Language profiling service

A system comprising:

• a memory comprising instructions;
• and
• one or more processors, wherein the instructions, when executed by the one or more processors, cause the system to perform operations comprising:
• for each language in a plurality of languages, utilizing, by the one or more processors, a plurality of counters to log use of the language by a user of an online service, the plurality of counters for each language including at least a first counter for logging a Top-Ranked-Count, indicating how many times the language was a top prediction among all language predictions, and a second counter for logging a Probability-Sum, indicating the sum of probabilities for the language across all the language predictions, the plurality of counters being updated based on detecting interactions of the user in the online service;
• for each language in the plurality of languages, calculating a weighted sum of the values of the plurality of counters, for the user;
• detecting, by the one or more processors, a text entered by the user in the online service;
• obtaining, by a language classifier using the one or more processors, an initial prediction having probabilities for the plurality of languages that the text is in the language;
• calculating, by the one or more processors, a language distribution prediction based on the initial prediction and the weighted sum of the counters for each language for the user by multiplying the initial prediction by the weighted sum of the counters, the language distribution prediction comprising a probability, for each language, that the text is in the language;
• selecting, by the one or more processors, a language used in the text based on the language distribution prediction;
• and
• causing, by the one or more processors, presentation on a display of a message in the selected language.

A non-transitory machine-readable storage medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:

• for each language in a plurality of languages, utilizing, by one or more processors, a plurality of counters to log use of the language by a user of an online service, the plurality of counters for each language including at least a first counter for logging a Top-Ranked-Count, indicating how many times the language was a top prediction among all language predictions, and a second counter for logging a Probability-Sum, indicating the sum of probabilities for the language across all the language predictions, the plurality of counters being updated based on detecting interactions of the user via the online service using a language of the plurality of languages;
• for each language in the plurality of languages, calculating a weighted sum of the values of the plurality of counters, for the user;
• detecting, by the one or more processors, a text entered by the user in the online service;
• obtaining, by a language classifier using the one or more processors, an initial prediction having probabilities for the plurality of languages that the text is in the language;
• calculating, by the one or more processors, a language distribution prediction based on the initial prediction and the weighted sum of the counters for each language for the user by multiplying the initial prediction by the weighted sum of the counters, the language distribution prediction comprising a probability, for each language, that the text is in the language;
• selecting, by the one or more processors, a language used in the text based on the language distribution prediction;
• and
• causing, by the one or more processors, presentation on a display of a message in the selected language.