its members. Unlike the conventional process of translation, the present invention, according to the patent documents filed by Facebook, discloses a system for selecting a translated text phrase from a set of translations for a text phrase in a social media network. At first, the system identifies users who could perform the translation for the text phrase based on profile information. The translation of the text phrase is performed by users or members identified in the social network.
The system receives the translated text from the members of the social media network. Once the translations are received, the system employs a process to select an optimised translated text among translations. To perform this, the system uses the votes from the users on the translations and determines quality scores for the translations. The system assigns weights, or credibility coefficients, to the votes for determining quality scores. A weight of a vote may be based on quality scores for translations submitted by the voter. The system finally selects a translation, based on quality scores for available translations and the selected translation is then presented to the member.
The patent, for which an application was filed at the Chennai office in the year 2010, will give a leg-up to Facebook’s effort to provide error-free translations on its platform which has been seeing increased interactions, in various languages. Facebook said that in the existing techniques, manual interventions are necessary and the text is translated manually by a system operator into various desired languages. However, this requires effort and expense by the system operator.
Another option is automated translation of text based on foreign language dictionaries, translation algorithms, or other methods. Automated translation is generally difficult and error- prone, often changing the meaning of the translated text or producing non-idiomatic translations. As a result, desired translations may not be produced, and members who need the translations will not be able to properly use the social network and will not be able to connect well with members who use other languages.
Facebook claimed that in the present invention, members of the social networking system vote on the received translations. The votes indicate the perceived quality of the translations by the voting members. A translation can then be selected from among available translations based on the votes and other factors. The system uses the votes and determines quality scores for the translations. Thereafter, the system assigns weights, or credibility coefficients, to the votes for determining quality scores. A weight of a vote may be based on quality scores for translations submitted by the voter. Votes by voters who are a friend of the particular member may be weighted more heavily than votes by others.
By using the existing resources of the social network, that is, allowing the members of the social media network to translate the text rather than implementing translation algorithms, the present invention is able to reduce the system resource usage, which would lead to economical solution and to provide an optimised translation of text, thereby allowing effective communication between various members in the social media network. Since the system performs validation of the translated text, the system assures that the best translated text is provided to the user and hence it is error free, the company said.