Automatic Learning of Language Models

A method for speech recognition comprising:

• determining a context vector for speech audio spoken by a user, the context vector including one or more elements of context;
• determining, based on the context vector, a context-dependent library associated with the context vector;
• determining, based in part on the context-dependent library, that the speech audio includes a phrase in the context-dependent library, the phrase in the context-dependent library previously identified by analyzing a text-based communication;
• and
• outputting text for the included phrase.

A system for altering a context-dependent library of phrases, the system comprising:

• at least one processor;
• at least one computer-readable storage medium storing instructions that are executable by the at least one processor to perform operations including:
• analyzing a text-based communication of a user within an organization to:
• determine a phrase within the text-based communication, the phrase being one or more words in length;
• and
• determine a context vector for the text-based communication;
• determining, based on the context vector for the text-based communication, a context-dependent library associated with the context vector, the context-dependent library shared with one or more users within the organization;
• and
• altering the context-dependent library based the determined phrase, the context-dependent library usable by the one or more users within the organization to improve speech recognition of speech audio having another context vector associated with the context-dependent library.