Automatic speech recognition incorporating word usage information

A computing system, comprising:

• at least one processor;
• at least one memory including instructions that, when executed by the at least one processor, cause the system to:
• receive audio data;
• determine first text data corresponding to-the audio data, the first text data including a first word sequence and a next word following the first word sequence;
• determine, using a language model having data for a plurality of states, where each state of the plurality of states corresponds to a respective word sequence, a first state of the plurality of states corresponding to the first word sequence;
• determine, using the language model, that no second state of the language model following the first state corresponds to the next word;
• determining, using the language model, a first score corresponding to a similarity between use of the first word sequence in a text corpus and use of a second word sequence in the text corpus;
• and
• determine, based on determining that no second state following the first state corresponds to the next word and based on the first score, second text data by replacing the first word sequence in the first text data with the second word sequence;
• and
• determine output text data corresponding to the audio data, the output text data including the second text data.