CN114661899A - Task creating method and device, computer equipment and storage medium - Google Patents
Task creating method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN114661899A CN114661899A CN202210138389.0A CN202210138389A CN114661899A CN 114661899 A CN114661899 A CN 114661899A CN 202210138389 A CN202210138389 A CN 202210138389A CN 114661899 A CN114661899 A CN 114661899A
- Authority
- CN
- China
- Prior art keywords
- task
- information
- conversation
- machine learning
- learning model
- 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 67
- 238000010801 machine learning Methods 0.000 claims abstract description 68
- 238000012549 training Methods 0.000 claims description 47
- 230000008569 process Effects 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 230000004048 modification Effects 0.000 claims description 10
- 238000012986 modification Methods 0.000 claims description 10
- 230000015654 memory Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000010200 validation analysis Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 6
- 239000013589 supplement Substances 0.000 claims description 5
- 230000003993 interaction Effects 0.000 description 9
- 238000004891 communication Methods 0.000 description 6
- 230000002457 bidirectional effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000002452 interceptive effect Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000002618 waking effect Effects 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000007704 transition 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
- G06F16/35—Clustering; Classification
-
- 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
- G06F16/38—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- 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
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
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)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Library & Information Science (AREA)
- Human Computer Interaction (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a task creating method, a task creating device, computer equipment and a storage medium, wherein in a specific implementation mode, the method comprises the following steps: s2, obtaining conversation contents among users in a variable time window of t minutes in a conversation group and obtaining conversation information in the conversation contents; s4, inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information; and S6, creating a task according to the task information. According to the embodiment of the invention, the requirement of the user is automatically identified by setting the variable time window, so that the intelligent conversation assistant does not need to have conversation with the user in group chat to obtain complete task elements, thereby creating a task, and solving the problems that the traditional intelligent assistant forcibly intervenes in the conversation and the user experience is poor.
Description
Technical Field
The present invention relates to the field of electronic communications technologies, and in particular, to a task creation method and apparatus, a computer device, and a storage medium.
Background
With the rapid development of mobile intelligent terminals and cloud computing, the wave of artificial intelligence is quietly turning over the point of life, intelligent conversation user interaction is rapidly developed as a new field, and more new requirements on the aspects of linguistics, emotional modeling, logic building and the like are provided for user experience. The intelligent dialogue user interaction is a new generation interaction mode based on voice input or text input, and feedback results can be obtained by speaking or typing. A typical application scenario is an intelligent conversation assistant. The intelligent conversation assistant is an intelligent application, and can help a user to solve problems and create tasks such as meeting reservation room, conference participant reminding and meeting through intelligent conversation.
The existing mode of creating tasks by the intelligent conversation assistant is a circular mode in which a user actively provides input and the intelligent conversation assistant passively provides feedback in a private chat mode between the user and the intelligent conversation assistant. The intelligent conversation assistant will only be subject to the user instructions singularly. However, this passive mechanism does not enable a solution for intelligent conversation assistants to create tasks without having to talk to the user in group chat.
In chinese patent document entitled "a voice interaction apparatus, method and computer readable storage medium", publication number CN110827821A describes a voice interaction apparatus, including: the control module is used for sending information determined based on the user intention to a user when the user intention is determined to have an intervention requirement and an intervention occasion is determined to arrive based on the session parameters. Although the invention is an active mechanism, the related session parameters such as the speed of speech, the conversation frequency and the like are only suitable for a voice interaction scene and are not suitable for a text chat scene of the instant messaging software, and the applied scenes are several scenes such as chat, discussion, question and answer and fixed conversation programs and are not suitable for task type conversation scenes such as an intelligent assistant, and even if the invention is applied to the voice chat scene of the instant messaging software, the group chat experience can be greatly influenced according to the frequent intervention of the session parameters.
Disclosure of Invention
An object of the present invention is to provide a task creation method, apparatus, computer device, and storage medium.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a task creating method in a first aspect, which comprises the following steps:
s2, obtaining conversation contents among users in a variable time window of t minutes in a conversation group and obtaining conversation information in the conversation contents;
s4, inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information;
and S6, creating a task according to the task information.
Optionally, the dialog content comprises text dialog content and/or voice dialog content.
Optionally, before the S2, the task creating method further includes:
and S0, training according to the plurality of dialogue information training samples to obtain an intention recognition machine learning model and a named entity recognition machine learning model.
Optionally, the S0 further includes:
training a machine learning model for intent recognition according to a plurality of dialogue information training samples comprises the following steps:
the data set uses the historical conversation content of the instant chat tool as a training sample, and is divided into a training set, a verification set and a test set according to hierarchical sampling;
training the intent recognition machine learning model using a training set and a validation set;
verifying the intention recognition machine learning model by using a test set to obtain a trained intention recognition machine learning model;
the training of the training samples according to the plurality of dialogue information to obtain the named entity recognition machine learning model comprises the following steps:
the data set uses the historical conversation content of the instant chat tool as a training sample, and is divided into a training set, a verification set and a test set according to hierarchical sampling;
training the named entity recognition machine learning model using a training set and a validation set;
and verifying the named entity recognition machine learning model by using a test set to obtain the trained named entity recognition machine learning model.
Optionally, after the S4, before the S6, the task creating method further includes:
s5, extracting task elements according to the task information, and judging whether the task elements are complete: if yes, the dialog content covered in the currently set variable time window t minutes includes all required task elements, and the process proceeds to the step S6; if not, adjusting the variable time window value to be t + n minutes, acquiring the conversation content among the users in the variable time window t + n minutes in the conversation group, and acquiring the conversation information in the conversation content; inputting the dialogue information into a trained intention recognition machine learning model and a named entity recognition machine learning model to obtain task information; extracting task elements according to task information, and judging whether the task elements are complete: if yes, go to S6; if not, the dialog group issues an inquiry statement for guiding the user to supplement information until the task elements are complete, and the process proceeds to S6.
Optionally, after S6, the task creating method further includes:
and S7, publishing the created task in the conversation group.
Optionally, after S7, the task creating method further includes:
and S8, obtaining the dialogue content of the user in the dialogue group and obtaining the dialogue information in the dialogue content, inputting the dialogue information into the trained intention recognition machine learning model and named entity recognition machine learning model for intention recognition and named entity recognition to obtain task modification information, and modifying the created task according to the task modification information.
A second aspect of the present invention provides a task creating device that uses the task creating method according to the first aspect of the present invention, including:
the acquisition module is used for acquiring conversation contents among users in a variable time window of t minutes in a conversation group and acquiring conversation information in the conversation contents;
the input module is used for inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information;
and the creating module is used for creating the task according to the task information.
A third aspect of the invention provides a computer apparatus comprising a processor and a memory, the memory having stored thereon a computer program which, when executed by the processor, performs the method according to the first aspect of the invention.
A fourth aspect of the invention provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the method according to the first aspect of the invention.
The invention has the following beneficial effects:
aiming at the existing problems, the invention sets a task creating method, provides an intelligent conversation assistant based on intention identification and named entity identification, and automatically identifies the requirements of a user by setting a variable time window, so that the intelligent conversation assistant does not need to have conversation with the user in group chat to obtain complete task elements, thereby creating a task and solving the problems that the traditional intelligent assistant forcibly intervenes in the conversation and the user experience is poor.
Drawings
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
Fig. 1 shows a flowchart of a task creation method according to an embodiment of the present invention.
FIG. 2 illustrates a named entity recognition machine learning model structure diagram.
Fig. 3 is a schematic diagram of a task creation apparatus according to an embodiment of the present invention.
Fig. 4 shows a schematic structural diagram of a computer device.
Detailed Description
In order to more clearly illustrate the invention, the invention is further described below with reference to preferred embodiments and the accompanying drawings. Similar parts in the figures are denoted by the same reference numerals. It is to be understood by persons skilled in the art that the following detailed description is illustrative and not restrictive, and is not to be taken as limiting the scope of the invention.
The inventor analyzes the way of creating tasks by the intelligent conversation assistant in the prior art and finds that a plurality of problems exist. For example, taking the current intelligent conversation assistant as an example, the technical problems existing in the intelligent conversation user interaction technology in the prior art are explained.
Currently, there are many technical problems for intelligent conversation assistants in multi-person scenes and natural interactions, including and not limited to:
(1) the user and the intelligent conversation assistant or the voice assistant are in a private chat state and are not in a group chat scene, namely the current intelligent conversation assistant or voice assistant is usually aimed at a single user and generally ignored for a multi-user chat scene.
(2) The user and the intelligent conversation assistant can only provide input actively and provide feedback passively in a private chat mode. The intelligent conversation assistant will only be subject to the user instructions singularly. However, this passive mechanism does not enable a solution for intelligent conversation assistants to create tasks without having to talk to the user in group chat.
(3) The intelligent conversation assistant, such as various APP customer service robots, AI sound boxes and the like, need to wake up words, the interactive technology adopting the wake-up words is a scheme adopted by most of the existing intelligent assistants or voice assistant products, and is based on one-time conversation design, namely, a user wakes up first, inputs a problem, and the intelligent assistant or voice assistant identifies and gives the most possible answer. The interactive dialog is lack of relevance, the intelligent assistant or the voice assistant does not understand the context, the intelligent assistant or the voice assistant cannot recognize the user intention due to incomplete information, and no response or wrong instruction execution is possible, so that the problems of 'waking up without waking up' or 'mistaken waking up' and the like occur. And each conversation needs to wake up a word, and the rhythm is stuck, so that the interactive experience of the user is influenced.
(4) When the intervention requirement is determined based on the user intention and the intervention opportunity is determined based on the session parameters, sending information determined based on the user intention to the user, wherein the related session parameters are the speed of speech, the conversation frequency and the like, are only suitable for a voice interaction scene and are not suitable for an instant messaging software text chat scene, the applied scenes are several scenes such as a chat scene, a discussion scene, a question-answer scene and a fixed conversation program and are not suitable for a task type conversation scene like an intelligent assistant, and even if the application is applied to the instant messaging software voice chat scene, the group chat experience can be greatly influenced according to the frequent intervention session of the session parameters.
In view of this, as shown in fig. 1, an embodiment of the present invention provides a task creating method, including:
and S0, training according to the plurality of dialogue information training samples to obtain an intention recognition machine learning model and a named entity recognition machine learning model.
The method comprises the steps of collecting historical conversation contents of an instant chat tool as training samples, training two machine learning models of a natural language understanding module, wherein one model is used for identifying intentions, the other model is used for identifying named entities, different intentions correspond to different named entity identification models, and each model is trained by using corresponding corpora.
The intention-recognition machine learning model applies the Bert model, which is a pre-trained model and is a two-stage model that includes pre-training and Fine-tuning. The model of the downstream task (Ruyi graph recognition) can be fine-tuned by loading a pre-trained model (e.g., BERT-based-uncapped, BERT-based-chinese). Adding softmax activation function to the output layer of the Bert network for intention identification.
The specific process is as follows:
the data set uses the historical conversation content of the instant chat tool as a training sample and comprises intentions of booking a conference room, following up project progress, having no task and the like;
dividing a data set into a training set, a verification set and a test set according to hierarchical sampling;
training the intent recognition machine learning model using a training set and a validation set;
verifying the intention recognition machine learning model by using a test set to obtain a trained intention recognition machine learning model, which specifically comprises the following steps:
applying a Bert token to map tokens to word embeddings in all samples;
encoding the data set;
loading a pre-training model bert-base-chip;
compiling and training the model: adopting an Adam optimizer, and setting a learning rate learning _ rate to be 2e-5 and a training round number of _ epoch to be 8;
the trained machine learning model is deployed on a server, and the task creation tool communicates with the server.
The named entity recognition machine learning model applies a BERT + BilSTM + CRF model to carry out named entity recognition, namely entity recognition. Named entity recognition refers to recognition of entities with specific meanings in text, and mainly includes names of people, places, organizations, proper nouns, time and the like.
BERT (bidirectional Encoder retrieval from transformers), the Encoder of a bidirectional Transformer. The model respectively captures expression of word and Sentence levels by using a Masked LM method and a Next sequence Prediction method on a pre-training method.
BilSTM is an abbreviation of Bi-directional Long Short-Term Memory, and is formed by combining forward LSTM and backward LSTM.
CRF is a conditional random field that can be used to construct a conditional probability distribution model of one set of output random variables given a set of input random variables.
The model BERT + BilSTM + CRF consists of BERT modules, BilSTM and CRF 3 modules. The overall model is shown in fig. 2. Firstly, obtaining a word vector by using a BERT model, and extracting important features of a text; then named entity recognition is carried out through BiILSTM deep learning context feature information; and finally, processing the output sequence of the BilSTM by the CRF layer, and obtaining a global optimal sequence according to the labels between adjacent layers by combining the state transition matrix in the CRF.
The first layer of the model is to obtain the word vector in the input text information by using the initialization of the pre-trained BERT language model as the sequence X ═ X (X)1,x2,x3,…,xn) The obtained word vector can effectively extract the characteristics in the text by utilizing the mutual relation between the words;
the second layer of the model is a bidirectional LSTM layer, the n-dimensional word vector acquired by the first layer is used as the input of each time step of a bidirectional long-time memory neural network to obtain the hidden state sequence of the bidirectional LSTM layer
And
after the forward and backward processing is finished, all the hidden state sequences are spliced according to the positions to obtain a complete hidden state sequence ht＝(h1,h2,…,hn）∈Rn×mThen, the linear output layer maps the complete hidden state sequence to s dimension (s dimension is the number of label categories of the labeled set), and records the extracted sentence features as the sequence after all mappings as a matrix L (L)1,l2,…,ln)∈Rn×s，li∈RsEach dimension l ofi,jRespectively correspond to the character xiLabel y for each categoryiFractional value of (c). If the score values of all positions are directly and independently classified at the moment, the output result is directly obtained by selecting the highest score value, information between adjacent sentences cannot be considered, the global optimum cannot be obtained, and the classification result is not ideal. The last layer of the model is introduced.
The specific process is as follows:
the data set uses the instant chat tool historical conversation content as a training sample, which is processed into the following format:
@ tomorrow afternoon three-point # TIME/at @ Shanghai conference room # MEETINGROOM/conference
Good, how long the conference lasts?
About @ two hours # TIME/Bar
Good
Where @ is followed by the entity and # is followed by the label of the entity.
Dividing a data set into a training set, a verification set and a test set according to level sampling;
training the named entity recognition machine learning model using a training set and a validation set;
verifying the named entity recognition machine learning model by using a test set to obtain a trained named entity recognition machine learning model, which specifically comprises the following steps:
loading a pre-training model, namely, Chinese _ L-12_ H-768_ A-12;
compiling and training the model: an Adam optimizer is adopted, the learning rate is selected to be 0.001, LSTM _ dim is set to be 200, batch _ size is set to be 64, and max _ seq _ len is set to be 128. In order to prevent the over-fitting problem, Dropout is used in the input and the output of the BiLSTM, and the value is 0.5;
the trained machine learning model is deployed on a server, and the task creation tool communicates with the server.
In one particular example, the conversation content includes textual conversation content and/or voice conversation content.
And S2, obtaining the conversation content among the users in the variable time window t minutes in the conversation group and obtaining the conversation information in the conversation content.
In the embodiment of the invention, a variable time window value t is set to be 5 minutes, when group chat information exists, a dialogue information storage module starts to store dialogue information, and if the dialogue information is character information, the character information is directly used as the dialogue information; if the dialogue information is voice information, the voice information is firstly converted into character information, and then the converted character information is used as the dialogue information.
And acquiring the conversation contents including voice and characters among the users in the last five minutes in the conversation group from a conversation information storage module, and acquiring the conversation information in the conversation contents according to the conversation contents.
And S4, inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information.
In the following example, the obtained dialogue information is input into the intention recognition machine learning model in the natural language understanding module to obtain intentions related to the dialogue information, such as intentions of meeting room reservation, item follow-up progress, no task and the like, if the obtained dialogue information is recognized as no task, the dialogue content stored in the dialogue information storage module is emptied, the dialogue information in t minutes to 2t minutes is continuously stored, the step is repeated, and if the obtained dialogue information is recognized as the meeting room reservation intentions, the dialogue information is input into the named entity recognition machine learning model of the corresponding intentions in the natural language understanding module to obtain the task information.
In one possible implementation manner, as shown in fig. 1, after the S4 and before the S6, the task creating method further includes:
s5, extracting task elements according to the task information, and judging whether the task elements are complete: if yes, the dialog content covered in the currently set variable time window t minutes includes all required task elements, and the process proceeds to the step S6; if not, adjusting the variable time window value to be t + n minutes, acquiring the conversation content among the users in the variable time window t + n minutes in the conversation group, and acquiring the conversation information in the conversation content; inputting the dialogue information into a trained named entity recognition machine learning model to obtain task information; extracting task elements according to task information, and judging whether the task elements are complete: if yes, go to S6; if not, the dialog group issues an inquiry statement for guiding the user to supplement information until the task elements are complete, and the process proceeds to S6.
Continuing with the above example, for example, if it is recognized as the conference room reservation intention, the dialog information is input into the named entity recognition machine learning model of the corresponding intention in the natural language understanding module, a specific task element is recognized, for example, the task element included in the conference room reservation intention has { "conference room reservation" [ "start time", "duration", "conference room name" ], … … } and the like, the specific task element is matched with the corresponding relation of the recognition intention, it is determined whether the obtained task element is complete, if complete, the process goes to S6, if incomplete adjusting the variable time window value t to 5+2 minutes, that is, the dialog information storage module adds 2 minutes of dialog information, the natural language understanding module named entity recognition machine learning model is input again, if complete, the process goes to S6, if not complete, there is a high possibility that no task element which is lacked in the dialog process appears any more, then the task management module issues an inquiry statement to the group that guides the user to supplement information, such as "whether to create a task for booking a conference room, start time: XXX, conference room location: XXX, if the task is created and the conference duration is required to be supplemented, or else, the task management module is required to ignore ", after the task management module issues a word and sentence query statement, the dialogue information storage module starts another branch to store n ═ 10 group chat dialogue information after the query statement is stored, the n ═ 10 group chat dialogue information after the query statement is stored is input into the natural language understanding module named entity recognition machine learning model, if the missing task element is obtained, the complete task element is supplemented, the dialogue information storage module clears the n ═ 10 group chat dialogue information after the query statement is stored, the operation goes to S6, and if the missing task element is not obtained, the user is considered not to create the task by default.
And S6, creating a task according to the task information.
In a specific example, a task is created according to the complete task elements, after the dialog information and the query statement stored in the dialog information storage module in the latest variable time window t minutes are cleared, n is 10 pieces of group chat dialog information, the variable time window value t is restored to 5 minutes, and when there is group chat information in the group again, the process proceeds to S2.
In a possible implementation manner, as shown in fig. 1, after S6, the task creating method further includes:
and S7, publishing the created task in the dialog group.
In a specific example, the task management module directly performs task operations such as booking a conference room and displaying the task operations in a conversation group, clears the conversation information stored in the conversation information storage module, and restores the variable time window value t to 5 minutes.
In one possible implementation manner, as shown in fig. 1, after S7, the task creating method further includes:
s8, obtaining the dialogue content of the user in the dialogue group and obtaining the dialogue information in the dialogue content, inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to perform intention recognition and named entity recognition so as to obtain task modification information, and modifying the created task according to the task modification information.
In a specific example, if the user needs to cancel or change, the user can say the corresponding words "cancel task" and "modify task", the dialogue information storage module stores the query sentence, then n is 10 pieces of group chat dialogue information, then the stored query sentence, then n is 10 pieces of group chat dialogue information are input into the natural language understanding module, the intention recognition machine learning model is used for recognizing the intention involved in the dialogue information, the dialogue information storage module removes the stored query sentence, then n is 10 pieces of group chat dialogue information, if the user recognizes that the user cancels the task, the task management module deletes the task just created, the user switches to the process of removing the dialogue information and the query sentence stored in the dialogue information storage module within the latest variable time window t minute, then n is 10 pieces of group chat dialogue information, and the variable time window value t is reduced to 5 minutes; if a modification task is identified, the task management module presents the just created task and task elements in the conversation group and asks which element to modify, the dialogue information storage module stores the inquiry sentence, n is 10 pieces of group chat dialogue information, if the user sends the task element and the content to be modified, the n is 10 pieces of group chat dialogue information after the inquiry sentence, inputting 10 pieces of group chat dialogue information after the query sentence into a natural language understanding module named entity recognition machine learning model, if the task elements are recognized, correspondingly modifying the task elements of the task by a task management module, if the task elements are not recognized, ignoring the modification operation, the method proceeds to clearing the dialog information and the query statement stored in the dialog information storage module within the latest variable time window t minutes, then changing n to 10 group chat dialog information, and changing the variable time window value t to 5 minutes.
It should be noted that specific values of t and n may be adjusted according to an actual group chat scenario, which is not limited in the present invention.
In a specific example, based on the task creation method provided by the embodiment of the invention, the conversation content is analyzed through an instant chat tool and an AI algorithm, automatically creating tasks and managing tasks according to task elements, such as intelligent assistants (secretaries) realized in plug-in form in office type APPs of enterprise WeChat, nailing and the like, according to the conversation content (text or voice) in the conversation group, the task intention (for example, to reserve a meeting room to meet a client) is identified through the trained AI model, task elements are collected "silently" in a variable time window mode, only if the conversation content in the group does not relate to the task elements, information inquiring whether to establish the task is sent out to the group, and leads to supplement of the missing task elements, and if the acquiescent enough task elements are obtained, the task elements such as a reserved conference room are directly operated and displayed in the conversation group.
The intelligent conversation assistant provided by the invention is different from the small conversation assistant in that the intelligent conversation assistant does not need to be awakened and is automatically intervened. The distinction from an automated customer service assistant is that instead of asking the user on his own initiative, the task intent is automatically recognized through a dialog between the users.
For example, Zhang III in the enterprise WeChat group sends a reminder for me to participate in a sharing meeting at ten am of the next Monday, and the small assistant automatically replies a reminder of the added schedule, Zhang III and adds a reminder item in the enterprise WeChat;
for another example, zhang san in the enterprise wechat group sends "@ li si, tomorrow communication project progress", the small assistant automatically replies "whether a task @ zhang needs to be established", zhang "@ li si shang hai conference room in morning on tomorrow", the small assistant automatically replies "specific several points @ zhang", zhang san "nine am", the small assistant automatically subscribes the conference room and automatically replies "ordered conference room @ li";
for example, Zhang III directly says which meeting room is clear, then the assistant automatically orders the meeting room and replies in the group.
According to the task creating method provided by the embodiment of the invention, according to the conversation content (characters or voice) among the users in the conversation group, the intention of the task is identified through a trained AI model (for example, a meeting client in a conference room needs to be reserved), the task elements are collected through a variable time window mode, only if the conversation content in the group does not relate to the task elements, information inquiring whether to establish the task is sent out to the group, and the missing task elements are guided to be supplemented, if enough task information is obtained through the acquiescence mode, the task operation such as the conference room reservation and the like is directly carried out and displayed in the conversation group, so that the intelligent conversation assistant in the group chat does not need to have the conversation with the users, the task creation is carried out after the complete task elements are obtained, and the problems that the traditional intelligent assistant forcibly intervenes the conversation and the user experience is poor are solved.
Another embodiment of the present invention provides a task creating device applying the task creating method, as shown in fig. 3, the device includes:
the acquisition module is used for acquiring conversation contents among users in a variable time window of t minutes in a conversation group and acquiring conversation information in the conversation contents;
the input module is used for inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information;
and the creating module is used for creating the task according to the task information.
As shown in fig. 4, a computer system suitable for implementing the task creation method provided by the above-described embodiments includes a central processing module (CPU) that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) or a program loaded from a storage section into a Random Access Memory (RAM). In the RAM, various programs and data necessary for the operation of the computer system are also stored. The CPU, ROM, and RAM are connected thereto via a bus. An input/output (I/O) interface is also connected to the bus.
The following components are connected to the I/O interface: an input section including a keyboard, a mouse, and the like; an output section including a speaker and the like such as a Liquid Crystal Display (LCD); a storage section including a hard disk and the like; and a communication section including a network interface card such as a LAN card, a modem, or the like. The communication section performs communication processing via a network such as the internet. The drive is also connected to the I/O interface as needed. A removable medium such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive as necessary, so that a computer program read out therefrom is mounted into the storage section as necessary.
In particular, the processes described in the above flowcharts may be implemented as computer software programs according to the present embodiment. For example, the present embodiments include a computer program product comprising a computer program tangibly embodied on a computer-readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication section, and/or installed from a removable medium.
The flowchart and schematic diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to the present embodiments. In this regard, each block in the flowchart or schematic diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the schematic and/or flowchart illustration, and combinations of blocks in the schematic and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
On the other hand, the present embodiment also provides a nonvolatile computer storage medium, which may be the nonvolatile computer storage medium included in the apparatus in the foregoing embodiment, or may be a nonvolatile computer storage medium that exists separately and is not assembled into a terminal. The non-volatile computer storage medium stores one or more programs that, when executed by a device, cause the device to:
obtaining conversation contents among users in a variable time window of t minutes in a conversation group and obtaining conversation information in the conversation contents;
inputting the dialogue information into a trained intention recognition machine learning model and a named entity recognition machine learning model to obtain task information;
and creating a task according to the task information.
In the description of the present invention, it should be noted that relational terms such as first and second, and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
It should be understood that the above-described embodiments of the present invention are examples for clearly illustrating the invention, and are not to be construed as limiting the embodiments of the present invention, and it will be obvious to those skilled in the art that various changes and modifications can be made on the basis of the above description, and it is not intended to exhaust all embodiments, and obvious changes and modifications can be made on the basis of the technical solutions of the present invention.
Claims (10)
1. A task creation method, comprising:
s2, obtaining conversation contents among users in a variable time window of t minutes in a conversation group and obtaining conversation information in the conversation contents;
s4, inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information;
and S6, creating a task according to the task information.
2. The method of claim 1, wherein the dialog content comprises textual dialog content and/or voice dialog content.
3. The method according to claim 1, wherein before the S2, the task creating method further comprises:
and S0, training according to the plurality of dialogue information training samples to obtain an intention recognition machine learning model and a named entity recognition machine learning model.
4. The method according to claim 3, wherein the S0 further comprises:
training a purpose recognition machine learning model according to a plurality of dialogue information training samples comprises the following steps:
the data set uses the historical conversation content of the instant chat tool as a training sample, and is divided into a training set, a verification set and a test set according to hierarchical sampling;
training the intent recognition machine learning model using a training set and a validation set;
verifying the intention recognition machine learning model by using a test set to obtain a trained intention recognition machine learning model;
the training of the training samples according to the plurality of dialogue information to obtain the named entity recognition machine learning model comprises the following steps:
the data set uses the historical conversation content of the instant chat tool as a training sample, and is divided into a training set, a verification set and a test set according to hierarchical sampling;
training the named entity recognition machine learning model using a training set and a validation set;
and verifying the named entity recognition machine learning model by using a test set to obtain the trained named entity recognition machine learning model.
5. The method of claim 1, wherein after the S4, prior to the S6, the task creation method further comprises:
s5, extracting task elements according to the task information, and judging whether the task elements are complete: if yes, the dialog content covered in the currently set variable time window t minutes includes all required task elements, and the process proceeds to the step S6; if not, adjusting the variable time window value to be t + n minutes, acquiring the conversation content among the users in the variable time window t + n minutes in the conversation group, and acquiring the conversation information in the conversation content; inputting the dialogue information into a trained intention recognition machine learning model and a named entity recognition machine learning model to obtain task information; extracting task elements according to the task information, and judging whether the task elements are complete: if yes, go to S6; if not, the dialog group issues an inquiry statement for guiding the user to supplement information until the task elements are complete, and the process proceeds to S6.
6. The method according to claim 1, wherein after the S6, the task creation method further comprises:
and S7, publishing the created task in the conversation group.
7. The method according to claim 6, wherein after the S7, the task creating method further comprises:
s8, obtaining the dialogue content of the user in the dialogue group and obtaining the dialogue information in the dialogue content, inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to perform intention recognition and named entity recognition so as to obtain task modification information, and modifying the created task according to the task modification information.
8. A task creating apparatus to which the task creating method according to any one of claims 1 to 7 is applied, characterized by comprising:
the acquisition module is used for acquiring conversation contents among users in a variable time window of t minutes in a conversation group and acquiring conversation information in the conversation contents;
the input module is used for inputting the dialogue information into the trained intention recognition machine learning model and the named entity recognition machine learning model to obtain task information;
and the creating module is used for creating the task according to the task information.
9. A computer device comprising a processor and a memory, the memory having stored thereon a computer program, characterized in that the processor, when executing the program, implements the method according to any of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210138389.0A CN114661899A (en) | 2022-02-15 | 2022-02-15 | Task creating method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210138389.0A CN114661899A (en) | 2022-02-15 | 2022-02-15 | Task creating method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114661899A true CN114661899A (en) | 2022-06-24 |
Family
ID=82026838
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210138389.0A Pending CN114661899A (en) | 2022-02-15 | 2022-02-15 | Task creating method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114661899A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106201424A (en) * | 2016-07-08 | 2016-12-07 | 北京甘为乐博科技有限公司 | A kind of information interacting method, device and electronic equipment |
| US20180268818A1 (en) * | 2017-03-20 | 2018-09-20 | Ebay Inc. | Detection of mission change in conversation |
| CN112131885A (en) * | 2019-06-24 | 2020-12-25 | 北京猎户星空科技有限公司 | Semantic recognition method and device, electronic equipment and storage medium |
| CN112417894A (en) * | 2020-12-10 | 2021-02-26 | 上海方立数码科技有限公司 | Conversation intention identification method and system based on multi-task learning |
| CN112579758A (en) * | 2020-12-25 | 2021-03-30 | 北京百度网讯科技有限公司 | Model training method, device, equipment, storage medium and program product |
-
2022
- 2022-02-15 CN CN202210138389.0A patent/CN114661899A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106201424A (en) * | 2016-07-08 | 2016-12-07 | 北京甘为乐博科技有限公司 | A kind of information interacting method, device and electronic equipment |
| US20180268818A1 (en) * | 2017-03-20 | 2018-09-20 | Ebay Inc. | Detection of mission change in conversation |
| CN112131885A (en) * | 2019-06-24 | 2020-12-25 | 北京猎户星空科技有限公司 | Semantic recognition method and device, electronic equipment and storage medium |
| CN112417894A (en) * | 2020-12-10 | 2021-02-26 | 上海方立数码科技有限公司 | Conversation intention identification method and system based on multi-task learning |
| CN112579758A (en) * | 2020-12-25 | 2021-03-30 | 北京百度网讯科技有限公司 | Model training method, device, equipment, storage medium and program product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3582119A1 (en) | Spoken language understanding system and method using recurrent neural networks | |
| CN111651609A (en) | Multi-turn dialogue method and system integrating knowledge graph and emotion supervision | |
| US20150279366A1 (en) | Voice driven operating system for interfacing with electronic devices: system, method, and architecture | |
| CN114691852B (en) | Man-machine conversation system and method | |
| KR101322486B1 (en) | General dialogue service apparatus and method | |
| CN111666380A (en) | Intelligent calling method, device, equipment and medium | |
| CN115309877A (en) | Dialog generation method, dialog model training method and device | |
| CN112632244A (en) | Man-machine conversation optimization method and device, computer equipment and storage medium | |
| CN116049360A (en) | Intelligent voice dialogue scene conversation intervention method and system based on client image | |
| CN112100353A (en) | Man-machine conversation method and system, computer device and medium | |
| CN112199486A (en) | Task type multi-turn conversation method and system for office scene | |
| CN111399629A (en) | Operation guiding method of terminal equipment, terminal equipment and storage medium | |
| CN113360001A (en) | Input text processing method and device, electronic equipment and storage medium | |
| CN112199502A (en) | Emotion-based poetry sentence generation method and device, electronic equipment and storage medium | |
| CN112307166A (en) | Intelligent question and answer method and device, storage medium and computer equipment | |
| CN114661899A (en) | Task creating method and device, computer equipment and storage medium | |
| CN112150103B (en) | Schedule setting method, schedule setting device and storage medium | |
| CN115983282A (en) | Prompt-based high-efficiency small sample dialogue semantic understanding method | |
| CN112506405B (en) | Artificial intelligent voice large screen command method based on Internet supervision field | |
| CN115687581A (en) | Power customer service interactive question and answer method and device based on attention mechanism | |
| CN114036268A (en) | Task type multi-turn dialogue method and system based on intention gate | |
| CN114596842A (en) | Voice interaction method and device, computer equipment and storage medium | |
| CN114220425A (en) | Chat robot system and conversation method based on voice recognition and Rasa framework | |
| CN114373443A (en) | Speech synthesis method and apparatus, computing device, storage medium, and program product | |
| CN114328857A (en) | Statement extension method, device and computer readable storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220624 |