US 11,727,302 B2
Method and apparatus for building a conversation understanding system based on artificial intelligence, device and computer-readable storage medium
Ke Sun, Beijing (CN); Shiqi Zhao, Beijing (CN); Dianhai Yu, Beijing (CN); and Haifeng Wang, Beijing (CN)
Assigned to BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed by BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD., Beijing (CN)
Filed on Jun. 12, 2018, as Appl. No. 16/6,208.
Claims priority of application No. 201710443686.5 (CN), filed on Jun. 13, 2017.
Prior Publication US 2018/0357570 A1, Dec. 13, 2018
Int. Cl. G06N 20/00 (2019.01); G06F 7/14 (2006.01); G06F 40/30 (2020.01); G06N 5/02 (2023.01)
CPC G06N 20/00 (2019.01) [G06F 7/14 (2013.01); G06N 5/027 (2013.01); G06F 40/30 (2020.01)] 18 Claims
OG exemplary drawing
 
1. A method for building a conversation understanding system based on artificial intelligence, wherein the conversation understanding system comprises a basic conversation understanding system and an adjustment system, wherein the basic conversation understanding system is built by:
obtaining application scenario information of a specific conversation service scenario provided by a developer through a visualized customization page, the application scenario information including intent information, parameter information and corresponding execution actions; and
according to the application scenario information, building the basic conversation understanding system having basic service logic,
and the method further comprises:
training the conversation understanding system during real-time conversation between a user and the conversation understanding system, both of which having a common task and target, in both a training stage and a using stage, with the following processing:
obtaining, by the basic conversation understanding system, a request from the user;
outputting to the user, by the basic conversation understanding system, a conversation response for the request indicating explicitly the semantic understanding of the request including an identified intent and an identified slot recognized from the request by the basic conversation understanding system;
obtaining, from the user, training feedback information for the conversation response, the training feedback information including feedback for changing the identified intent to a replacement intent;
according to the training feedback information, performing adjustment processing for a service state of the basic conversation understanding system by replacing the identified intent with the replacement intent, and re-executing, by the adjustment system, to obtain an execution result by the basic conversation understanding system, to obtain an adjustment state of the basic conversation understanding system corresponding to the adjusted understanding of the request; and
performing, by the adjustment system, data merging processing according to the training feedback information and the adjustment state of the basic conversation understanding system, to obtain correct annotation data as model training data which is stored in a feedback annotation area to be used to re-optimize the basic conversation understanding system by the developer as required.