US 11,783,173 B2
Multi-domain joint semantic frame parsing
Dilek Z Hakkani-Tur, Kirkland, WA (US); Asli Celikyilmaz, Kirkland, WA (US); Yun-Nung Chen, New Taipei (TW); Li Deng, Redmond, WA (US); Jianfeng Gao, Woodinville, WA (US); Gokhan Tur, Kirkland, WA (US); and Ye-Yi Wang, Redmond, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Aug. 4, 2016, as Appl. No. 15/228,990.
Claims priority of provisional application 62/354,064, filed on Jun. 23, 2016.
Prior Publication US 2017/0372199 A1, Dec. 28, 2017
Int. Cl. G06N 3/08 (2023.01); G10L 15/16 (2006.01); G10L 15/22 (2006.01); G10L 15/18 (2013.01); G06N 3/044 (2023.01)
CPC G06N 3/08 (2013.01) [G06N 3/044 (2023.01); G10L 15/16 (2013.01); G10L 15/1822 (2013.01); G10L 15/22 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A system comprising:
one or more processing unit(s);
one or more computer-readable media coupled to one or more of the processing unit(s), the one or more computer readable media having thereon one or more modules of computer-executable instructions to configure a computer to perform operations comprising:
building a multi-domain classifier model configured to jointly model domain, intent, and a sequence of one or more slot values for a natural-language sequence of words, the model comprising a recurrent neural network (RNN) that operates on an input sequence comprising (n+1) input elements to generate an output sequence comprising (n+1) output elements each uniquely associated with a corresponding one of the input elements, the input sequence comprising a sequence of n tokens representing the words of the natural-language sequence and the output sequence comprising a sequence of n slot tags associated with respective tokens representing the words of the natural-language sequence, an (n+1)-th element of the input sequence being an end-of-sequence token and an associated (n+1)-th element of the output sequence being a concatenation of domain and intent tags.