CN112199486A - Task type multi-turn conversation method and system for office scene - Google Patents
Task type multi-turn conversation method and system for office scene Download PDFInfo
- Publication number
- CN112199486A CN112199486A CN202011134560.8A CN202011134560A CN112199486A CN 112199486 A CN112199486 A CN 112199486A CN 202011134560 A CN202011134560 A CN 202011134560A CN 112199486 A CN112199486 A CN 112199486A
- Authority
- CN
- China
- Prior art keywords
- task
- slot
- model
- office
- intention
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000000605 extraction Methods 0.000 claims abstract description 41
- 238000013136 deep learning model Methods 0.000 claims abstract description 20
- 238000003672 processing method Methods 0.000 claims abstract description 9
- 230000014509 gene expression Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 description 10
- 239000013598 vector Substances 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 241000282414 Homo sapiens Species 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
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
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Human Computer Interaction (AREA)
- Evolutionary Computation (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a task-type multi-turn conversation method and system for an office scene. The method comprises the following steps: scene definition is carried out on different office application tasks, and intentions and slot values of the different office application tasks are obtained; acquiring data of the intentions and the slot values of the different office application tasks to obtain acquired data; determining an intention recognition model and a slot extraction model according to the acquired data by using a deep learning model training method and a customized rule processing method; performing dialogue management until necessary information is collected; determining a current user intent and a current slot value according to the intent recognition model, the slot extraction model and the necessary information; and executing the service according to the current user intention and the current slot value. The invention can cover different office task scenes, accurately meet the requirements of users and improve the office efficiency.
Description
Technical Field
The invention relates to the field of task conversation of office scenes, in particular to a task-type multi-turn conversation method and system of the office scenes.
Background
Human-computer interaction is a basic technology for information exchange between human beings and computers in the information age, and is widely concerned by the academic and industrial fields. Man-machine conversation is the core field of man-machine interaction technology, and aims to maximally imitate the way of conversation between people, so that people can communicate with machines in a more natural way. In general, man-machine dialog systems can be roughly divided into two types: task dialog systems and non-task dialog systems, which are also called chatting robots.
The chat robot represented by microsoft ice makes the man-machine conversation technology have more practical value and commercial value, but has certain gap with human beings in the aspects of naturalness, logicality, fluency and the like. The task type conversation system has the characteristics of clear scene, easiness in evaluating conversation quality and the like, and has higher application value compared with a non-task type conversation system. Task-based dialog systems are oriented in the vertical domain, with the aim of helping users to complete predetermined tasks or actions, such as booking airline tickets, hotels and restaurants, etc., using as few dialog turns as possible.
With the development of internet technology, office systems are also becoming popular and developed, and a task-based dialog system is added to an office system to help users complete tasks and better handle office scenes. The current task-type multi-turn dialogue construction process is completely based on a deep learning algorithm and can utilize the advantage of big data. However, the data volume in the office field is small, and certain requirements are placed on the accuracy of service completion, so that a good effect cannot be obtained completely based on a deep learning algorithm. In addition, there are systems using a single round of conversation or a simple multiple round of conversation, and the demands cannot be fully satisfied for office systems.
Disclosure of Invention
The invention aims to provide a task-based multi-turn conversation method and system for office scenes, which can cover different office task scenes, accurately meet the requirements of users and improve the office efficiency.
In order to achieve the purpose, the invention provides the following scheme:
a task-based multi-turn conversation method for an office scenario, comprising:
scene definition is carried out on different office application tasks, and intentions and slot values of the different office application tasks are obtained;
acquiring data of the intentions and the slot values of the different office application tasks to obtain acquired data;
determining an intention recognition model and a slot extraction model according to the acquired data by using a deep learning model training method and a customized rule processing method;
performing dialogue management until necessary information is collected;
determining a current user intent and a current slot value according to the intent recognition model, the slot extraction model and the necessary information;
and executing the service according to the current user intention and the current slot value.
Optionally, the determining an intention recognition model and a slot extraction model according to the acquired data by using a deep learning model training method and a customized rule processing method specifically includes:
training a deep learning model Bert + TextCn according to the acquired data to obtain an intention recognition model;
training a Bert-BilSTM-CRF model according to the acquired data to obtain a non-time slot extraction model;
and processing the collected data by adopting a regular expression rule based on common sense date to obtain a time slot extraction model.
Optionally, the performing session management until collecting necessary information specifically includes:
dialogs and templates for replies are specified for different office application tasks, with multiple rounds of dialog with the user until the necessary information is collected.
Optionally, the executing the service according to the current user intention and the current slot value specifically includes:
judging whether the task meets the task requirement or not according to the current user intention;
if the task requirements are met, executing appointed service calling according to the slot value;
and if the task requirements are not met, executing the service by a question-answer similarity matching method.
A task-based multi-turn dialog system for an office scenario, comprising:
the scene definition module is used for carrying out scene definition on different office application tasks to obtain intents and slot values of the different office application tasks;
the data acquisition module is used for acquiring data of the intentions and the slot values of the different office application tasks to obtain acquired data;
the natural language understanding module is used for determining an intention recognition model and a slot extraction model by using a deep learning model training method and a customized rule processing method according to the collected data;
the dialogue management module is used for carrying out dialogue management until necessary information is collected;
a current user intention and current slot value determining module for determining a current user intention and a current slot value according to the intention recognition model, the slot extraction model and the necessary information;
and the service execution module is used for executing the service according to the current user intention and the current slot value.
Optionally, the natural language understanding module specifically includes:
the intention recognition model determining unit is used for training a deep learning model Bert + TextCn according to the acquired data to obtain an intention recognition model;
the non-time slot extraction model determining unit is used for training a Bert-BilSTM-CRF model according to the acquired data to obtain a non-time slot extraction model;
and the time slot extraction model determining unit is used for processing the acquired data by adopting a regular expression rule based on common knowledge date to obtain a time slot extraction model.
Optionally, the dialog management module specifically includes:
and the conversation management unit is used for specifying the conversation and the template of the reply for different office application tasks and carrying out multiple rounds of conversation with the user until necessary information is collected.
Optionally, the service execution module specifically includes:
the judging unit is used for judging whether the task meets the task requirement or not according to the current user intention;
the first service execution unit is used for executing appointed service calling when meeting the task requirement;
and the second service execution unit is used for executing the service by a question-answer similarity matching method when the task requirements are not met.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a task-based multi-turn conversation method and system for an office scene, which can more accurately finish the requirements of users in a specific business process and facilitate the addition and management of task scenes; the data are generated by using the template to achieve the effect of data enhancement; in the groove extraction process, deep learning and customized rules are combined, so that the extraction result is more definite; for the demands not in the office task scene, question-answer matching based on similarity is used. The whole multi-turn conversation method can cover different office task scenes, accurately meets the requirements of users, and improves office efficiency.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of a task-based multi-turn conversation method of an office scenario of the present invention;
FIG. 2 is a diagram illustrating a scenario definition according to the present invention;
FIG. 3 is a schematic view of session management according to the present invention;
FIG. 4 is a schematic diagram of the service implementation of the present invention;
FIG. 5 is a diagram of a task-based multi-turn dialog system according to an exemplary embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a task-based multi-turn conversation method and system for office scenes, which can cover different office task scenes, accurately meet the requirements of users and improve the office efficiency.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
FIG. 1 is a flow chart of a task-based multi-turn conversation method in an office scenario according to the present invention. As shown in fig. 1, a task-based multi-turn dialog method for an office scenario includes:
step 101: and scene definition is carried out on different office application tasks, and the intentions and the slot values of the different office application tasks are obtained.
In the process of scene definition, according to different office tasks, the natural language and intention of users in different tasks are divided, and a slot value is set according to the task requirement of the scene.
The scene requirement is defined in the following specific mode:
and defining corresponding tasks according to the requirements of users on different functions in the office scene.
Such as leave requests, car dispatches, meeting room reservations, etc., the execution of which requires the user to provide certain parameters (hereinafter referred to as slots) such as time, place, leave type, etc.;
the provision of these slots is not possible with only a single round of dialog, and therefore multiple rounds of dialog need to be designed.
According to different tasks in an office scene, the intention of a user and a corresponding slot are defined, and the reply and the flow of a plurality of rounds of conversations are designed, so that a conversation system can completely and accurately acquire required information. As shown in fig. 2.
Step 102: and carrying out data acquisition on the intentions and the slot values of the different office application tasks to obtain acquired data.
Firstly, through user collection, in the process of using the office system by the user, the user has certain habit on using office business, so that the real conversation corpus of the user can be collected, for example, "I want to please leave things in the tomorrow". In addition to the user-provided real data, the slots in the user-provided real data are removed, such as "i want to please see false", to obtain a template, and a portion of new data is generated using the template.
The template data is mainly used for making up for the deficiency of data, such as real data provided by a user, and all leave types are not included. In the task scene of leave, a plurality of leave types are provided, so that the leave types in the user data can be replaced to perform data enhancement.
The fill template requires different slot values, some of which are proper terms in office scenarios. For example, in the scheduled task of the meeting room, the names of the meeting room are from fixed sources, and the types of vehicles in the dispatching task are also available. These terms are collected for use in generating new data.
Step 103: determining an intention recognition model and a slot extraction model according to the acquired data by using a deep learning model training method and a customized rule processing method, and specifically comprising the following steps of:
training a deep learning model Bert + TextCn according to the acquired data to obtain an intention recognition model;
training a Bert-BilSTM-CRF model according to the acquired data to obtain a non-time slot extraction model;
and processing the collected data by adopting a regular expression rule based on common sense date to obtain a time slot extraction model.
The purpose of natural language understanding is to map user input into semantic slots that are defined in advance from different scenarios, typically including intent recognition and semantic slot extraction.
Natural language understanding should translate user input into a form that can be understood by a computer as completely, clearly and accurately as possible. The method comprises two modes of separate modeling and joint modeling, and the method respectively models the intention recognition and the groove extraction.
Intention recognition:
the intention recognition belongs to a text classification task, and judges the intention of the user and the related field according to the input of the current user, wherein the intention of the user and the related field are from requirement definitions in the step 101, namely vehicle dispatching, leave asking, meeting room presetting and the like.
For text classification tasks, statistical learning models, such as SVM, na iotave bayes, etc., were used early. With the development of artificial intelligence technology, the deep learning model has good effect in the text classification task. User data is collected and data enhancement is performed using template generation. By using a deep learning model in these data, a higher-level feature can be obtained, and the effect of intention recognition is better.
Training a deep learning model Bert + TextCn model by using the data obtained in the step 102, performing intention recognition on the dialogue of the user by using the trained model, and judging whether the dialogue is a task type dialogue, namely car dispatching, leave asking, meeting room presetting and the like; or non-task type such as inquiry reimbursement process.
Groove extraction:
slot extraction is different from domain recognition, intent detection, which essentially belongs to the problem of sequence labeling, aiming at recognizing semantic slots and their corresponding values in sentences. Different requirements in an office scenario have different parameters. For example, the leave request needs to extract parameters such as leave request type and leave request time of the user. Therefore, according to the scenario requirements defined in step 101, parameters (slots) required by different functions are determined, and the collected data is applied to model training to obtain a slot extraction model.
And (3) training the Bert-BilSTM-CRF model by using the data obtained in the step 102, and accurately extracting other slots except the time slot in different scenes by using the trained model.
In different office scenarios, the slot is required to be used for "time". There are many expressions for "time" slots, such as a specific date, or "tomorrow", "next three weeks", etc., and duration "please leave three days", etc., and the regular expression rules based on common sense dates are used for processing to identify the time slots and the times of start and end during the user's dialog.
And combining deep learning and rules, extracting the slots required by the service from the user conversation, and providing information for subsequent service execution.
Step 104: performing session management until necessary information is collected, specifically including:
dialogs and templates for replies are specified for different office application tasks, with multiple rounds of dialog with the user until the necessary information is collected. FIG. 3 is a schematic diagram of session management according to the present invention.
The user's dialog process includes intentions for specific task requirements, such as asking for a leave application, and requires multiple rounds of dialog to collect the slots needed to complete the task, and thus multiple rounds of dialog management are required. The main functions of dialog management are to update dialog state and information and to select one or more predefined system actions. Each predefined action is associated with a slot that needs to be collected, such as "what is your leave asking type? "slot for guiding user to answer the leave type.
Considering that the task in the office scene is relatively clear, when the intention identifies the specific intention of the user, in the next multiple rounds of conversations, the reply and guidance are carried out around the current intention until the slots needed by the task are collected. A plurality of slots exist in one dialogue of a user, for example, "tomorrow me please leave things", the current dialogue system obtains two slots of "tomorrow" and "leave things" through slot filling, then, the two slots of "start time" and "leave things type" are not inquired, and repeated inquiry efficiency is prevented from being reduced.
The user may be currently in a leave intention and have switched intentions in between, such as being changed to a carriage dispatch intention. At this time, the dialog management needs to clarify the intention of the user, that is, ask the user and confirm whether to jump to a new intention. After the user confirms, jump to the new intention and re-run the multi-turn dialog for the intended task.
In addition to multiple rounds of conversation for a task, the user may ask other intentions, such as asking a more fixed answer question, such as an reimbursement process. At this point, the most relevant answer to the question is returned to the user using similarity-based matching.
Step 105: determining a current user intent and a current slot value according to the intent recognition model, the slot extraction model and the necessary information;
step 106: executing a service according to the current user intention and the current slot value, specifically comprising:
judging whether the task meets the task requirement or not according to the current user intention;
if the task requirements are met, executing appointed service calling according to the slot value;
and if the task requirements are not met, executing the service by a question-answer similarity matching method. As shown in fig. 4.
After collecting the required slots for a specific task, these slots are used as parameters to call the interface in the office scene, such as making a leave application, meeting room reservation, etc.
For non-task type conversation requirements, such as a question and answer process, question and answer pairs are predefined, similar questions and corresponding answers are found through a cosine similarity matching method based on word vectors, and the answers are returned.
a. Pre-training word vectors, pre-training word vectors using the public corpus obtained in step 102.
b. And performing jieba word segmentation on the input sentences to obtain word vectors of each word, and splicing the word vectors into a word vector sequence.
c. And c, performing cosine similarity matching on the word vector of the input statement and the word vector of the question statement trained in the step a, and finding out the question statement most similar to the input statement. The cosine similarity matching formula is as follows:
where a and B are the word vector sequences of the two sentences.
d. And taking the answer corresponding to the question sentence as output.
The invention provides a task-type multi-turn dialogue system construction method for an office scene. Firstly, defining different task scenes according to different requirements of office scenes; collecting real data of a user, and generating a part of data by using a template to enhance the data; natural language understanding, namely combining a deep learning model and rules, recognizing intentions and extracting groove values through the trained deep learning model, and processing partial groove values by using domain rules; conversation management, namely formulating a reply template and a plurality of conversation processes according to a slot required by a task; and executing the service, wherein the service calling and the question answering are matched.
The invention customizes the task of the office scene, can accurately process different requirements, utilizes the user data to formulate the template to enhance the data, and combines deep learning and rules to more effectively acquire the information.
Corresponding to the task-based multi-turn dialog method of an office scenario of the present invention, the present invention further provides a task-based multi-turn dialog system of an office scenario, as shown in fig. 5, the task-based multi-turn dialog system of an office scenario includes:
and the scene definition module 201 is configured to perform scene definition on different office application tasks to obtain intents and slot values of the different office application tasks.
And the data acquisition module 202 is configured to perform data acquisition on the intents and slot values of the different office application tasks to obtain acquired data.
And the natural language understanding module 203 is used for determining an intention recognition model and a slot extraction model by using a deep learning model training method and a customized rule processing method according to the collected data.
And a session management module 204 for performing session management until necessary information is collected.
A current user intent and current slot value determining module 205 for determining a current user intent and a current slot value based on the intent recognition model, the slot extraction model, and the necessary information.
And a service execution module 206, configured to execute a service according to the current user intention and the current slot value.
The natural language understanding module 203 specifically includes:
and the intention recognition model determining unit is used for training the deep learning model Bert + TextCn according to the acquired data to obtain an intention recognition model.
And the non-time slot extraction model determining unit is used for training the Bert-BilSTM-CRF model according to the acquired data to obtain a non-time slot extraction model.
And the time slot extraction model determining unit is used for processing the acquired data by adopting a regular expression rule based on common knowledge date to obtain a time slot extraction model.
The dialog management module 204 specifically includes:
and the conversation management unit is used for specifying the conversation and the template of the reply for different office application tasks and carrying out multiple rounds of conversation with the user until necessary information is collected.
The service execution module 206 specifically includes:
and the judging unit is used for judging whether the task meets the task requirement according to the current user intention.
And the first service execution unit is used for executing the appointed service call when the task requirement is met.
And the second service execution unit is used for executing the service by a question-answer similarity matching method when the task requirements are not met.
According to the invention, through office scene analysis, data enhancement and deep learning and rules, a task-type multi-turn conversation method and system of an office scene are constructed, multi-turn conversation collection information can be automatically carried out with office users, corresponding services are executed, and the office efficiency is improved.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (8)
1. A task-based multi-turn dialog method for an office scenario, comprising:
scene definition is carried out on different office application tasks, and intentions and slot values of the different office application tasks are obtained;
acquiring data of the intentions and the slot values of the different office application tasks to obtain acquired data;
determining an intention recognition model and a slot extraction model according to the acquired data by using a deep learning model training method and a customized rule processing method;
performing dialogue management until necessary information is collected;
determining a current user intent and a current slot value according to the intent recognition model, the slot extraction model and the necessary information;
and executing the service according to the current user intention and the current slot value.
2. The task-based multi-turn dialog method for the office scenario as claimed in claim 1, wherein the determining the intention recognition model and the slot extraction model using a deep learning model training method and a customized rule processing method according to the collected data specifically comprises:
training a deep learning model Bert + TextCn according to the acquired data to obtain an intention recognition model;
training a Bert-BilSTM-CRF model according to the acquired data to obtain a non-time slot extraction model;
and processing the collected data by adopting a regular expression rule based on common sense date to obtain a time slot extraction model.
3. The task-based multi-turn conversation method for the office scenario as claimed in claim 1, wherein the performing conversation management until collecting necessary information specifically comprises:
dialogs and templates for replies are specified for different office application tasks, with multiple rounds of dialog with the user until the necessary information is collected.
4. The task-based multi-turn dialog method for an office scenario as claimed in claim 1, wherein the performing a service according to the current user intent and the current slot value specifically comprises:
judging whether the task meets the task requirement or not according to the current user intention;
if the task requirements are met, executing appointed service calling according to the slot value;
and if the task requirements are not met, executing the service by a question-answer similarity matching method.
5. A task-based multi-turn dialog system for an office scenario, comprising:
the scene definition module is used for carrying out scene definition on different office application tasks to obtain intents and slot values of the different office application tasks;
the data acquisition module is used for acquiring data of the intentions and the slot values of the different office application tasks to obtain acquired data;
the natural language understanding module is used for determining an intention recognition model and a slot extraction model by using a deep learning model training method and a customized rule processing method according to the collected data;
the dialogue management module is used for carrying out dialogue management until necessary information is collected;
a current user intention and current slot value determining module for determining a current user intention and a current slot value according to the intention recognition model, the slot extraction model and the necessary information;
and the service execution module is used for executing the service according to the current user intention and the current slot value.
6. The task-based multi-turn dialog system of an office scenario of claim 5, wherein the natural language understanding module specifically comprises:
the intention recognition model determining unit is used for training a deep learning model Bert + TextCn according to the acquired data to obtain an intention recognition model;
the non-time slot extraction model determining unit is used for training a Bert-BilSTM-CRF model according to the acquired data to obtain a non-time slot extraction model;
and the time slot extraction model determining unit is used for processing the acquired data by adopting a regular expression rule based on common knowledge date to obtain a time slot extraction model.
7. The task-based multi-turn dialog system of an office scenario of claim 5, wherein the dialog management module specifically comprises:
and the conversation management unit is used for specifying the conversation and the template of the reply for different office application tasks and carrying out multiple rounds of conversation with the user until necessary information is collected.
8. The task-based multi-turn dialog system of an office scenario of claim 5, wherein the service execution module specifically comprises:
the judging unit is used for judging whether the task meets the task requirement or not according to the current user intention;
the first service execution unit is used for executing appointed service calling when meeting the task requirement;
and the second service execution unit is used for executing the service by a question-answer similarity matching method when the task requirements are not met.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011134560.8A CN112199486A (en) | 2020-10-21 | 2020-10-21 | Task type multi-turn conversation method and system for office scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011134560.8A CN112199486A (en) | 2020-10-21 | 2020-10-21 | Task type multi-turn conversation method and system for office scene |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112199486A true CN112199486A (en) | 2021-01-08 |
Family
ID=74010540
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011134560.8A Pending CN112199486A (en) | 2020-10-21 | 2020-10-21 | Task type multi-turn conversation method and system for office scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112199486A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112925897A (en) * | 2021-04-12 | 2021-06-08 | 辽宁工程技术大学 | Human-computer dialogue system based on task type and its realizing method |
| CN113821620A (en) * | 2021-09-18 | 2021-12-21 | 湖北亿咖通科技有限公司 | Multi-turn conversation task processing method and device and electronic equipment |
| CN114117024A (en) * | 2022-01-27 | 2022-03-01 | 永鼎行远(南京)信息科技有限公司 | Platform construction method for multi-round conversation function scene |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446306A (en) * | 2018-10-16 | 2019-03-08 | 浪潮软件股份有限公司 | A kind of intelligent answer method of more wheels dialogue of task based access control driving |
| CN109493166A (en) * | 2018-10-23 | 2019-03-19 | 深圳智能思创科技有限公司 | A kind of construction method for e-commerce shopping guide's scene Task conversational system |
| CN109816106A (en) * | 2017-11-20 | 2019-05-28 | 北京智合大方科技有限公司 | One kind carrying out call center's customer service knowledge class response quality evaluation system based on speech recognition and natural language processing technique |
| CN110209791A (en) * | 2019-06-12 | 2019-09-06 | 百融云创科技股份有限公司 | It is a kind of to take turns dialogue intelligent speech interactive system and device more |
| CN110321418A (en) * | 2019-06-06 | 2019-10-11 | 华中师范大学 | A kind of field based on deep learning, intention assessment and slot fill method |
| CN110555095A (en) * | 2018-05-31 | 2019-12-10 | 北京京东尚科信息技术有限公司 | Man-machine conversation method and device |
| CN110825865A (en) * | 2020-01-09 | 2020-02-21 | 浩鲸云计算科技股份有限公司 | Multi-round conversation intelligent customer service system based on special word correction and cold start |
| CN111078846A (en) * | 2019-11-25 | 2020-04-28 | 青牛智胜(深圳)科技有限公司 | Multi-turn dialog system construction method and system based on business scene |
| CN111078844A (en) * | 2018-10-18 | 2020-04-28 | 上海交通大学 | Task-based dialog system and method for software crowdsourcing |
| CN111104495A (en) * | 2019-11-19 | 2020-05-05 | 深圳追一科技有限公司 | Information interaction method, device, equipment and storage medium based on intention recognition |
| CN111414195A (en) * | 2019-02-03 | 2020-07-14 | 北京邮电大学 | Universal configurable, multi-model-compatible and domain-transportable multi-round dialogue system |
-
2020
- 2020-10-21 CN CN202011134560.8A patent/CN112199486A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109816106A (en) * | 2017-11-20 | 2019-05-28 | 北京智合大方科技有限公司 | One kind carrying out call center's customer service knowledge class response quality evaluation system based on speech recognition and natural language processing technique |
| CN110555095A (en) * | 2018-05-31 | 2019-12-10 | 北京京东尚科信息技术有限公司 | Man-machine conversation method and device |
| CN109446306A (en) * | 2018-10-16 | 2019-03-08 | 浪潮软件股份有限公司 | A kind of intelligent answer method of more wheels dialogue of task based access control driving |
| CN111078844A (en) * | 2018-10-18 | 2020-04-28 | 上海交通大学 | Task-based dialog system and method for software crowdsourcing |
| CN109493166A (en) * | 2018-10-23 | 2019-03-19 | 深圳智能思创科技有限公司 | A kind of construction method for e-commerce shopping guide's scene Task conversational system |
| CN111414195A (en) * | 2019-02-03 | 2020-07-14 | 北京邮电大学 | Universal configurable, multi-model-compatible and domain-transportable multi-round dialogue system |
| CN110321418A (en) * | 2019-06-06 | 2019-10-11 | 华中师范大学 | A kind of field based on deep learning, intention assessment and slot fill method |
| CN110209791A (en) * | 2019-06-12 | 2019-09-06 | 百融云创科技股份有限公司 | It is a kind of to take turns dialogue intelligent speech interactive system and device more |
| CN111104495A (en) * | 2019-11-19 | 2020-05-05 | 深圳追一科技有限公司 | Information interaction method, device, equipment and storage medium based on intention recognition |
| CN111078846A (en) * | 2019-11-25 | 2020-04-28 | 青牛智胜(深圳)科技有限公司 | Multi-turn dialog system construction method and system based on business scene |
| CN110825865A (en) * | 2020-01-09 | 2020-02-21 | 浩鲸云计算科技股份有限公司 | Multi-round conversation intelligent customer service system based on special word correction and cold start |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112925897A (en) * | 2021-04-12 | 2021-06-08 | 辽宁工程技术大学 | Human-computer dialogue system based on task type and its realizing method |
| CN113821620A (en) * | 2021-09-18 | 2021-12-21 | 湖北亿咖通科技有限公司 | Multi-turn conversation task processing method and device and electronic equipment |
| CN113821620B (en) * | 2021-09-18 | 2023-12-12 | 亿咖通(湖北)技术有限公司 | Multi-round dialogue task processing method and device and electronic equipment |
| CN114117024A (en) * | 2022-01-27 | 2022-03-01 | 永鼎行远(南京)信息科技有限公司 | Platform construction method for multi-round conversation function scene |
| CN114117024B (en) * | 2022-01-27 | 2022-05-03 | 永鼎行远(南京)信息科技有限公司 | Platform construction method for multi-round conversation function scene |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110413746B (en) | Method and device for identifying intention of user problem | |
| CN112199486A (en) | Task type multi-turn conversation method and system for office scene | |
| Merdivan et al. | Dialogue systems for intelligent human computer interactions | |
| CN110543554A (en) | Classification method and device for multi-turn conversations | |
| CN111339302A (en) | Method and device for training element classification model | |
| Chi et al. | Speaker role contextual modeling for language understanding and dialogue policy learning | |
| Wu et al. | A probabilistic framework for representing dialog systems and entropy-based dialog management through dynamic stochastic state evolution | |
| CN113268610A (en) | Intent skipping method, device and equipment based on knowledge graph and storage medium | |
| Petukhova et al. | Modelling multi-issue bargaining dialogues: Data collection, annotation design and corpus | |
| CN110890088A (en) | Voice information feedback method and device, computer equipment and storage medium | |
| CN110473543B (en) | Voice recognition method and device | |
| CN113761156A (en) | Data processing method, device and medium for man-machine interaction conversation and electronic equipment | |
| CN117573834B (en) | Multi-robot dialogue method and system for software-oriented instant service platform | |
| CN111399629B (en) | Operation guiding method of terminal equipment, terminal equipment and storage medium | |
| CN116821290A (en) | Multitasking dialogue-oriented large language model training method and interaction method | |
| CN112784024B (en) | Man-machine conversation method, device, equipment and storage medium | |
| CN111414732A (en) | Text style conversion method and device, electronic equipment and storage medium | |
| CN117370512A (en) | Method, device, equipment and storage medium for replying to dialogue | |
| CN111966809A (en) | Method and device for carrying out knowledge question answering based on multiple rounds of conversations | |
| CN111680148B (en) | Method and device for intelligently responding to question of user | |
| CN113468306A (en) | Voice conversation method, device, electronic equipment and storage medium | |
| CN114201595A (en) | Sentence recommendation method and device in conversation, storage medium and electronic equipment | |
| CN115408500A (en) | Question-answer consistency evaluation method and device, electronic equipment and medium | |
| CN112328774A (en) | Method for realizing task type man-machine conversation task based on multiple documents | |
| CN110888971A (en) | Multi-round interaction method and device for robot customer service and user |
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 |