CN114281962A - Intelligent dialogue system, method and storage medium based on robot deep learning - Google Patents
Intelligent dialogue system, method and storage medium based on robot deep learning Download PDFInfo
- Publication number
- CN114281962A CN114281962A CN202111357137.9A CN202111357137A CN114281962A CN 114281962 A CN114281962 A CN 114281962A CN 202111357137 A CN202111357137 A CN 202111357137A CN 114281962 A CN114281962 A CN 114281962A
- Authority
- CN
- China
- Prior art keywords
- subject
- branches
- predicate
- sentence
- sub
- 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.)
- Withdrawn
Links
Images
Landscapes
- Machine Translation (AREA)
Abstract
The invention discloses an intelligent dialogue system, method and storage medium based on robot deep learning, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring voice data to perform voice recognition to obtain a transcription text of the voice data; constructing an information tree based on the transcribed text, and creating sentence break nodes according to the Chinese grammar; segmenting the transcribed text based on the sentence break nodes to obtain a plurality of groups of parallel short sentences; acquiring a questioning voice, identifying a subject in the questioning voice and a corresponding questioning type, and generating a retrieval sequence; and inputting the retrieval sequence into a matching model, traversing the sub-branches, the branches and the information tree by the matching model to find the best answer, and broadcasting. The invention has the learning and scene deep conversation capability, and can deeply memorize and understand the meaning of the question and search the answer of the question besides the language processing method of one question and one answer.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to an intelligent dialogue system, method and storage medium based on robot deep learning.
Background
With the increasing development of artificial intelligence technology and the wide application of artificial intelligence customer service, such as chat robots, the means for identifying the user intention by the artificial intelligence customer service is becoming mature.
The user intention identification means that when a user communicates with the chat robot, the robot can quickly judge the real intention of the user according to direct or indirect information provided by the user. For example, the chat robot can identify travel consumption intentions of the user, such as booking air tickets, hotels and the like, according to the problems of the user.
Although intelligent voice robots have gained popularity, flexible answers are not available during the course of a conversation. For example, the robot first updates the data in the database: please remember that there are 3 families of information academy in schools, which are computer science and technology system, communication engineering system and electronic information engineering system, respectively, 200 people are recruited in computer science and technology system, 150 people are recruited in communication engineering system and 100 people are recruited in electronic information engineering system in this year. The user starts to ask questions: in the small beauty, the information colleges in the schools of the Zan have several lines. The answer is: three of them. Asking questions: there are information colleges in the schools of Zan people. The answer is: there is no answer you want.
That is, when the user asks the robot a question that matches the database to a very high degree, the user can only call the answer from the database and answer the question, but if the user changes the language order or the query method, the robot cannot call the answer from the database and answer the question, so the flexibility is low and the use feeling is poor.
Disclosure of Invention
The invention provides an intelligent dialogue system, a method and a storage medium based on robot deep learning, aiming at solving the technical problems in the background technology.
The invention adopts the following technical scheme: an intelligent dialogue method based on deep learning of a robot comprises the following steps:
acquiring voice data to perform voice recognition to obtain a transcription text of the voice data;
constructing an information tree based on the transcribed text, and creating sentence break nodes according to the Chinese grammar; segmenting the transcribed text based on the sentence break nodes to obtain a plurality of groups of parallel short sentences, wherein the short sentences are used as branches and respectively extend on the information tree;
creating an identification model, and judging each group of branches: if at least two sentence-breaking nodes of the same type still exist in the branch, continuously segmenting to obtain at least one group of sub-branches based on the sentence-breaking nodes of the same type until only one sentence-breaking node of the same type exists in any group of sub-branches;
acquiring a questioning voice, identifying a subject in the questioning voice and a corresponding questioning type, and generating a retrieval sequence; and inputting the retrieval sequence into a matching model, traversing the sub-branches, the branches and the information tree by the matching model to find the best answer, and broadcasting.
In a further embodiment, segmentation is stopped for an information tree, branch or sub-branch if there is one and only one punctuation node of the same type in the information tree, branch or sub-branch.
By adopting the technical scheme, each sentence is split into extremely short phrases, and the flexibility of conversation is increased.
In a further embodiment, the sentence break node is one or more of a subject, a predicate or an object;
the types of the sentence-breaking nodes are divided through subjects, predicates and objects.
By adopting the technical scheme, the subjects belong to sentence-break nodes of the same type, the predicates belong to sentence-break nodes of the same type, and the objects belong to sentence-break nodes of the same type.
In a further embodiment, segmenting the transcribed text based on the sentence break node and obtaining a plurality of groups of parallel short sentences specifically comprises the following steps:
counting all vocabularies in the transcribed text to generate a vocabulary data set V;
traversing the data set, extracting each subject in the data set to generate a subject data set S, and removing and updating the subject data set from the vocabulary data set to obtain a new vocabulary data set V';
using mappingsThe formula calls at least one predicate and/or object adapted to each subject from the new lexical data set V' to obtain a related short sentence f starting with the subject(s，p，o)And generating a set of phrases based on the subject.
The mapping formula is expressed as
f(s，p，o)＝Map({es}，{ep}，{eo}, Os, Op, Oo); wherein s represents a subject, p predicate, and o represents an object; given a subject, predicate and object body, Os, Op and Oo respectively, the mapping from the ontology Os to Op and Oo means that for each entity in the ontology Os, a corresponding entity is found in the ontology Op and Oo that conforms to the chinese grammar, where es∈Os，ep∈Op，eoE is equal to Oo, and,
{es}，{ep}，{eodenotes the elements in the subject data set S, the predicate set and the object set in the new vocabulary data set V', respectively. When f is(s，p，o)Null, then { es}，{ep}，{eoThere is no relation between them, no relevant short sentence is generated.
In a further embodiment, the subject in the short sentence fits the predicate and/or object invoked as follows:
the current relationship is in accordance with the original relationship of the subject and the corresponding predicate and/or object when the subject is in the transcribed text.
In a further embodiment, the creation process of the recognition model is as follows:
classifying the vocabulary in each group of branches or sub-branches according to the subjects, predicates and objects, accumulating the occurrence times of the subjects, predicates and objects from zero, and adding 1 to the corresponding type if the occurrence times of the subjects, predicates and objects do not occur once;
when the number of times Ns of the subject is more than or equal to 2, segmenting the subject based on the subject, and dividing the subject into Ns segments, wherein the subject of each segment is different;
when the number of times Np of the predicates is larger than or equal to 2, segmenting based on the predicates, and dividing into Np sections, wherein the gastric language of each section is different;
when the number No of times of the object is more than or equal to 2, the object is segmented based on the object and is divided into No segments, and the object of each segment is different.
In a further embodiment, the search sequence is represented as [ subject, predicate, object, question type ];
wherein the question types at least include: the judgment is yes or no, one-out-of-multiple, multiple-out-of-multiple, and answer type.
In a further embodiment, the workflow of the matching model is at least as follows:
on the basis of traversal of the retrieval sequence from the sub-branches, respectively calculating the similarity between the subject and the predicate in the retrieval sequence and the subject and the predicate in the sub-branches, and if the similarity is within a threshold range, generating an answer in the object in the sub-branches on the basis of the question type;
otherwise, traversing from the branch based on the retrieval sequence, respectively calculating the similarity between the subject and the predicate in the retrieval sequence and the subject and the predicate in the branch, and if the similarity is in a threshold range, generating an answer in the object in the branch based on the question type;
and if the answer is not matched, generating feedback information.
The calculation formula of the similarity is as follows:
m is the number of subjects in the search sequence, wxmIs the weight of the subject numbered m, w1Is the weight of the subject in the sub-branch; n is the number of subjects in the search sequence, wxnIs the weight of the predicate numbered n, w2Is the weight of the predicate in the child branch.
Alpha is a question type, and the value of alpha is as follows
A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method as described above when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method as set forth above.
The invention has the following beneficial effects: the invention splits the text in the robot data by adopting a tree structure, splits the overlapped sentences into a plurality of identical short sentences with main and predicate objects according to the Chinese grammar, and is convenient for the robot to flexibly identify question sentences during answering, thereby searching out the best answer. The use feeling of the client is improved.
Drawings
FIG. 1 is a flow chart of the operation of the present invention.
Detailed Description
The invention is further described with reference to the drawings and examples in the following description.
At present, when a robot carries out conversation, the robot has poor language analysis capability, namely, the robot cannot identify the same semantic question method. Examples are: the robot first updates the data in the database: please remember that there are 3 families of information academy in schools, which are computer science and technology system, communication engineering system and electronic information engineering system, respectively, 200 people are recruited in computer science and technology system, 150 people are recruited in communication engineering system and 100 people are recruited in electronic information engineering system in this year. The user starts to ask questions: the beauty is the number of people recruited in the computer science and technology system of this year. The answer is: 200 persons. Asking questions: the computer science and technology of this year is attracting people. If so, how many people are recruited. The answer is: there is no answer you want.
The embodiment discloses an intelligent dialogue method based on deep learning of a robot, which comprises the following steps:
acquiring voice data to perform voice recognition to obtain a transcription text of the voice data;
constructing an information tree based on the transcribed text, and creating sentence break nodes according to the Chinese grammar; segmenting the transcribed text based on the sentence break nodes to obtain a plurality of groups of parallel short sentences, wherein the short sentences are used as branches and respectively extend on the information tree; in a further embodiment, the sentence break node is one or more of a subject, a predicate or an object; and is divided by subject, predicate and object according to the type of sentence-breaking node.
In other words, subjects belong to sentence break nodes of the same type, predicates belong to sentence break nodes of the same type, and objects belong to sentence break nodes of the same type.
That is, when two subjects still exist in the split phrase, that is, the predicate and the object are copied to generate two copies with the two subjects as bases, and the two copies correspond to the two subjects respectively, the split phrase is split into two phrases with the same meaning: the current transcribed text is: there are 3 families of information academy in schools, which are respectively computer science and technology system, communication engineering system and electronic information engineering system, wherein 200 people are introduced in the computer science and technology system, 150 people are introduced in the communication engineering system and 100 people are introduced in the electronic information engineering system in this year. Wherein, there are two subjects "school" and "information college" in "there are 3 lines of" information colleges of schools ". Therefore, the transcribed text is segmented based on the sentence break nodes (two subjects) to obtain a plurality of groups of parallel short sentences, and the sentence break nodes are divided into a 'college of information in schools' and a 'college of information' with 3 lines.
Creating an identification model, and judging each group of branches: if at least two sentence-breaking nodes of the same type still exist in the branch, continuously segmenting to obtain at least one group of sub-branches based on the sentence-breaking nodes of the same type until only one sentence-breaking node of the same type exists in any group of sub-branches; if there is one and only one punctuation node of the same type in the information tree, branch or sub-branch, the segmentation is stopped for the information tree, branch or sub-branch.
That is, when 3 lines of 'schools and information colleges' are identified, two sentence-breaking nodes (subjects) of the same type are found, one is 'school' and the other is 'information college'. Therefore, the sentences can be further split into 3 lines of information colleges of people and schools and 3 lines of information colleges. The two split short sentences are identified, and each short sentence has only one node of the same type, so that the split short sentences are not split, and the 'information college in schools' and the 'information college has 3 lines' are sub-branches obtained by splitting respectively.
Acquiring a questioning voice, identifying a subject in the questioning voice and a corresponding questioning type, and generating a retrieval sequence; and inputting the retrieval sequence into a matching model, traversing the sub-branches, the branches and the information tree by the matching model to find the best answer, and broadcasting.
In a further embodiment, segmenting the transcribed text based on the sentence break node and obtaining a plurality of groups of parallel short sentences specifically comprises the following steps:
counting all vocabularies in the transcribed text to generate a vocabulary data set V;
traversing the data set, extracting each subject in the data set to generate a subject data set S, and removing and updating the subject data set from the vocabulary data set to obtain a new vocabulary data set V';
using a mapping formula to invoke at least one predicate and/or object adapted to each subject from the new lexical data set V' resulting in a related short sentence f starting with the subject(s，p，o)And generating a set of phrases based on the subject.
The mapping formula is expressed as
f(s，p，o)＝Map({es}，{ep}，{eo}, Os, Op, Oo); wherein s represents a subject, p predicate, and o represents an object; given a subject, predicate and object body, Os, Op and Oo respectively, the mapping from the ontology Os to Op and Oo means that for each entity in the ontology Os, a corresponding entity is found in the ontology Op and Oo that conforms to the chinese grammar, where es∈Os，ep∈Op，eoE is equal to Oo, and,
{es}，{ep}，{eodenotes the elements in the subject data set S, the predicate set and the object set in the new vocabulary data set V', respectively. When f is(s，p，o)Null, then { es}，{ep}，{eoThere is no relation between them, no relevant short sentence is generated.
For example, the following steps are carried out: based on the transcribed text, the Xiaomei please remember that there are 3 families of the information academy of schools, which are the computer science and technology system, the communication engineering system and the electronic information engineering system, respectively, 200 people are recruited in the computer science and technology system, 150 people are recruited in the communication engineering system and 100 people are recruited in the electronic information engineering system in this year.
The following calculation is obtained through the mapping formula: when the subject is school, the predicate having mapping relation with the subject is 'present', the object is 'information college', 'computer science and technology system', 'communication engineering system' and 'electronic information engineering system'; the subject "electronic information engineering system", the predicate having a mapping relation with the subject "present", the object "information college", "computer science and technology system", "communication engineering system", "electronic information engineering system"; the subject "information college", the predicate having a mapping relationship with the subject "information college" is "recruit", the object "200", and similarly, the subject "communication engineering system", the predicate having a mapping relationship with the subject "communication engineering system" is "recruit", and the object "150"; the subject "electronic information engineering", the predicate having a mapping relation with the subject "recruit", and the object "100", therefore, under the action of the above mapping formula, the transcribed text is first converted into a phrase set based on the subject as a school { a person school has an information institute, a person school has a computer science and technology system, a person school has a communication engineering system, a person school has an electronic information engineering system }.
Secondly, obtaining a phrase set based on subject as informatics { the information college has a computer science and technology system, the information college has a communication engineering system, and the information college has a communication engineering system };
then respectively obtaining 200 persons of the computer science and technology system, 150 persons of the communication engineering system and 100 persons of the electronic information engineering system based on the information college, the computer science and technology system, the communication engineering system and the electronic information engineering system.
In a further embodiment, the subject in the short sentence fits the predicate and/or object invoked as follows:
the current relationship is in accordance with the original relationship of the subject and the corresponding predicate and/or object when the subject is in the transcribed text.
In a further embodiment, the creation process of the recognition model is as follows:
classifying the vocabulary in each group of branches or sub-branches according to the subjects, predicates and objects, accumulating the occurrence times of the subjects, predicates and objects from zero, and adding 1 to the corresponding type if the occurrence times of the subjects, predicates and objects do not occur once;
when the number of times Ns of the subject is more than or equal to 2, segmenting the subject based on the subject, and dividing the subject into Ns segments, wherein the subject of each segment is different;
when the number of times Np of the predicates is larger than or equal to 2, segmenting based on the predicates, and dividing into Np sections, wherein the gastric language of each section is different;
when the number No of times of the object is more than or equal to 2, the object is segmented based on the object and is divided into No segments, and the object of each segment is different.
In a further embodiment, the search sequence is represented as [ subject, predicate, object, question type ];
wherein the question types at least include: the judgment is yes or no, one-out-of-multiple, multiple-out-of-multiple, and answer type.
Asking questions whether the judgment is yes: the schools of the people have school information colleges. The answer is: there are. Asking questions: there is English college in the schools of Zan. The answer is: none.
Asking questions of one more choice: the number of recruits in the information college was 100 for which line. The answer is: electronic information engineering system.
Multi-choice question asking: which lines are in the information academy to find the person. The answer is: computer science and technology system, communication engineering system and electronic information engineering system.
Answer type question: which lines are in the information academy to find the person. Each recruits more or less people. The answer is: computer science and technology is recruited 200 people, communication engineering is recruited 150 people, electronic information engineering is recruited 100 people.
In order to implement the above technical solution, the work flow of the matching model is at least as follows:
on the basis of traversal of the retrieval sequence from the sub-branches, respectively calculating the similarity between the subject and the predicate in the retrieval sequence and the subject and the predicate in the sub-branches, and if the similarity is within a threshold range, generating an answer in the object in the sub-branches on the basis of the question type;
otherwise, traversing from the branch based on the retrieval sequence, respectively calculating the similarity between the subject and the predicate in the retrieval sequence and the subject and the predicate in the branch, and if the similarity is in a threshold range, generating an answer in the object in the branch based on the question type;
and if the answer is not matched, generating feedback information.
The calculation formula of the similarity is as follows:
m is the number of subjects in the search sequence, wxmIs the weight of the subject numbered m, w1Is the weight of the subject in the sub-branch; n is the number of subjects in the search sequence, wxnIs the weight of the predicate numbered n, w2Is the weight of the predicate in the child branch.
Alpha is a question type, and the value of alpha is as follows
A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method as described above when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method as set forth above.
The robot of the embodiment has the learning and scene deep conversation capability, and can deeply memorize and understand the meaning of a question and search answers of questions besides a language processing method of one question and one answer.
Claims (10)
1. An intelligent dialogue method based on deep learning of a robot is characterized by comprising the following steps:
acquiring voice data to perform voice recognition to obtain a transcription text of the voice data;
constructing an information tree based on the transcribed text, and creating sentence break nodes according to the Chinese grammar; segmenting the transcribed text based on the sentence break nodes to obtain a plurality of groups of parallel short sentences, wherein the short sentences are used as branches and respectively extend on the information tree;
creating an identification model, and judging each group of branches: if at least two sentence-breaking nodes of the same type still exist in the branch, continuously segmenting to obtain at least one group of sub-branches based on the sentence-breaking nodes of the same type until only one sentence-breaking node of the same type exists in any group of sub-branches;
acquiring a questioning voice, identifying a subject in the questioning voice and a corresponding questioning type, and generating a retrieval sequence; and inputting the retrieval sequence into a matching model, traversing the sub-branches, the branches and the information tree by the matching model to find the best answer, and broadcasting.
2. The intelligent dialogue method based on the deep robot learning of claim 1, wherein if there is only one and only one sentence-break node of the same type in the information tree, branch or sub-branch, the segmentation is stopped for the information tree, branch or sub-branch.
3. The intelligent dialogue method based on the robot deep learning of claim 1, wherein the sentence-break node is one or more of a subject, a predicate or an object;
the types of the sentence-breaking nodes are divided through subjects, predicates and objects.
4. The intelligent dialogue method based on the robot deep learning of claim 1, wherein segmenting the transcribed text based on the sentence break node and obtaining a plurality of groups of parallel short sentences specifically comprises the following steps:
counting all vocabularies in the transcribed text to generate a vocabulary data set;
traversing the data set, extracting each subject in the data set to generate a subject data set, and removing and updating the subject data set from the vocabulary data set to obtain a new vocabulary data set;
at least one predicate and/or object adapted to each subject is invoked from the new lexical data set using a mapping formula, resulting in related phrases starting with the subject, and generating a set of phrases based on the subject.
5. The intelligent dialogue method based on the deep robot learning of claim 4, wherein the subject in the short sentence is adapted to the called predicate and/or object as follows:
the current relationship is in accordance with the original relationship of the subject and the corresponding predicate and/or object when the subject is in the transcribed text.
6. The intelligent dialogue method based on the deep robot learning of claim 1, wherein the identification model is created as follows:
classifying the vocabulary in each group of branches or sub-branches according to the subjects, predicates and objects, accumulating the occurrence times of the subjects, predicates and objects from zero, and adding 1 to the corresponding type if the occurrence times of the subjects, predicates and objects do not occur once;
when the number of times Ns of the subject is more than or equal to 2, segmenting the subject based on the subject, and dividing the subject into Ns segments, wherein the subject of each segment is different;
when the number of times Np of the predicates is larger than or equal to 2, segmenting based on the predicates, and dividing into Np sections, wherein the gastric language of each section is different;
when the number No of times of the object is more than or equal to 2, the object is segmented based on the object and is divided into No segments, and the object of each segment is different.
7. The intelligent dialogue method based on robot deep learning of claim 1, wherein the search sequence is expressed as [ subject, predicate, object, question type ];
wherein the question types at least include: the judgment is yes or no, one-out-of-multiple, multiple-out-of-multiple, and answer type.
8. The intelligent dialogue method based on deep robot learning of claim 7, wherein the workflow of the matching model is at least as follows:
on the basis of traversal of the retrieval sequence from the sub-branches, respectively calculating the similarity between the subject and the predicate in the retrieval sequence and the subject and the predicate in the sub-branches, and if the similarity is within a threshold range, generating an answer in the object in the sub-branches on the basis of the question type;
otherwise, traversing from the branch based on the retrieval sequence, respectively calculating the similarity between the subject and the predicate in the retrieval sequence and the subject and the predicate in the branch, and if the similarity is in a threshold range, generating an answer in the object in the branch based on the question type;
and if the answer is not matched, generating feedback information.
9. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1 to 7 are implemented by the processor when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111357137.9A CN114281962A (en) | 2021-11-16 | 2021-11-16 | Intelligent dialogue system, method and storage medium based on robot deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111357137.9A CN114281962A (en) | 2021-11-16 | 2021-11-16 | Intelligent dialogue system, method and storage medium based on robot deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114281962A true CN114281962A (en) | 2022-04-05 |
Family
ID=80869493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111357137.9A Withdrawn CN114281962A (en) | 2021-11-16 | 2021-11-16 | Intelligent dialogue system, method and storage medium based on robot deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114281962A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115186092A (en) * | 2022-07-11 | 2022-10-14 | 贝壳找房(北京)科技有限公司 | Online interaction processing method and apparatus, storage medium, and program product |
-
2021
- 2021-11-16 CN CN202111357137.9A patent/CN114281962A/en not_active Withdrawn
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115186092A (en) * | 2022-07-11 | 2022-10-14 | 贝壳找房(北京)科技有限公司 | Online interaction processing method and apparatus, storage medium, and program product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107451126B (en) | Method and system for screening similar meaning words | |
| CN107273355B (en) | Chinese word vector generation method based on word and phrase joint training | |
| CN110096567B (en) | QA knowledge base reasoning-based multi-round dialogue reply selection method and system | |
| CN106570708B (en) | Management method and system of intelligent customer service knowledge base | |
| CN109271537B (en) | Text-to-image generation method and system based on distillation learning | |
| CN108681574B (en) | Text abstract-based non-fact question-answer selection method and system | |
| WO2019015269A1 (en) | Korean named entities recognition method based on maximum entropy model and neural network model | |
| CN113094578B (en) | Deep learning-based content recommendation method, device, equipment and storage medium | |
| CN108733647B (en) | Word vector generation method based on Gaussian distribution | |
| CN108038106B (en) | Fine-grained domain term self-learning method based on context semantics | |
| CN113672718B (en) | Dialogue intention recognition method and system based on feature matching and field self-adaption | |
| Harwath et al. | Zero resource spoken audio corpus analysis | |
| CN112052318A (en) | Semantic recognition method and device, computer equipment and storage medium | |
| CN111625633A (en) | Knowledge graph-based enterprise system question-answer intention identification method and device | |
| CN115858750A (en) | Power grid technical standard intelligent question-answering method and system based on natural language processing | |
| CN112100212A (en) | Case scenario extraction method based on machine learning and rule matching | |
| CN110795544A (en) | Content search method, device, equipment and storage medium | |
| CN107562907B (en) | Intelligent lawyer expert case response device | |
| CN114281962A (en) | Intelligent dialogue system, method and storage medium based on robot deep learning | |
| Mandge et al. | Revolutionize cosine answer matching technique for question answering system | |
| Narimatsu et al. | Duration and interval hidden Markov model for sequential data analysis | |
| CN117454898A (en) | Method and device for realizing legal entity standardized output according to input text | |
| CN115905187B (en) | Intelligent proposition system oriented to cloud computing engineering technician authentication | |
| CN114372454A (en) | Text information extraction method, model training method, device and storage medium | |
| CN107609096B (en) | Intelligent lawyer expert response method |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220405 |