CN114817499A - Method and system for flexibly constructing service scene multi-turn conversations - Google Patents
Method and system for flexibly constructing service scene multi-turn conversations Download PDFInfo
- Publication number
- CN114817499A CN114817499A CN202210404999.0A CN202210404999A CN114817499A CN 114817499 A CN114817499 A CN 114817499A CN 202210404999 A CN202210404999 A CN 202210404999A CN 114817499 A CN114817499 A CN 114817499A
- Authority
- CN
- China
- Prior art keywords
- node
- user
- scene
- complex
- conversation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3347—Query execution using vector based model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
- G06F40/35—Discourse or dialogue representation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a method and a system for flexibly constructing a multi-turn conversation of a service scene, which comprises a data flow for constructing a starting node, a logic judgment node, an ending node and a connecting line; the starting node is used as the start of the whole process and supports the configuration of the common mandatory ending corpora of the whole conversation; a connecting line between the starting node and the logic judgment node constructs the data flow graph into a complex conversation scene; the logic judgment node provides an external service to provide a function externally and supports a plurality of complex conversation scenes in series; the end node is used for summarizing information, processing the information desired by the user and returning the processed information to the user. The invention extracts information through the nodes, supports the user-defined coding operation to complete the scene design, and solves the problems that the application scene is relatively fixed and the flexible switching of the scene can not be realized to meet the complex scene link in the prior art; the judgment of subsequent logic is carried out by adopting the connecting line, so that the method is clearer and clearer, and the conversation flow can be quickly and intuitively established aiming at various complex scenes.
Description
Technical Field
The invention relates to the technical field of intelligent conversation, in particular to a method and a system for flexibly constructing a service scene and multi-turn conversation.
Background
The intelligent dialogue system is widely applied to the fields of intelligent customer service, robots, automobiles, navigation and the like, and under the large background that the automobile field is wholly transformed into intelligent modeling, the automobile industry is undergoing a great change, and various standardized processes are gradually replaced by machines. Likewise, a revolution is also experienced for car sales or after-sales service personnel. The appearance of intelligent customer service establishes a better communication bridge between enterprises and users. Long-term data accumulation can give intelligent customer service unlimited intelligence. For example, CN201911166714.9 discloses a method and a system for constructing a multi-round conversation system based on service scenarios, which provides a method for designing a multi-round conversation process according to different service scenarios, and configures corpora and keywords for each user node in the multi-round conversation process, and configures titles and dialogs in each machine node; and respectively matching the linguistic data and the keywords of each user node in the multi-turn conversation process according to a text classification technology and a rule matching training intention judgment model and a rule matching model. According to the method, the node linguistic data are configured, and then the intention is judged through the nodes for distribution. However, the node in the method only supports intention judgment, and cannot support part of complex services or construct more complex business logic. The method is essentially a template configuration method, and is limited to partial simple scenes and does not support expansion by configuring a conversation template in advance and returning conversation template information by matching input corpus information. For another example, CN 202110609449.8 discloses a human-machine multi-turn dialogue method and device, where the human-machine multi-turn dialogue method includes: acquiring a process node; acquiring intention libraries, wherein each intention library comprises intention information and basic linguistic data; obtaining a statement to be analyzed; respectively carrying out similarity comparison on the sentence to be analyzed and each basic corpus so as to obtain the similarity value of the sentence to be analyzed and each basic corpus; and acquiring a corresponding response action according to the identification intention information. Because the intention library can be associated with a plurality of process nodes, a plurality of basic corpora do not need to be configured for each process node. However, this method disposes an error situation in which a plurality of intentions conflict easily occurs when similarity calculations are calculated separately from a sentence to be analyzed.
Therefore, how to flexibly configure and construct a complex conversation scene of a complex flow is a problem to be solved by the intelligent customer service.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a service scene-based multi-turn conversation flexible configuration construction method and system, and solves the problems that the application scene is relatively fixed/single and the flexible switching of the scene cannot be realized to meet the complex scene link in the prior art.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method for flexibly constructing a service scene multi-turn dialogue is characterized in that a data flow is constructed by constructing a starting node, a logic node, an ending node and a connecting line, so that the flexible construction of the service scene multi-turn dialogue is realized;
the starting node is used as the start of the whole process and supports the configuration of the common mandatory ending corpora of the whole conversation; a connecting line between the starting node and the logic judgment node constructs the data flow graph into a complex conversation scene; the logic judgment node provides an external service to provide a function externally and supports a plurality of complex conversation scenes in series; the end node is used for summarizing information, processing the information desired by the user and returning the processed information to the user.
Further, the method for flexibly constructing the multi-turn dialog of the service scene comprises the following steps:
step 1: stripping a service data flow diagram according to a complex service scene, and creating an intention module in advance;
step 2: constructing a data flow graph into a complex conversation scene through a starting node, a logic judgment node, an ending node and a connecting line;
and step 3: the external service is provided to provide external functions, and a plurality of complex conversation scenes are supported to be connected in series.
Further, a rule in a connecting line between the start node and other logical nodes is used as a logical judgment; in various conversation scenes, related intents are required to be established in advance according to business scenes, the whole conversation process can be triggered by selecting the set intents in advance in a connecting line, and meanwhile, corpus triggering is supported to be set in the connecting line.
The related intention is that the user input is directly output through related linguistic data by means of an NGT intention searching method; the method can efficiently and quickly match the most relevant intentions input by the user through approximate nearest neighbor vector search in a high-dimensional space.
Triggering a subsequent sub-graph when the similarity between the user input corpus and the corpus configured in the scene reaches a set threshold value by configuring a trigger corpus; in the logic judgment node, node recurrence is carried out on the complex flow, and the linguistic data of multiple rounds of question return is edited in the logic node.
Furthermore, the logic judgment node is at least one, and the information provided by the user in the complex question-answering scene is acquired according to the service logic editing logic judgment node. After the user triggers the intention, the logic node can acquire some information which is possibly input by the user in the scene; at the logic judgment node, a question-answering operation returned to the user can be directly configured, and the setting can be carried out by compiling a custom script; or inquiring for many times, recording effective information input by the user in a contact manner, and then performing node circulation according to the obtained information.
Furthermore, semantic analysis is carried out according to the content returned by the user in the logic judgment node, and the analysis work is supported by constructing semantic rules, requesting external resources and self-development scripts. The semantic rule is constructed by analyzing a large number of dialogue logs, extracting information with obvious distinguishing characteristics in the logs through various data analysis means and constructing the semantic rule through the information. The external resource request is to directly analyze the content input by the user by calling an external service to acquire necessary information. The self-development script is a self-defined editing script in a logic judgment node under the condition of supporting some complex scenes, and various complex functions are realized through the script, including but not limited to the construction of complex timing tasks.
In various conversation scenes, related intents are required to be established in advance according to business scenes, the intents set in the step 1 can be selected in advance in a connection line in a manner of selecting intention triggering to trigger the whole conversation process, and simultaneously, corpus triggering can be supported to be set in the connection line. And triggering the linguistic data through configuration, and entering a subsequent flow of the conversation flow when the similarity between the user input linguistic data and the linguistic data configured in the scene reaches a set threshold value. In the logic judgment node, by reproducing nodes in a complex flow and editing the linguistic data of multiple rounds of back questions in the logic judgment node, a certain scene can be obtained and completed from a user side, but the content returned by the user side can not directly meet the requirement, so that semantic analysis needs to be carried out on the content returned by the user in the logic judgment node, and the analysis work is supported by constructing a semantic rule, requesting external resources and self-developing scripts. Establishing semantic rules, analyzing a large number of dialogue logs, extracting information with obvious distinguishing characteristics in the logs by various data analysis means, and establishing the semantic rules through the information; meanwhile, the content input by the user can be directly analyzed by directly requesting external resources and calling external services to acquire necessary information; under the condition of supporting some complex scenes, scripts can be edited in a self-defining mode in the logic nodes, and various complex functions can be realized through the scripts, including but not limited to constructing complex timing tasks and the like.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention is based on a business scene multi-turn dialogue flexible configuration construction method, information is extracted through nodes, custom coding operation is supported to complete scene design, and all scenes are theoretically supported; the problem of prior art application scene relatively fixed, can not realize the nimble switching of scene in order to satisfy complicated scene link is solved.
2. The invention adopts the connecting line to judge the subsequent logic, is clearer and clearer, and can quickly and intuitively establish conversation flow aiming at various complex scenes.
3. The intention matching method provided by the invention supports the method, and also supports the user-defined addition of other judgment methods for verification while the intention is judged, for example, a regular expression, and the intention judgment logic can be judged to be triggered only when a plurality of conditions are met simultaneously.
Drawings
FIG. 1 is a data flow diagram of a method for constructing a flexible configuration of multiple sessions based on a service scenario according to an embodiment of the present invention;
FIG. 2 is a simplified logic diagram of a business scenario-based construction method and system for flexible multi-turn dialog configuration according to the present invention;
FIG. 3 is a logical flow to node slot extraction diagram according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood by those skilled in the art, the present invention is further described below with reference to the accompanying drawings and specific examples, but the embodiments of the present invention are not limited thereto.
A method for flexibly constructing service scene multi-turn conversations comprises the following steps: and constructing a data flow by constructing a starting node, a logic node, an ending node and a connecting line, thereby realizing flexible construction of a service scene multi-turn conversation.
Referring to fig. 1, a data flow chart abstracted from a concrete example is shown, and the flow chart includes a start node, an end node and two logical judgment nodes, and performs information transmission and condition judgment between the nodes through a connection line.
The starting node is used as the start of the whole process and supports the configuration of common forced ending corpora of the whole conversation; a connecting line between the starting node and the logic judgment node constructs the data flow graph into a complex conversation scene; the logic judgment node provides an external service to provide a function externally and supports a plurality of complex conversation scenes in series; the end node is used for summarizing information, processing the information desired by the user and returning the processed information to the user.
Referring to FIG. 2, a simplified logic diagram of the present invention is shown, abstractly representing the steps necessary to construct a dialog scenario. In order to realize flexible switching of scenes, meet complex scene links and solve the problem of single scene in the prior art, the invention adopts a method of nodes and connecting lines to construct various complex conversation processes, and the method specifically comprises the following steps:
step 1: firstly, stripping a service data flow diagram according to a complex service scene, and creating an intention module in advance;
step 2: constructing a data flow graph into a complex conversation scene through nodes and connecting lines;
and step 3: the external service is provided to provide external functions, and a plurality of complex conversation scenes are supported to be connected in series.
The intention module created in advance in the step 1 is a method for providing NGT intention search, and supports direct output of the related intention input by the user through related linguistic data. The method can efficiently and quickly match the most relevant intentions input by the user through approximate nearest neighbor vector search in a high-dimensional space.
The specific implementation method of the step 2 is as follows: and constructing a data flow chart by constructing a starting node, a logic judgment node, an ending node and a connecting line. Firstly, a starting node is created, and the node serves as the start of the whole process and supports the configuration of the common forced ending corpus of the whole dialog. And the rule in the connecting line between the starting node and other logic nodes is used as logic judgment. The connection line connecting the start nodes will typically set a trigger condition for the dialog flow.
In various conversation scenes, related intents need to be established in advance according to business scenes, the intents set in the step 1 can be selected in advance in a connection line in a mode of selecting intention triggering to trigger the whole conversation process, and meanwhile, corpus triggering can be set in the connection line. And triggering the linguistic data through configuration, and entering a subsequent flow chart when the similarity between the user input linguistic data and the linguistic data configured in the scene reaches a set threshold value. In the logic judgment node, by reproducing nodes in a complex flow and editing the linguistic data of multiple rounds of back questions in the logic judgment node, a certain scene can be obtained and completed from a user side, but the content returned by the user side can not directly meet the requirement, so that semantic analysis needs to be carried out on the content returned by the user in the logic judgment node, and the analysis work is supported by constructing a semantic rule, requesting external resources and self-developing scripts. Establishing semantic rules, analyzing a large number of dialogue logs, extracting information with obvious distinguishing characteristics in the logs by various data analysis means, and establishing the semantic rules through the information; meanwhile, the content input by the user can be directly analyzed by directly requesting external resources and calling external services to acquire necessary information; under the condition of supporting some complex scenes, scripts can be edited in a self-defining mode in the logic nodes, and various complex functions can be realized through the scripts, including but not limited to constructing complex timing tasks and the like.
The following further describes embodiments of the present invention with respect to a complete dialog scenario.
1. According to the triggering language material of the business scene preparation response, for example, under the car buying scene, the language material of the concrete direct associated business scene such as 'I want to buy cars', 'how xx car price' and the like is prepared, the car buying intention can be newly built according to the language material, and the language material input by the user can be fully understood to belong to the car buying scene through training an intention classification model or simply building a text semantic matching model.
2. And constructing a starting node which can be used as a brief description of the whole conversation scene, or setting a forced interruption corpus at the starting node, and forcibly ending the whole complex conversation process through the corpus. The concrete implementation mode is to judge whether the query is a forced ending corpus or not, directly reset the conversation state, change the type of the conversation behavior into an ending state, and return to a preset scene in a starting node to end the conversation.
3. Creating a scenario trigger condition (edit stream diversion), and the scenario trigger has various implementation modes including but not limited to intention identification, corpus trigger, keyword verification and the like. The intention identification is to call a pre-trained intention classification model, judge which intention is the most matched with the problem input by the user, and further transfer the intention to a subsequent node; the corpus triggering refers to editing relevant triggering corpora in a flow line, and then judging whether user input is similar to a triggering text or not through a semantic model trained in advance; the keyword verification mainly provides a verification method through a regular and equal rule, and whether the user input meets the condition can be quickly judged. The flow line is mainly used for judging flow conditions, subsequent flow conditions and answer operation can be set in the flow line in advance, and then flexible response is carried out according to user input.
4. And editing the logic judgment node according to the service logic. The logic judgment node is mainly used for acquiring information provided by users in a complex question and answer scene. For example, in a car buying scenario, after the user triggers a car buying intention, the logical node may obtain some information that the user may input in the scenario, including but not limited to an intended vehicle, a budget, and the like. At the logic judgment node, a question-back operation returned to the user can be directly configured, and a user-defined script can be programmed for setting, for example, price calculation and screening can be carried out through information input by the user, inquiry can also be carried out for multiple times, effective information input by the user is recorded in a context mode, and node circulation is carried out according to the obtained information.
5. And finishing the node. The end node is used for summarizing information and processing information desired by the user and returning the processed information to the user. For example, in a car buying scene of a user, the comparison of various pre-selected car models can be returned to the user, but almost all platforms do not directly support the comparison at present, but by the method, the data of various car models are acquired in the end node and the customized arrangement of pages is carried out according to a desired format. The end node supports a plurality of coding modes, and the main implementation mode is coding embedding.
Referring to fig. 3, a partial enlarged view of a logical node in the conversation process flow diagram is shown. Fig. 3 shows key information extracted from dialogue data. When the user carries out dialogue interaction with the system, the system analyzes the dialogue information, and extracts the coordinate or longitude and latitude information in the dialogue information in a self-defined coding mode for finishing the execution of the subsequent dialogue flow.
Therefore, the method can solve the problem of difficult construction of multiple rounds of questions and answers in a complex scene.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and not for limiting the technical solutions, and those skilled in the art should understand that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all that should be covered by the claims of the present invention.
Claims (11)
1. A method for flexibly constructing a multi-turn dialogue of a service scene is characterized by comprising the steps of constructing a starting node, a logic judgment node, an ending node and a connecting line to construct a data flow;
the starting node is used as the start of the whole process and supports the configuration of the common mandatory ending corpora of the whole conversation; a connecting line between the starting node and the logic judgment node constructs the data flow graph into a complex conversation scene; the logic judgment node provides an external service to provide a function externally and supports a plurality of complex conversation scenes in series; the end node is used for summarizing information, processing the information desired by the user and returning the processed information to the user.
2. The method for flexibly building multiple turns of dialogue for business scenario according to claim 1, comprising the steps of:
step 1: stripping a service data flow diagram according to a complex service scene, and creating an intention module in advance;
step 2: constructing a data flow graph into a complex conversation scene through a starting node, a logic judgment node, an ending node and a connecting line;
and step 3: the external service is provided to provide external functions, and a plurality of complex conversation scenes are supported to be connected in series.
3. The method for flexibly constructing multi-turn conversations of business scenarios according to claim 1 or 2, characterized in that rules in the connection lines between the start node and other logical nodes are used as logical decisions; in various conversation scenes, related intents are required to be established in advance according to business scenes, the whole conversation process can be triggered by selecting the set intents in advance in a connecting line, and simultaneously corpus triggering is supported to be set in the connecting line.
4. The method for flexibly constructing multi-turn conversations in a business scenario according to claim 1 or 2, wherein at least one logical decision node is used, and the logical decision node is edited according to the business logic to obtain the information provided by the user in the complex question-answer scenario.
5. The method for flexibly building multi-turn conversations of business scenes according to claim 2, wherein the related intention module supports direct output of the related intention inputted by the user by inputting related corpora through a high-dimensional vector retrieval algorithm, NGT intention search; the method can efficiently and quickly match the most relevant intentions input by the user through approximate nearest neighbor vector search in a high-dimensional space.
6. The method for flexibly building multiple rounds of conversations in a business scene according to claim 3, wherein after a user triggers an intention, the logic judgment node obtains some information that the user may input in the scene; at the logic judgment node, a question-answering operation returned to the user can be directly configured, and the setting can be carried out by compiling a custom script; or inquiring for many times, recording effective information input by the user in a contact manner, and then performing node circulation according to the obtained information.
7. The method according to claim 3, wherein the triggering corpus is configured, and when the similarity between the corpus input by the user and the corpus configured in the scene reaches a preset threshold, the subsequent sub-graph is triggered; in the logic judgment node, by performing node recurrence on the complex flow, and by editing the linguistic data of multiple rounds of question return in the logic node.
8. The method for flexibly building multi-turn conversations of a business scenario according to claim 3, wherein semantic parsing is performed according to the content returned by the user in the logical decision node, and the parsing work is supported by building semantic rules, requesting external resources and self-developing scripts.
9. The method for flexibly constructing multi-turn conversations of a business scenario according to claim 8, wherein the semantic rule is constructed by analyzing a large number of conversation logs, extracting information with clearly distinguished features in the logs by various data analysis means, and constructing the semantic rule by the information; the external resource request is to directly analyze the content input by the user by calling an external service to acquire necessary information.
10. The method for flexibly building multi-turn conversations of business scenes according to claim 8, wherein the self-developed script is a custom-edited script in a logic decision node under the condition of supporting some complex scenes, and various complex functions are realized through the script, including but not limited to building complex timing tasks.
11. A system for flexibly building multiple rounds of conversations of a business scenario, characterized in that the method of any of claims 1-10 is used.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210404999.0A CN114817499A (en) | 2022-04-18 | 2022-04-18 | Method and system for flexibly constructing service scene multi-turn conversations |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210404999.0A CN114817499A (en) | 2022-04-18 | 2022-04-18 | Method and system for flexibly constructing service scene multi-turn conversations |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114817499A true CN114817499A (en) | 2022-07-29 |
Family
ID=82536249
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210404999.0A Pending CN114817499A (en) | 2022-04-18 | 2022-04-18 | Method and system for flexibly constructing service scene multi-turn conversations |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114817499A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117151449A (en) * | 2023-10-30 | 2023-12-01 | 国网浙江省电力有限公司 | Data platform chain type information interaction method based on full scene linkage |
| CN117611095A (en) * | 2023-12-06 | 2024-02-27 | 阿帕数字科技有限公司 | Design method of multifunctional combination collocation system applied to supply chain |
-
2022
- 2022-04-18 CN CN202210404999.0A patent/CN114817499A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117151449A (en) * | 2023-10-30 | 2023-12-01 | 国网浙江省电力有限公司 | Data platform chain type information interaction method based on full scene linkage |
| CN117151449B (en) * | 2023-10-30 | 2024-02-06 | 国网浙江省电力有限公司 | Data platform chain type information interaction method based on full scene linkage |
| CN117611095A (en) * | 2023-12-06 | 2024-02-27 | 阿帕数字科技有限公司 | Design method of multifunctional combination collocation system applied to supply chain |
| CN117611095B (en) * | 2023-12-06 | 2024-04-26 | 阿帕数字科技有限公司 | Design method of multifunctional combination collocation system applied to supply chain |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108090177B (en) | Multi-round question-answering system generation method, equipment, medium and multi-round question-answering system | |
| CN110309283B (en) | Answer determination method and device for intelligent question answering | |
| KR102119468B1 (en) | System and method for learning counseling chatter robot based on counceling contents of counselor | |
| CN114817499A (en) | Method and system for flexibly constructing service scene multi-turn conversations | |
| CN106649825B (en) | Voice interaction system and creation method and device thereof | |
| US8903711B2 (en) | System and methods for semiautomatic generation and tuning of natural language interaction applications | |
| CN105159977A (en) | Information interaction processing method and apparatus | |
| CN111177338B (en) | Context-based multi-round dialogue method | |
| CN110008308B (en) | Method and device for supplementing information for user question | |
| CN111611368A (en) | Method and device for backtracking common scene conversation in multi-turn conversation | |
| CN111930912A (en) | Dialogue management method, system, device and storage medium | |
| CN110399473B (en) | Method and device for determining answers to user questions | |
| CN115148212A (en) | Voice interaction method, intelligent device and system | |
| CN111368029A (en) | Interaction method, device and equipment based on intention triples and storage medium | |
| CN117573834B (en) | Multi-robot dialogue method and system for software-oriented instant service platform | |
| CN115658875B (en) | Data processing method based on chat service and related products | |
| CN116974554A (en) | Code data processing method, apparatus, computer device and storage medium | |
| CN115391503A (en) | Application method and device of session flow | |
| CN114626388A (en) | Intention recognition method and device, electronic equipment and storage medium | |
| CN114970559A (en) | Intelligent response method and device | |
| CN114187388A (en) | Animation production method, device, equipment and storage medium | |
| CN113515274A (en) | Visual modeling design method about human-computer interaction logic | |
| CN112925889A (en) | Natural language processing method, device, electronic equipment and storage medium | |
| CN115577092B (en) | User speech processing method and device, electronic equipment and computer storage medium | |
| CN112395402A (en) | Depth model-based recommended word generation method and device and computer equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |