CN108228637B - Automatic response method and system for natural language client - Google Patents
Automatic response method and system for natural language client Download PDFInfo
- Publication number
- CN108228637B CN108228637B CN201611185779.4A CN201611185779A CN108228637B CN 108228637 B CN108228637 B CN 108228637B CN 201611185779 A CN201611185779 A CN 201611185779A CN 108228637 B CN108228637 B CN 108228637B
- Authority
- CN
- China
- Prior art keywords
- answer
- module
- candidate
- user
- subject
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3334—Selection or weighting of terms from queries, including natural language queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3338—Query expansion
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method and a system for automatically responding to a natural language client. The method comprises the following steps: inquiring a database to obtain an answer of an original question of a user, wherein the answer is an example meeting conditions in the database, and all attributes of the example are used as candidate slot positions; judging whether the number of answers is larger than a preset value or not; if the number of the answers is larger than a preset value, a question is asked to the user; if the number of the answers is not more than the preset value, judging whether the answers are inquired or not; and if the answer is inquired and the number of the answers is not more than the preset value, returning the answer to the user. The invention takes the user requirement as the starting point of the service, the user only needs to describe the problem, and the user does not need to know the service content in advance; the information required by the user can be acquired directly according to the natural language description of the actual requirement of the user.
Description
Technical Field
The invention relates to the field of natural language processing, in particular to a method and a system for automatically responding to a natural language client.
Background
The customer service system is a system with coordinated business processes, strategies and technologies, provides a proper channel for obtaining information for users, and creates customer value and enterprise value in an interactive conversation mode.
The manual agent response and the automatic response are two most important customer question and answer systems at present.
While the human agent response can provide a better user service experience, it still has the following drawbacks:
1. the labor cost is high, and a large amount of personnel is needed to reduce the waiting service time of the user;
2. the repeated work is more, and a large amount of simple and repeated labor exists when general service contents are processed;
3. the professional requirement is high, and a lot of time is needed to improve the service proficiency when complex service flow and business information are processed.
The automatic answering system can greatly make up the deficiency of manual seat answering, and has the following advantages:
1. the labor cost is saved, and the function of manual seat response can be partially replaced;
2. the response speed is high, and the problem with clear content can be responded efficiently;
3. the service is uninterrupted, and the automatic response can be carried out for 7x24 hours.
However, most of the existing automatic response systems are simple in logic and single in function, and ignore context information of user history problems; when the user intention is complex or not clear enough, the user often answers questions or refuses to answer, and the user experience is poor; meanwhile, the conventional automatic response system has poor capability of understanding user requirements, and a user can correctly select and receive services only after needing to know service items, so that correct service contents cannot be provided according to the user requirements. Meanwhile, more application scenarios of the existing chat robots such as siri and microsoft mini-ice are chatting instead of inquiring accurate knowledge, and the chat robots belong to different fields from the professional field question-answering system.
The existing automatic question-answering systems are mainly divided into three types, namely, automatic answer based on key instruction, automatic answer based on keyword extraction and automatic answer based on general natural language processing. The following disadvantages exist in the three systems respectively:
automatic answering based on key commands:
1. passive service: the user must know all service items and then can select the correct service type;
2. poor flexibility: the interaction logic is fixed, the ability to understand the user requirements is poor, and the user experience is poor.
Automatic response based on keyword extraction:
1. the intelligence is poor: the system does not have natural language processing and understanding capacity, cannot understand user requirements in a fuzzy mode, does not have the capacity of understanding user requests, and users must provide correct keywords to develop services.
2. Poor expansibility: the rules for extracting the keywords are manually set, do not have the capability of rapid automatic expansion, and need to be continuously updated manually.
Automatic response based on generic natural language processing:
1. the ability to handle complex service logic is not available: only simple response function can be performed, and the processing capability in a specific application field is insufficient.
Disclosure of Invention
In view of the above technical problems, the present invention provides an automatic response method and system for a natural language client, which can start a service process without requiring a user to know a service item in advance.
According to an aspect of the present invention, there is provided a natural language client automatic response method, including:
inquiring a database to obtain an answer of an original question of a user, wherein the answer is an example meeting conditions in the database, and all attributes of the example are used as candidate slot positions;
judging whether the number of answers is larger than a preset value or not;
if the number of the answers is not more than the preset value, judging whether the answers are inquired or not;
and if the answer is inquired and the number of the answers is not more than the preset value, returning the answer to the user.
In one embodiment of the invention, the method further comprises:
if the answer is not inquired, judging whether the subject matching fails;
if the subject matching fails, fuzzy matching is carried out, and a possible subject list is returned;
if the subject matching fails, judging whether the attribute matching of the query fails;
if the attribute matching of the query fails, returning an attribute value list corresponding to the subject instance;
if the attribute matching of the query does not fail, the subject is expanded and the restriction condition is relaxed, and then the query is performed.
In an embodiment of the present invention, the performing fuzzy matching and returning a list of possible subjects includes:
analyzing the original problem of the user by using a natural language processing method, and extracting a subject, a predicate and a fixed language limiting condition of the original problem of the user;
if the precise matching of the subject language fails, recalling the examples meeting the conditions according to the predicate and the predicate limiting conditions as a candidate set;
if the predicate exact matching fails, recalling the attribute corresponding to the subject as a candidate set according to the subject and the subject limiting condition;
in the candidate instance set, scoring each match using a fuzzy matching method;
and selecting the candidate set with the score larger than the preset threshold value to return to the user, and carrying out next confirmation by the user.
In an embodiment of the present invention, the query after expanding the subject and relaxing the constraint condition includes:
if the subject is an example in the database, all classes to which the example belongs are inquired upwards recursively, whether a certain class can meet the inquiry condition exists or not, and the inquiry is stopped when an answer is obtained;
if the subject is a class in the database, all father classes of the class are inquired upwards and recursively, whether a certain father class meets the inquiry condition or not is judged, and the inquiry is stopped when an answer is inquired;
if too many subject limiting conditions result in no answer to be inquired, then each limiting condition is tried to be removed in sequence, the number of answers is recorded, and the answer with the minimum number is returned as the answer.
In one embodiment of the invention, the method further comprises:
judging whether an answer is obtained after re-inquiry;
if the answer is obtained after the inquiry, the answer is returned to the user;
if the answer can not be obtained after the inquiry, the original question of the user is not answered.
In one embodiment of the invention, the method further comprises: and if the number of the answers is larger than the preset value, a question is asked to the user.
In one embodiment of the invention, said asking the user a question comprises:
judging whether a candidate slot position exists;
if no candidate slot position exists, the answer is truncated;
if the candidate slot position exists, scoring the candidate slot position;
judging whether the candidate slot position with the highest score is directly asked in a reverse manner;
if the candidate slot position with the highest score is directly asked in a reverse way, the possible value corresponding to the candidate slot position is asked in a reverse way;
if the candidate slot position with the highest score is not directly asked, the sorted candidate slot position list is returned, and the user selects the interested candidate slot position.
In one embodiment of the present invention, the scoring the candidate slot includes:
acquiring an answer set meeting the original problem of a user and an attribute set of a candidate slot position;
calculating entropy values of each slot position after grouping according to the slot position;
obtaining the interestingness score of each slot position of a user;
performing weighted summation on the interestingness score and the entropy value;
and updating the answer set into a new answer set after the user interaction.
In an embodiment of the present invention, after the answer is returned to the user, the method further includes:
recalling all attributes of the instance or class corresponding to the answer;
counting the frequency of each candidate attribute;
obtaining the correlation degree of each candidate attribute and the original problem of the user;
sorting the candidate attributes according to the frequency and the correlation;
and generating a recommended candidate question according to the sorted candidate attributes.
According to another aspect of the present invention, there is provided a natural language client automatic response system, comprising a query module, an answer number judgment module, a query result judgment module, and an answer return module, wherein:
the query module is used for querying a database to obtain an answer of an original question of a user, wherein the answer is an example which meets the conditions in the database, and all attributes of the example are used as candidate slot positions;
the answer quantity judging module is used for judging whether the answer quantity is greater than a preset value or not;
the query result judging module is used for judging whether the answer is queried or not according to the judgment result of the answer quantity judging module under the condition that the answer quantity is not greater than the preset value;
and the answer returning module is used for returning the answer to the user under the condition of inquiring the answer according to the judgment result of the inquiry result judgment module.
In an embodiment of the present invention, the system further includes a subject matching judgment module, a fuzzy matching module, an attribute matching judgment module, an attribute list returning module, and a subject extension module, wherein:
the subject matching judgment module is used for judging whether subject matching fails or not under the condition that an answer is not inquired according to the judgment result of the inquiry result judgment module;
the fuzzy matching module is used for performing fuzzy matching according to the judgment result of the subject matching judgment module under the condition that the subject matching fails and returning a possible subject list;
the attribute matching judgment module is used for judging whether the attribute matching of the query fails or not under the condition that the subject matching fails according to the judgment result of the subject matching judgment module;
the attribute list returning module is used for returning an attribute value list corresponding to the subject instance under the condition that the inquired attribute matching fails according to the judgment result of the attribute matching judgment module;
and the subject extension module is used for extending the subject and relaxing the limiting conditions and then inquiring according to the judgment result of the attribute matching judgment module under the condition that the inquired attribute matching is not failed.
In one embodiment of the present invention, the fuzzy matching module includes a condition extraction unit, a candidate set determination unit, a fuzzy scoring unit, and a candidate set return unit, wherein:
the condition extraction unit is used for analyzing the original user problem by using a natural language processing method and extracting a subject, a predicate and a fixed language limiting condition of the original user problem;
the candidate set determining unit is used for recalling the examples meeting the conditions according to the predicates and the fixed-phrase limiting conditions as the candidate set under the condition that the precise matching of the subjects fails; under the condition that the accurate matching of the predicates fails, recalling the attributes corresponding to the subject as a candidate set according to the subject and the subject limiting conditions;
the fuzzy scoring unit is used for scoring each match in the candidate example set by using a fuzzy matching method;
and the candidate set returning unit is used for selecting the candidate set with the score larger than the preset threshold value to return to the user and the user confirms next step.
In one embodiment of the present invention, the subject extension module includes a first query unit, a second query unit, and a condition releasing unit, wherein:
the first query unit is used for querying all classes to which the instance belongs upwards and recursively under the condition that the subject is the instance in the database, judging whether a certain class can meet query conditions, and stopping querying the answer;
the second query unit is used for inquiring all father classes of the class upwards and recursively under the condition that the subject is the class in the database, judging whether a certain father class meets the query condition, and stopping when the answer is queried;
and the condition removing unit is used for sequentially trying to remove each limiting condition under the condition that the query cannot be answered due to excessive subject limiting conditions, recording the number of answers, and returning the answer with the minimum number as the answer.
In an embodiment of the present invention, the system further includes a re-query result determination module, wherein:
the re-query result judging module is used for judging whether an answer is obtained after re-query; under the condition that the answer is obtained after re-inquiry, indicating an answer returning module to return the answer to the user; and not answering the original question of the user under the condition that the answer can not be obtained after the re-inquiry.
In one embodiment of the invention, the system further comprises a question-back module, wherein:
and the question-returning module is used for providing a question-returning to the user under the condition that the number of the answers is greater than the preset value according to the judgment result of the answer number judgment module.
In an embodiment of the present invention, the question-backing module includes a candidate slot position judging unit, an answer truncating unit, a slot position scoring unit, and a question-backing judging unit, wherein:
the candidate slot position judging unit is used for judging whether a candidate slot position exists or not under the condition that the answer quantity is greater than a preset value according to the judgment result of the answer quantity judging module;
the answer truncation unit is used for truncating the answer under the condition that no candidate slot position exists according to the judgment result of the candidate slot position judgment unit;
the slot position scoring unit is used for scoring the candidate slot position under the condition that the candidate slot position exists according to the judgment result of the candidate slot position judging unit;
the back-questioning judging unit is used for judging whether to directly ask the candidate slot position with the highest back-questioning score after the slot position scoring unit scores the candidate slot positions; under the condition of directly asking the candidate slot position with the highest score, asking the possible value corresponding to the candidate slot position; and under the condition that the candidate slot position with the highest score is not directly asked, returning the sorted candidate slot position list, and selecting the interested candidate slot position by the user.
In an embodiment of the present invention, the slot scoring unit includes an answer set obtaining sub-module, an entropy determining sub-module, an interestingness determining sub-module, a weighted sum sub-module, and an answer updating sub-module, wherein:
the answer set acquisition submodule is used for acquiring an answer set meeting the original problem of the user and an attribute set of the candidate slot position;
the entropy value determining submodule is used for calculating the entropy value grouped according to the slot position for each slot position;
the interest level determining submodule is used for obtaining the interest level score of each slot position of the user;
the weighted summation submodule is used for carrying out weighted summation on the interestingness score and the entropy value;
and the answer updating submodule is used for updating the answer set into a new answer set after the user interacts.
In an embodiment of the present invention, the system further includes an attribute recall module, a frequency statistics module, a relevancy determination module, a ranking module, and a recommendation problem generation module, wherein:
the attribute recalling module is used for recalling all attributes of the instance or the class corresponding to the answer after the answer returning module returns the answer to the user;
the frequency counting module is used for counting the frequency of each candidate attribute;
the relevancy determining module is used for acquiring the relevancy of each candidate attribute and the original problem of the user;
the sorting module is used for sorting the candidate attributes according to the frequency and the relevancy;
and the recommendation problem generation module is used for generating recommended candidate problems according to the candidate attributes sorted by the sorting module.
The invention takes the user requirement as the starting point of the service, the user only needs to describe the problem, and the user does not need to know the service content in advance; the information required by the user can be acquired directly according to the natural language description of the actual requirement of the user.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic diagram of a first embodiment of an automatic response method for a natural language client according to the present invention.
Fig. 2 is a diagram illustrating a second embodiment of the method for automatically responding to a natural language client according to the present invention.
FIG. 3 is a diagram illustrating a dynamic slot selection method based on dynamic entropy according to an embodiment of the present invention.
FIG. 4 is a flowchart illustrating an intelligent interactive function based on fuzzy matching according to an embodiment of the present invention.
Fig. 5 is a flowchart illustrating an intelligent answer retrieval based on an ontology structure according to an embodiment of the present invention.
Fig. 6 is a diagram illustrating a third embodiment of the method for automatically responding to a natural language client according to the present invention.
Fig. 7 is a diagram illustrating a first embodiment of the automatic response system for natural language clients according to the present invention.
Fig. 8 is a diagram illustrating a second embodiment of the automatic response system for natural language clients according to the present invention.
FIG. 9 is a diagram of a fuzzy matching module in accordance with an embodiment of the present invention.
FIG. 10 is a diagram of a subject expansion module according to an embodiment of the invention.
FIG. 11 is a diagram of a question-back module in accordance with an embodiment of the present invention.
FIG. 12 is a schematic diagram of a slot scoring element in accordance with an embodiment of the present invention.
Fig. 13 is a diagram illustrating a third embodiment of the automatic response system for natural language clients according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The relative arrangement of the components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.
Meanwhile, it should be understood that the sizes of the respective portions shown in the drawings are not drawn in an actual proportional relationship for the convenience of description.
Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate.
In all examples shown and discussed herein, any particular value should be construed as merely illustrative, and not limiting. Thus, other examples of the exemplary embodiments may have different values.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.
Fig. 1 is a schematic diagram of a first embodiment of an automatic response method for a natural language client according to the present invention. Preferably, this embodiment can be performed by the natural language client automatic response system of the present invention. The method comprises the following steps:
In one embodiment of the present invention, the database is an ontology repository, i.e. a service business repository.
And step 104, if the answer is inquired and the number of the answers is not more than the preset value, returning the answer to the user.
Based on the automatic response method for the natural language client provided by the embodiment of the invention, the user only needs to describe the problem by taking the user requirement as the starting point of the service, and the user does not need to know the service content in advance; the information required by the user can be acquired directly according to the natural language description of the actual requirement of the user; the service content which best meets the design target of the service logic can be obtained by intelligently searching and reasoning in the service knowledge base.
Fig. 2 is a diagram illustrating a second embodiment of the method for automatically responding to a natural language client according to the present invention. Preferably, this embodiment can be performed by the natural language client automatic response system of the present invention. The method comprises the following steps:
step 201, receiving a user original question input by a user.
In step 203, it is determined whether the number of answers is greater than a predetermined value (i.e., whether there are too many answers). If the number of answers is greater than the predetermined value (i.e. too many answers), go to step 204; otherwise, if the answer number is not greater than the predetermined value, step 210 is executed.
In step 204, it is determined whether there is a candidate slot (optional slot). If there is no candidate slot, go to step 205; otherwise, if there is a candidate slot, step 206 is executed.
And step 206, scoring the candidate slot position according to the answer initialization result. And scoring each slot position based on a dynamic slot position evaluation algorithm to obtain the priority of each slot position.
In an embodiment of the present invention, after step 208 or step 209, the method may further include: and generating question-back sentences and selectable items, namely generating natural language sentences based on natural language generation technology according to question-back logic and presenting the natural language sentences to the user. Thereafter, the other steps of the present embodiment are not performed.
Step 213, fuzzy matching is carried out, and a possible subject list is returned; after which no further steps of the present implementation are performed.
Step 215, returning an attribute value list corresponding to the subject instance; after which no further steps of the present implementation are performed.
Steps 204-209 in the above embodiments of the present invention are detailed steps of the step of asking the user a question.
The embodiment of the invention takes the user requirement as the starting point of the service, and the user only needs to describe the problem and does not need to know the service content in advance.
The embodiment of the invention can directly obtain the information required by the user according to the natural language description of the actual requirement of the user.
According to the embodiment of the invention, questions can be actively asked to the user, and a proper interaction sequence is obtained based on calculation, so that the interaction efficiency is improved.
The embodiment of the invention can intelligently search and reason in the service business knowledge base to obtain the service content which best meets the business logic design target.
The embodiment of the invention has the capability of processing complex business logic and the capability of pertinently meeting the exact requirements of users.
The embodiment of the invention can make clear user query intention step by step in a hierarchical manner through conversation interaction, thereby realizing complex service logic of the service.
The embodiment of the invention can help the user to select the service meeting the user requirement according to the user requirement and the service logic.
In an embodiment of the present invention, in the fig. 1 or fig. 2 embodiment, the method may further include: and saving and restoring conversation history information.
The above embodiments of the present invention may be applied to a dialogue interaction mechanism of a natural language client question-and-answer system.
The following further illustrates some of the key steps of the embodiment of fig. 2 of the present invention by way of specific examples.
FIG. 3 is a diagram illustrating a dynamic slot selection method based on dynamic entropy according to an embodiment of the present invention. As shown in fig. 3, step 206 in the embodiment of fig. 2, namely, the step of scoring the candidate slot may include:
The final score, score (slot) ═ w1×entr+w2× pop, known as w from empirical values1Preferably 0.4, w2Preferably 0.6.
Step 2065, the answer set is updated to a new answer set after the user interaction, and then step 2061 is executed. Until the number of answers meets the requirement, the steps 2061 to 2065 are calculated iteratively.
The embodiment of the invention realizes the dynamic slot position evaluation based on the information entropy, thereby enabling a user to obtain a final answer more quickly and reducing the interaction times.
FIG. 4 is a flowchart illustrating an intelligent interactive function based on fuzzy matching according to an embodiment of the present invention. As shown in fig. 4, step 213 in the embodiment of fig. 2, that is, the step of performing fuzzy matching, returning to the possible subject list, may include:
step 2131, analyzing the user original problem by using a natural language processing method, and extracting a subject, a predicate and a predicate restriction condition of the user original problem.
And step 2132, judging whether the precise matching of the subject language fails. If the exact match of the subject fails, go to step 2133; if the exact match of the subject does not fail, then step 2134 is performed.
And step 2133, recalling the examples meeting the conditions according to the predicates and the predicate limiting conditions as a candidate set.
And a step 2134 of judging whether the precise matching of the subject language fails. If the exact match of the subject fails, go to step 2135; if the exact match of the subject does not fail, then step 2136 is performed.
And step 2135, recalling the attributes corresponding to the subject as a candidate set according to the subject and the language-fixing limiting conditions.
In an embodiment of the present invention, after step 2133 and step 2135, the method may further include: scoring each match in the set of candidate instances using a fuzzy matching method (edit distance, longest common subsequence); and selecting the candidate set with the score larger than the preset threshold value to return to the user, and carrying out next confirmation by the user.
Since the problem of user input often has some ambiguity, it is not always exactly matched to an instance or class in the ontology repository. Therefore, fuzzy matching methods are used in the entity linking stage, namely fuzzy matching based on instance names and fuzzy matching based on attributes, and an interaction mechanism based on fuzzy matching is provided.
Fig. 5 is a flowchart illustrating an intelligent answer retrieval based on an ontology structure according to an embodiment of the present invention. As shown in fig. 5, step 216 in the embodiment of fig. 2, that is, the step of expanding the subject and relaxing the constraint and then querying, may include:
in step 2161, when the answer is not found, it is determined whether the subject is an instance in the database. If the subject is an instance in the database, go to step 2162; otherwise, if the subject is not an instance in the database, then step 2163 is performed.
In step 2165, it is determined whether an answer is obtained. If the answer is obtained, returning the answer; otherwise, if no answer is obtained, then go to step 2166.
Fig. 6 is a diagram illustrating a third embodiment of the method for automatically responding to a natural language client according to the present invention. Preferably, this embodiment can be performed by the natural language client automatic response system of the present invention. In the embodiment of fig. 1 or fig. 2, after the query returns an answer successfully, the method further includes the following steps:
In one embodiment of the present invention, step 603 may include: vectorizing the attribute, and calculating the cosine similarity of the attribute vector and an original attribute vector as the correlation information sim of the attribute and the original problem;
and step 604, sorting the candidate attributes according to the frequency and the relevancy.
In one embodiment of the present invention, step 604 may comprise: obtaining a final score (p1, pori) ═ w according to the frequency and the correlation1×cp1+w2× sim, where w is known from empirical values1Preferably 0.4, w2Preferably 0.6; sorting in descending order according to the final score.
In one embodiment of the present invention, step 605 may comprise: and according to the candidate attributes after descending sorting, taking the top five after descending sorting as recommended related problems.
When the user explicitly queries the specific attribute of a certain instance or class in the ontology, the embodiment of the invention indicates that the user has a query intention for the instance or class, and the system can intelligently take other attributes of the current instance or class as related questions which may be concerned by the user and return the related questions to the user while answering the user questions.
The embodiment of the invention can actively recommend other problems related to the query to the user and help the user to obtain more information.
Fig. 7 is a diagram illustrating a first embodiment of the automatic response system for natural language clients according to the present invention. As shown in fig. 7, the natural language client automatic response system may include a query module 701, an answer number judgment module 702, a query result judgment module 703 and an answer return module 704, wherein:
the query module 701 is configured to query a database to obtain an answer to an original question of a user, where the answer is an instance satisfying a condition in the database, and all attributes of the instance are used as candidate slots.
An answer quantity determining module 702 is configured to determine whether the answer quantity is greater than a predetermined value.
The query result determining module 703 is configured to determine whether an answer is queried according to the determination result of the answer number determining module 702, if the answer number is not greater than the predetermined value.
An answer returning module 704, configured to return an answer to the user according to the determination result of the query result determining module 703 when the answer is queried.
Based on the automatic response system of the natural language client provided by the embodiment of the invention, the user only needs to describe the problem by taking the user requirement as the starting point of the service, and the user does not need to know the service content in advance; the information required by the user can be acquired directly according to the natural language description of the actual requirement of the user; the service content which best meets the design target of the service logic can be obtained by intelligently searching and reasoning in the service knowledge base.
Fig. 8 is a diagram illustrating a second embodiment of the automatic response system for natural language clients according to the present invention. Compared with the embodiment shown in fig. 7, in the embodiment shown in fig. 8, the system may further include a subject matching judgment module 705, a fuzzy matching module 706, an attribute matching judgment module 707, an attribute list returning module 708, and a subject expansion module 709, wherein:
the subject matching determining module 705 is configured to determine whether the subject matching fails according to the determination result of the query result determining module 703 under the condition that no answer is queried.
And the fuzzy matching module 706 is configured to perform fuzzy matching according to the judgment result of the subject matching judgment module 705 and in the case that the subject matching fails, and return a possible subject list.
An attribute matching determining module 707, configured to determine, according to a determination result of the subject matching determining module 705, whether the subject matching fails or not when the subject matching fails.
An attribute list returning module 708, configured to return an attribute value list corresponding to the subject instance according to the determination result of the attribute matching determining module 707, when the attribute matching of the query fails.
A subject extension module 709, configured to, according to the determination result of the attribute matching determination module 707, extend the subject and relax the restriction condition before querying when the queried attribute matching fails.
FIG. 9 is a diagram of a fuzzy matching module in accordance with an embodiment of the present invention. As shown in fig. 9, the fuzzy matching module 706 in the embodiment of fig. 8 may include a condition extracting unit 7061, a candidate set determining unit 7062, a fuzzy scoring unit 7063, and a candidate set returning unit 7064, where:
a condition extracting unit 7061, configured to analyze the user original problem by using a natural language processing method, and extract a subject, a predicate, and a predicate restriction condition of the user original problem.
A candidate set determining unit 7062, configured to, when the subject exact match fails, recall, as a candidate set, an instance that meets the condition according to the predicate and the predicate restriction condition; and recalling the attribute corresponding to the subject as a candidate set according to the subject and the subject limiting condition under the condition that the accurate matching of the predicates fails.
A fuzzy scoring unit 7063 for scoring each match in the set of candidate instances using a fuzzy matching method.
And a candidate set returning unit 7064, configured to select a candidate set with a score greater than a predetermined threshold to return to the user, and the user performs the next confirmation.
FIG. 10 is a diagram of a subject expansion module according to an embodiment of the invention. As shown in fig. 10, the subject extension module 709 in the embodiment of fig. 8 may include a first query unit 7091, a second query unit 7092, and a condition releasing unit 7093, where:
a first querying unit 7091, configured to, when the subject is an instance in the database, query all classes to which the instance belongs recursively upward, whether any one of the classes can satisfy the query condition, and stop when an answer is queried.
A second querying unit 7092, configured to, in a case that the subject is a class in the database, query all parent classes of the class upwards and recursively, whether there is a certain parent class that satisfies the query condition, and stop when an answer is queried.
A condition removing unit 7093, configured to, when too many subject constraints result in no answer being queried, attempt to remove each constraint in sequence, record the number of answers, and return the answer with the smallest number of answers as the answer.
In an embodiment of the present invention, as shown in fig. 8, the system may further include a re-query result determining module 710, where:
a re-query result judgment module 710, configured to judge whether to obtain an answer after re-query by the subject extension module 709; in the case of obtaining an answer after re-query, the answer returning module 704 is instructed to return an answer to the user; and not answering the original question of the user under the condition that the answer can not be obtained after the re-inquiry.
In one embodiment of the present invention, as shown in fig. 8, the system may further include a question module 711, wherein:
the question-backing module 711 is configured to, according to the judgment result of the answer number judgment module, provide a question-backing to the user when the answer number is greater than the predetermined value.
FIG. 11 is a diagram of a question-back module in accordance with an embodiment of the present invention. As shown in fig. 11, the question-backing module 711 in the embodiment of fig. 8 may include a candidate slot determination unit 7111, an answer truncation unit 7112, a slot scoring unit 7113, and a question-backing determination unit 7114, wherein:
a candidate slot position determining unit 7111, configured to determine whether there is a candidate slot position according to the determination result of the answer number determining module 702, if the answer number is greater than the predetermined value.
An answer truncation unit 7112, configured to truncate the answer according to the determination result of the candidate slot determining unit 7111, if there is no candidate slot.
And a slot scoring unit 7113 configured to score the candidate slot if the candidate slot exists according to the determination result of the candidate slot determining unit 7111.
A question-back judging unit 7114, configured to judge whether to directly question back the candidate slot with the highest score after the slot scoring unit 7113 scores the candidate slot; under the condition of directly asking the candidate slot position with the highest score, asking the possible value corresponding to the candidate slot position; and under the condition that the candidate slot position with the highest score is not directly asked, returning the sorted candidate slot position list, and selecting the interested candidate slot position by the user.
FIG. 12 is a schematic diagram of a slot scoring element in accordance with an embodiment of the present invention. As shown in fig. 12, the slot scoring unit 7113 in the embodiment of fig. 11 may include an answer set obtaining sub-module 71131, an entropy value determining sub-module 71132, an interestingness determining sub-module 71133, a weighted sum sub-module 71134, and an answer updating sub-module 71135, wherein:
the answer set obtaining sub-module 71131 is used to obtain the answer set and the attribute set of the candidate slot that satisfy the original question of the user.
An entropy determination submodule 71132, configured to calculate, for each slot, an entropy value grouped according to the slot.
And the interestingness determining submodule 71133 is used for obtaining the interestingness score of each slot for the user.
A weighted sum sub-module 71134 for weighted sum of the interestingness score and the entropy value.
And an answer updating submodule 71135, configured to update the answer set to a new answer set after the user interaction.
The embodiment of the invention takes the user requirement as the starting point of the service, and the user only needs to describe the problem and does not need to know the service content in advance.
The embodiment of the invention can directly obtain the information required by the user according to the natural language description of the actual requirement of the user.
According to the embodiment of the invention, questions can be actively asked to the user, and a proper interaction sequence is obtained based on calculation, so that the interaction efficiency is improved.
The embodiment of the invention can intelligently search and reason in the service business knowledge base to obtain the service content which best meets the business logic design target.
The embodiment of the invention has the capability of processing complex business logic and the capability of pertinently meeting the exact requirements of users.
The embodiment of the invention can make clear user query intention step by step in a hierarchical manner through conversation interaction, thereby realizing complex service logic of the service.
The embodiment of the invention can help the user to select the service meeting the user requirement according to the user requirement and the service logic.
Fig. 13 is a diagram illustrating a third embodiment of the automatic response system for natural language clients according to the present invention. Compared with the embodiment shown in fig. 8, in the embodiment shown in fig. 13, the system may further include an attribute recall module 715, a frequency statistics module 716, a relevancy determination module 717, an ordering module 718, and a recommendation question generation module 719, wherein:
and the attribute recalling module 715 is used for recalling all the attributes of the instance or the class corresponding to the answer after the answer returning module 704 returns the answer to the user.
And a frequency statistics module 716, configured to count a frequency of each candidate attribute.
And a relevance determining module 717, configured to obtain a relevance of each candidate attribute to the original question of the user.
A ranking module 718, configured to rank the candidate attributes according to the frequency and the relevancy.
And a recommendation question generating module 719, configured to generate a recommended candidate question according to the candidate attribute ranked by the ranking module 718.
When the user explicitly queries the specific attribute of a certain instance or class in the ontology, the embodiment of the invention indicates that the user has a query intention for the instance or class, and the system can intelligently take other attributes of the current instance or class as related questions which may be concerned by the user and return the related questions to the user while answering the user questions.
The embodiment of the invention can actively recommend other problems related to the query to the user and help the user to obtain more information.
The natural language client automatic response system described above may be implemented as a general purpose processor, a Programmable Logic Controller (PLC), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any suitable combination thereof, for performing the functions described herein.
Thus far, the present invention has been described in detail. Some details well known in the art have not been described in order to avoid obscuring the concepts of the present invention. It will be fully apparent to those skilled in the art from the foregoing description how to practice the presently disclosed embodiments.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The description of the present invention has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to practitioners skilled in this art. The embodiment was chosen and described in order to best explain the principles of the invention and the practical application, and to enable others of ordinary skill in the art to understand the invention for various embodiments with various modifications as are suited to the particular use contemplated.
Claims (14)
1. A natural language client automatic response method, comprising:
inquiring a database to obtain an answer of an original question of a user, wherein the answer is an example meeting conditions in the database, and all attributes of the example are used as candidate slot positions;
judging whether the number of answers is larger than a preset value or not;
if the number of the answers is not more than the preset value, judging whether the answers are inquired or not;
if the answer is inquired and the number of the answers is not more than the preset value, returning the answer to the user;
if the number of the answers is larger than a preset value, a question is asked to the user;
wherein said asking the user a question comprises:
judging whether a candidate slot position exists;
if no candidate slot position exists, the answer is truncated;
if the candidate slot position exists, scoring the candidate slot position;
judging whether the candidate slot position with the highest score is directly asked in a reverse manner;
if the candidate slot position with the highest score is directly asked in a reverse way, the possible value corresponding to the candidate slot position is asked in a reverse way;
if the candidate slot position with the highest score is not directly asked, the sorted candidate slot position list is returned, and the user selects the interested candidate slot position.
2. The method of claim 1, further comprising:
if the answer is not inquired, judging whether the subject matching fails;
if the subject matching fails, fuzzy matching is carried out, and a possible subject list is returned;
if the subject matching fails, judging whether the attribute matching of the query fails;
if the attribute matching of the query fails, returning an attribute value list corresponding to the subject instance;
if the attribute matching of the query does not fail, the subject is expanded and the restriction condition is relaxed, and then the query is performed.
3. The method of claim 2, wherein said performing fuzzy matching to return a list of possible subjects comprises:
analyzing the original problem of the user by using a natural language processing method, and extracting a subject, a predicate and a fixed language limiting condition of the original problem of the user;
if the precise matching of the subject language fails, recalling the examples meeting the conditions according to the predicate and the predicate limiting conditions as a candidate set;
if the predicate exact matching fails, recalling the attribute corresponding to the subject as a candidate set according to the subject and the subject limiting condition;
in the candidate instance set, scoring each match using a fuzzy matching method;
and selecting the candidate set with the score larger than the preset threshold value to return to the user, and carrying out next confirmation by the user.
4. The method of claim 2, wherein said expanding the subject and relaxing the constraints and then querying comprises:
if the subject is an example in the database, all classes to which the example belongs are inquired upwards recursively, whether a certain class can meet the inquiry condition exists or not, and the inquiry is stopped when an answer is obtained;
if the subject is a class in the database, all father classes of the class are inquired upwards and recursively, whether a certain father class meets the inquiry condition or not is judged, and the inquiry is stopped when an answer is inquired;
if too many subject limiting conditions result in no answer to be inquired, then each limiting condition is tried to be removed in sequence, the number of answers is recorded, and the answer with the minimum number is returned as the answer.
5. The method according to any one of claims 2-4, further comprising:
judging whether an answer is obtained after re-inquiry;
if the answer is obtained after the inquiry, the answer is returned to the user;
if the answer can not be obtained after the inquiry, the original question of the user is not answered.
6. The method of any of claims 1-4, wherein scoring the candidate slots comprises:
acquiring an answer set meeting the original problem of a user and an attribute set of a candidate slot position;
calculating entropy values of each slot position after grouping according to the slot position;
obtaining the interestingness score of each slot position of a user;
performing weighted summation on the interestingness score and the entropy value;
and updating the answer set into a new answer set after the user interaction.
7. The method according to any one of claims 1-4, wherein after returning an answer to the user, further comprising:
recalling all attributes of the instance or class corresponding to the answer;
counting the frequency of each candidate attribute;
obtaining the correlation degree of each candidate attribute and the original problem of the user;
sorting the candidate attributes according to the frequency and the correlation;
and generating a recommended candidate question according to the sorted candidate attributes.
8. An automatic response system for natural language clients is characterized by comprising a query module, an answer quantity judging module, a query result judging module and an answer returning module, wherein:
the query module is used for querying a database to obtain an answer of an original question of a user, wherein the answer is an example which meets the conditions in the database, and all attributes of the example are used as candidate slot positions;
the answer quantity judging module is used for judging whether the answer quantity is greater than a preset value or not;
the query result judging module is used for judging whether the answer is queried or not according to the judgment result of the answer quantity judging module under the condition that the answer quantity is not greater than the preset value;
the answer returning module is used for returning answers to the user under the condition that the answers are inquired according to the judgment result of the inquiry result judgment module;
wherein the natural language client automatic response system further comprises a question-back module, wherein:
the question-returning module is used for providing a question-returning to the user under the condition that the number of the answers is greater than the preset value according to the judgment result of the answer number judgment module;
wherein, ask-back module includes candidate trench judge unit, answer truncation unit, trench grading unit and ask-back judge unit, wherein:
the candidate slot position judging unit is used for judging whether a candidate slot position exists or not under the condition that the answer quantity is greater than a preset value according to the judgment result of the answer quantity judging module;
the answer truncation unit is used for truncating the answer under the condition that no candidate slot position exists according to the judgment result of the candidate slot position judgment unit;
the slot position scoring unit is used for scoring the candidate slot position under the condition that the candidate slot position exists according to the judgment result of the candidate slot position judging unit;
the back-questioning judging unit is used for judging whether to directly ask the candidate slot position with the highest back-questioning score after the slot position scoring unit scores the candidate slot positions; under the condition of directly asking the candidate slot position with the highest score, asking the possible value corresponding to the candidate slot position; and under the condition that the candidate slot position with the highest score is not directly asked, returning the sorted candidate slot position list, and selecting the interested candidate slot position by the user.
9. The system of claim 8, further comprising a subject matching determination module, a fuzzy matching module, an attribute matching determination module, an attribute list returning module, and a subject augmenting module, wherein:
the subject matching judgment module is used for judging whether subject matching fails or not under the condition that an answer is not inquired according to the judgment result of the inquiry result judgment module;
the fuzzy matching module is used for performing fuzzy matching according to the judgment result of the subject matching judgment module under the condition that the subject matching fails and returning a possible subject list;
the attribute matching judgment module is used for judging whether the attribute matching of the query fails or not under the condition that the subject matching fails according to the judgment result of the subject matching judgment module;
the attribute list returning module is used for returning an attribute value list corresponding to the subject instance under the condition that the inquired attribute matching fails according to the judgment result of the attribute matching judgment module;
and the subject extension module is used for extending the subject and relaxing the limiting conditions and then inquiring according to the judgment result of the attribute matching judgment module under the condition that the inquired attribute matching is not failed.
10. The system of claim 9, wherein the fuzzy matching module comprises a condition extraction unit, a candidate set determination unit, a fuzzy scoring unit, and a candidate set return unit, wherein:
the condition extraction unit is used for analyzing the original user problem by using a natural language processing method and extracting a subject, a predicate and a fixed language limiting condition of the original user problem;
the candidate set determining unit is used for recalling the examples meeting the conditions according to the predicates and the fixed-phrase limiting conditions as the candidate set under the condition that the precise matching of the subjects fails; under the condition that the accurate matching of the predicates fails, recalling the attributes corresponding to the subject as a candidate set according to the subject and the subject limiting conditions;
the fuzzy scoring unit is used for scoring each match in the candidate example set by using a fuzzy matching method;
and the candidate set returning unit is used for selecting the candidate set with the score larger than the preset threshold value to return to the user and the user confirms next step.
11. The system of claim 9, wherein the subject augmentation module comprises a first query unit, a second query unit, and a condition removal unit, wherein:
the first query unit is used for querying all classes to which the instance belongs upwards and recursively under the condition that the subject is the instance in the database, judging whether a certain class can meet query conditions, and stopping querying the answer;
the second query unit is used for inquiring all father classes of the class upwards and recursively under the condition that the subject is the class in the database, judging whether a certain father class meets the query condition, and stopping when the answer is queried;
and the condition removing unit is used for sequentially trying to remove each limiting condition under the condition that the query cannot be answered due to excessive subject limiting conditions, recording the number of answers, and returning the answer with the minimum number as the answer.
12. The system according to any one of claims 9-11, further comprising a re-query result determination module, wherein:
the re-query result judging module is used for judging whether an answer is obtained after re-query; under the condition that the answer is obtained after re-inquiry, indicating an answer returning module to return the answer to the user; and not answering the original question of the user under the condition that the answer can not be obtained after the re-inquiry.
13. The system according to any one of claims 8-11, wherein the slot scoring unit comprises an answer set acquisition sub-module, an entropy determination sub-module, an interestingness determination sub-module, a weighted summation sub-module, and an answer update sub-module, wherein:
the answer set acquisition submodule is used for acquiring an answer set meeting the original problem of the user and an attribute set of the candidate slot position;
the entropy value determining submodule is used for calculating the entropy value grouped according to the slot position for each slot position;
the interest level determining submodule is used for obtaining the interest level score of each slot position of the user;
the weighted summation submodule is used for carrying out weighted summation on the interestingness score and the entropy value;
and the answer updating submodule is used for updating the answer set into a new answer set after the user interacts.
14. The system according to any one of claims 8-11, further comprising an attribute recall module, a frequency statistics module, a relevancy determination module, a ranking module, and a recommendation question generation module, wherein:
the attribute recalling module is used for recalling all attributes of the instance or the class corresponding to the answer after the answer returning module returns the answer to the user;
the frequency counting module is used for counting the frequency of each candidate attribute;
the frequency counting module is used for counting the frequency of each candidate attribute;
the relevancy determining module is used for acquiring the relevancy of each candidate attribute and the original problem of the user;
the sorting module is used for sorting the candidate attributes according to the frequency and the relevancy;
and the recommendation problem generation module is used for generating recommended candidate problems according to the candidate attributes sorted by the sorting module.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611185779.4A CN108228637B (en) | 2016-12-21 | 2016-12-21 | Automatic response method and system for natural language client |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611185779.4A CN108228637B (en) | 2016-12-21 | 2016-12-21 | Automatic response method and system for natural language client |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108228637A CN108228637A (en) | 2018-06-29 |
CN108228637B true CN108228637B (en) | 2020-09-04 |
Family
ID=62651726
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611185779.4A Active CN108228637B (en) | 2016-12-21 | 2016-12-21 | Automatic response method and system for natural language client |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108228637B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110737671A (en) * | 2018-07-03 | 2020-01-31 | 百度在线网络技术(北京)有限公司 | Table-based retrieval method and device |
CN109815320B (en) * | 2018-12-26 | 2021-03-05 | 出门问问信息科技有限公司 | Answer generation method, device, equipment and storage medium of question-answering system |
CN109753557B (en) * | 2018-12-26 | 2021-01-01 | 出门问问信息科技有限公司 | Answer output method, device, equipment and storage medium of question-answering system |
CN111694932A (en) * | 2019-03-13 | 2020-09-22 | 百度在线网络技术(北京)有限公司 | Conversation method and device |
WO2020186458A1 (en) | 2019-03-19 | 2020-09-24 | 京东方科技集团股份有限公司 | Problem generation method and apparatus, inquiry system, and computer-readable storage medium |
CN112337094B (en) * | 2020-10-27 | 2022-11-25 | 四川长虹电器股份有限公司 | Voice game interaction method |
CN112667895B (en) * | 2020-12-28 | 2024-08-13 | 百果园技术(新加坡)有限公司 | Recommended item queue determining method, device, equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102236677A (en) * | 2010-04-28 | 2011-11-09 | 北京大学深圳研究生院 | Question answering system-based information matching method and system |
CN102368260A (en) * | 2011-10-12 | 2012-03-07 | 北京百度网讯科技有限公司 | Method and device of producing domain required template |
CN104216913A (en) * | 2013-06-04 | 2014-12-17 | Sap欧洲公司 | Problem answering frame |
CN105824933A (en) * | 2016-03-18 | 2016-08-03 | 苏州大学 | Automatic question answering system based on main statement position and implementation method thereof |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8832064B2 (en) * | 2005-11-30 | 2014-09-09 | At&T Intellectual Property Ii, L.P. | Answer determination for natural language questioning |
CN101872349B (en) * | 2009-04-23 | 2013-06-19 | 国际商业机器公司 | Method and device for treating natural language problem |
US20140006012A1 (en) * | 2012-07-02 | 2014-01-02 | Microsoft Corporation | Learning-Based Processing of Natural Language Questions |
CN103021403A (en) * | 2012-12-31 | 2013-04-03 | 威盛电子股份有限公司 | Voice recognition based selecting method and mobile terminal device and information system thereof |
CN104809197A (en) * | 2015-04-24 | 2015-07-29 | 同程网络科技股份有限公司 | On-line question and answer method based on intelligent robot |
CN106095872A (en) * | 2016-06-07 | 2016-11-09 | 北京高地信息技术有限公司 | Answer sort method and device for Intelligent Answer System |
-
2016
- 2016-12-21 CN CN201611185779.4A patent/CN108228637B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102236677A (en) * | 2010-04-28 | 2011-11-09 | 北京大学深圳研究生院 | Question answering system-based information matching method and system |
CN102368260A (en) * | 2011-10-12 | 2012-03-07 | 北京百度网讯科技有限公司 | Method and device of producing domain required template |
CN104216913A (en) * | 2013-06-04 | 2014-12-17 | Sap欧洲公司 | Problem answering frame |
CN105824933A (en) * | 2016-03-18 | 2016-08-03 | 苏州大学 | Automatic question answering system based on main statement position and implementation method thereof |
Also Published As
Publication number | Publication date |
---|---|
CN108228637A (en) | 2018-06-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108228637B (en) | Automatic response method and system for natural language client | |
US9678992B2 (en) | Text to image translation | |
US9646260B1 (en) | Using existing relationships in a knowledge base to identify types of knowledge for addition to the knowledge base | |
CN111949758A (en) | Medical question and answer recommendation method, recommendation system and computer readable storage medium | |
Li et al. | A hybrid recommendation system for Q&A documents | |
CN112100396B (en) | Data processing method and device | |
JP6355840B2 (en) | Stopword identification method and apparatus | |
CN111078837A (en) | Intelligent question and answer information processing method, electronic equipment and computer readable storage medium | |
US20030212663A1 (en) | Neural network feedback for enhancing text search | |
CN111737608B (en) | Method and device for ordering enterprise information retrieval results | |
CN102968419A (en) | Disambiguation method for interactive Internet entity name | |
US20220083879A1 (en) | Inferring a comparative advantage of multi-knowledge representations | |
CN112507089A (en) | Intelligent question-answering engine based on knowledge graph and implementation method thereof | |
KR101780237B1 (en) | Method and device for answering user question based on q&a data provided on online | |
CN111160699A (en) | Expert recommendation method and system | |
Calado et al. | A Bayesian network approach to searching Web databases through keyword-based queries | |
US20210406291A1 (en) | Dialog driven search system and method | |
CN111737523B (en) | Video tag, generation method of search content and server | |
CN117609468A (en) | Method and device for generating search statement | |
CN108810640B (en) | Television program recommendation method | |
Mirzadeh et al. | Cooperative query rewriting for decision making support and recommender systems | |
CN116414940A (en) | Standard problem determining method and device and related equipment | |
CN108984582A (en) | A kind of inquiry request processing method | |
CN110222156B (en) | Method and device for discovering entity, electronic equipment and computer readable medium | |
Meng et al. | Adaptive query relaxation and top-k result ranking over autonomous web databases |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |