FIELD: calculating; counting.
SUBSTANCE: invention relates to computer engineering for user speech recognition. Technical result is achieved by receiving usage history information containing one or more sequences of user interaction events; determining probable intent of user, corresponding to event, for each event in one or more sequences; determining set of probabilities of intent transitions based on probable intentions of user, determined for each event; and using a set of intent transitions to form a customized language model for the user.
EFFECT: technical result consists in improvement of generation accuracy of language personalized for user.
5 cl, 10 dwg
| Title | Year | Author | Number |
|---|---|---|---|
| SYSTEM AND METHOD FOR DETECTING ORPHAN UTTERANCES | 2015 |
|
RU2699399C2 |
| CONTEXTUAL QUERIES | 2011 |
|
RU2573764C2 |
| USING CONTEXT INFORMATION TO FACILITATE COMMAND PROCESSING IN VIRTUAL ASSISTANT | 2012 |
|
RU2542937C2 |
| MACHINE TRAINING | 2005 |
|
RU2391791C2 |
| MAINTENANCE OF CONTEXTUAL INFORMATION BETWEEN USER INTERACTIONS WITH VOICE ASSISTANT | 2018 |
|
RU2785950C2 |
| QUERY COMPOSITION SYSTEM | 2016 |
|
RU2691851C1 |
| USER INTENTION OUTPUT BASED ON PREVIOUS INTERACTIONS WITH VOICE ASSISTANT | 2011 |
|
RU2544787C2 |
| SERVICE ORCHESTRATION FOR INTELLIGENT AUTOMATED ASSISTANT | 2011 |
|
RU2556416C2 |
| PERSONALISED DICTIONARY FOR DIGITAL ASSISTANT | 2011 |
|
RU2541219C2 |
| DETERMINING USER INTENT BASED ON ONTOLOGIES OF DOMAINS | 2011 |
|
RU2541221C2 |
