CN117972069A - Method for carrying out active dialogue and knowledge base vector search based on artificial intelligence - Google Patents
Method for carrying out active dialogue and knowledge base vector search based on artificial intelligence Download PDFInfo
- Publication number
- CN117972069A CN117972069A CN202410381555.9A CN202410381555A CN117972069A CN 117972069 A CN117972069 A CN 117972069A CN 202410381555 A CN202410381555 A CN 202410381555A CN 117972069 A CN117972069 A CN 117972069A
- Authority
- CN
- China
- Prior art keywords
- resource
- data
- dialogue
- precision
- trend
- 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.)
- Granted
Links
- 239000013598 vector Substances 0.000 title claims abstract description 106
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000011156 evaluation Methods 0.000 claims description 99
- 230000014759 maintenance of location Effects 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 abstract description 6
- 239000002699 waste material Substances 0.000 abstract description 6
- 230000015556 catabolic process Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 23
- 238000004458 analytical method Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 238000003058 natural language processing Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to the technical field of intelligent answer, and particularly relates to a method for carrying out active dialogue and knowledge base vector search based on artificial intelligence. According to the invention, the system can more effectively manage resources by monitoring the utilization rate and the reservation rate of the resources, avoid excessive use or waste of the resources, adjust the conversation quality according to the utilization rate and the grade of the resources so as to ensure user experience, more quickly find out a proper answer by the determination of pre-search and actual search precision, improve the search efficiency, and improve the stability and the reliability of the system by dynamically adjusting the conversation grade and the utilization rate of the resources, and avoid system breakdown or errors caused by insufficient resources.
Description
Technical Field
The invention belongs to the technical field of intelligent answer, and particularly relates to a method for carrying out active dialogue and knowledge base vector search based on artificial intelligence.
Background
With the continuous development and application of artificial intelligence technology, natural Language Processing (NLP) and knowledge-graph technology are increasingly popular in the fields of dialogue systems and information retrieval. The traditional dialogue system based on rules or templates has the problems of limited language expression capability, poor adaptability and the like, and the dialogue system based on artificial intelligence can more flexibly understand and generate natural language and can perform personalized interaction and information pushing according to the requirements of users.
The current artificial intelligent dialogue system is mainly based on deep learning, reinforcement learning and other technologies, and model training is carried out through a large amount of corpus data, so that dialogue understanding and generation are realized. These systems typically include Natural Language Understanding (NLU), dialog Management (DM), and Natural Language Generation (NLG) modules that enable semantic understanding and intent recognition of the natural language input by the user and generate a context-compliant natural language response.
On the other hand, a knowledge base is a structured data set that stores rich information, including various entities, attributes, and relationships between them. The knowledge base can provide a rich background knowledge to the dialog system, helping the system to better understand the user's queries and give accurate answers. However, the conventional knowledge base retrieval method is generally based on keyword matching or rule-based query, is difficult to process complex natural language query, and often needs to manually maintain the content and structure of the knowledge base, so that the resource utilization rate is poor, the search efficiency is low, the problems cause system performance degradation, the use of systematic resources is wasted, and the user experience is poor.
Disclosure of Invention
The invention aims to provide a method for carrying out active dialogue and knowledge base vector search based on artificial intelligence, which can realize the optimized management of a dialogue system through dynamic resource adjustment and precision evaluation, improves the system performance and resource utilization rate and provides wider and more accurate information search and answer service.
The technical scheme adopted by the invention is as follows:
A method for active dialogue and knowledge base vector search based on artificial intelligence, comprising:
acquiring first resource data of an artificial intelligent dialogue, and determining a first resource trend according to the first resource data;
According to the first resource trend, acquiring a first dialogue level of the artificial intelligence;
Acquiring system resource retention rate data, and acquiring second resource data of an artificial intelligent dialogue according to the first dialogue level of the artificial intelligent;
acquiring dialogue data and converting the dialogue data into first vector data;
Judging first precision data of the knowledge base vector search according to the first vector data;
Acquiring a knowledge base vector to search the second precision data according to the second resource data and the first precision data;
and obtaining answers of the artificial intelligence dialogue from the knowledge base according to the second precision data.
In a preferred embodiment, the step of obtaining first resource data of the artificial intelligence dialog and determining a first resource trend according to the first resource data includes:
Acquiring first resource data of a current artificial intelligence dialogue;
acquiring a first trend model, and extracting a first trend function from the first trend model;
inputting historical resource data and current resource data into a first trend function to obtain a first trend value;
Acquiring a first standard trend evaluation interval, and judging a first resource trend according to the first standard trend evaluation interval and a first trend value;
if the first trend value belongs to the first standard trend evaluation interval, the first resource trend is indicated to be stable;
if the first trend value is greater than the first standard trend evaluation interval, indicating that the first resource trend is enhanced;
if the first trend value is smaller than the first standard trend evaluation interval, the first resource trend is indicated to be reduced.
In a preferred embodiment, the step of obtaining the first resource data of the current artificial intelligence session includes:
Acquiring the acquisition frequency of resource data;
Constructing acquisition nodes according to the acquisition frequency;
And acquiring the resource data corresponding to each acquisition node, namely the first resource data of the current artificial intelligence dialogue.
In a preferred embodiment, the step of obtaining the first dialog level of the artificial intelligence according to the first resource trend includes:
acquiring a first trend value of a first resource trend;
acquiring a first dialogue evaluation level interval;
Judging a first dialogue grade of the artificial intelligence according to the first trend value and the first dialogue evaluation grade interval;
if the first trend value belongs to a first-stage evaluation interval in the first dialogue evaluation grade interval, the first dialogue grade is shown to be the first stage, and the occupied resource rate is the highest;
If the first trend value belongs to a second-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is indicated to be second-level, and the occupied resource rate is high;
If the first trend value belongs to a third-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is three-level, and the occupied resource rate belongs to the standard occupancy rate;
if the first trend value belongs to a four-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be four-level, and the occupied resource rate is low;
if the first trend value belongs to a five-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be five-level, and the occupied resource rate is the lowest.
In a preferred embodiment, the step of obtaining the system resource reservation rate data and obtaining the second resource data of the artificial intelligence session according to the first session level of the artificial intelligence includes:
acquiring system resource retention rate data;
acquiring first resource data and a first dialogue grade of all artificial intelligence dialogues;
Acquiring a second resource function;
And inputting the system resource retention rate data, the first resource data and the first dialogue grade into a second resource function to obtain a second resource value, namely the second resource data of the artificial intelligent dialogue.
In a preferred embodiment, the step of obtaining dialogue data and converting dialogue data into first vector data includes:
Acquiring dialogue data;
Extracting text information from the dialogue data;
splitting the text information into a plurality of characteristic information blocks;
and converting the dialogue data into a vector information set, namely the first vector data, according to the plurality of characteristic information blocks.
In a preferred embodiment, the step of determining the first precision data of the knowledge base vector search according to the first vector data includes:
Acquiring a standard precision evaluation interval;
Matching the first vector data with a standard precision evaluation interval, and judging first precision data of the knowledge base vector search according to a matched interval result;
If the first vector data belongs to a first-level precision interval in the standard precision evaluation interval, the first precision data indicating that the knowledge base vector search is fine search, namely, the occupied system resource is large;
If the first vector data belongs to a secondary precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is standard search, namely the occupied system resources are normal;
if the first vector data belongs to a three-level precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is rough search, namely, the occupied system resource is small.
In a preferred embodiment, the step of obtaining the knowledge base vector to search the second precision data according to the second resource data and the first precision data includes:
respectively extracting first precision compensation values from all the first precision data;
Acquiring a second compensation precision resource model, and extracting a second compensation precision resource function from the second compensation precision resource model;
inputting the second resource value and the first precision compensation value into a second compensation precision resource function to obtain a second compensation precision resource value;
judging whether the second compensation precision resource value is larger than a second resource value or not;
If the second compensation precision resource value is smaller than or equal to the second resource value, the second compensation precision resource value is indicated to be the second precision data;
and if the second compensation precision resource value is larger than the second resource data, indicating that the first precision data is the second precision data.
In a preferred embodiment, the step of obtaining the artificial intelligence dialogue answer from the knowledge base according to the second accuracy data includes:
Acquiring the resource occupancy rate of vector search from the second precision data;
And selecting corresponding time of vector search according to the resource occupancy rate, and acquiring answers of the artificial intelligent dialogue from a knowledge base according to the corresponding time of vector search.
And a terminal for active dialogue and knowledge base vector search based on artificial intelligence, comprising:
one or more processors;
A storage device having one or more programs stored thereon;
The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method for active dialogue and knowledge base vector search based on artificial intelligence.
The invention has the technical effects that:
according to the invention, the system can more effectively manage resources by monitoring the utilization rate and the reservation rate of the resources, avoid excessive use or waste of the resources, adjust the conversation quality according to the utilization rate and the grade of the resources so as to ensure user experience, more quickly find out proper answers by the determination of pre-search and actual search precision, improve the search efficiency, and improve the stability and the reliability of the system by dynamically adjusting the conversation grade and the utilization rate of the resources, thereby avoiding system breakdown or errors caused by insufficient resources.
Drawings
FIG. 1 is a flow chart of a method provided by the present invention;
Detailed Description
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one preferred embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Further, the present invention will be described in detail with reference to the drawings, which are only examples for convenience of illustration, and should not limit the scope of the present invention.
Referring to fig. 1, a method for performing active dialogue and knowledge base vector search based on artificial intelligence is provided, which includes:
S1, acquiring first resource data of an artificial intelligent dialogue, and determining a first resource trend according to the first resource data; the first resource trend is system resource utilization rate data, wherein the system resource utilization rate data comprises current dialogue resource utilization rate information and other dialogue resource utilization rate information;
S2, acquiring a first dialogue level of the artificial intelligence according to the first resource trend;
s3, acquiring system resource retention rate data, and acquiring second resource data of the artificial intelligence dialogue according to the first dialogue level of the artificial intelligence; the second resource data is the utilization rate of the available resources of the system;
s4, acquiring dialogue data and converting the dialogue data into first vector data;
S5, judging first precision data of the knowledge base vector search according to the first vector data; the first precision data is the pre-search precision
S6, acquiring a knowledge base vector according to the second resource data and the first precision data to search the second precision data; the second precision data is the actual searching precision;
s7, according to the second precision data, obtaining the artificial intelligent dialogue answer from the knowledge base.
By monitoring the utilization data of the system resources, as in steps S1 to S7 above, the system can learn about the load situation of the current dialog system, which may be one or more of CPU utilization, memory utilization, network bandwidth utilization, or algorithm resource utilization, etc., from which it can be determined whether the system' S resource trend, i.e. the utilization of the resources, is in a high-load or low-load state, and based on the resource utilization data, the system can determine the level of the current dialog system, e.g. if the resource utilization is high, it may be necessary to limit the functionality of the dialog system or reduce the dialog quality, to ensure the stability of the system, learn about the current available resource situation of the system, to decide whether more dialog processes can be performed, the system resource utilization can help the system determine whether there are more resources enough to process new dialog requests, the conversion of dialogue data into vector data is for facilitating subsequent processing and analysis, the vector representation can better express semantic information and can be used for subsequent searching and matching, based on the first vector data, the system can perform pre-searching to estimate possible searching accuracy, which can help the system to quickly screen out some possible answers before performing actual searching, the system can determine the actual searching accuracy according to available resources of the system and the pre-searching accuracy, which can help the system to improve searching accuracy and efficiency as much as possible on the premise of ensuring resource utilization, finally, the system can acquire the most suitable answer from a knowledge base according to the actual searching accuracy and provide the most suitable answer to a user, the system can more effectively manage resources by monitoring the resource utilization and reservation rate, avoid excessive use or waste of the resources, according to the resource utilization rate and the level, the system can adjust the dialogue quality to ensure the user experience, the system can find a proper answer faster through the determination of the pre-search and the actual search precision, the search efficiency is improved, and the stability and the reliability of the system can be improved through dynamically adjusting the dialogue level and the resource utilization rate, so that the system breakdown or error caused by insufficient resources is avoided.
The method comprises the steps of obtaining first resource data of an artificial intelligence dialogue and determining a first resource trend according to the first resource data, wherein the steps comprise:
S101, acquiring first resource data of a current artificial intelligence dialogue;
s102, acquiring a first trend model, and extracting a first trend function from the first trend model;
S103, inputting historical resource data and current resource data into a first trend function to obtain a first trend value;
s104, acquiring a first standard trend evaluation interval, and judging a first resource trend according to the first standard trend evaluation interval and a first trend value;
if the first trend value belongs to the first standard trend evaluation interval, the first resource trend is indicated to be stable;
if the first trend value is greater than the first standard trend evaluation interval, indicating that the first resource trend is enhanced;
if the first trend value is smaller than the first standard trend evaluation interval, the first resource trend is indicated to be reduced.
As in steps S104 to S104, collecting resource data of the current artificial intelligence dialogue system, where the data may include one or more of CPU utilization, memory utilization, network bandwidth utilization, algorithm resource utilization, etc., and the data is key to evaluating system resource utilization, the first trend model may be a mathematical model for describing a trend of resource utilization over time, and the first trend function extracted from the first trend model may help the system understand a rule of change of resource utilization, where the first trend function isWherein, Q represents a first trend value, i represents the number of the first resource data, n represents the total number of the first resource data, L represents the first resource data, B represents the first standard resource data, D represents the last data in the first resource data, Z represents the first data in the first resource data, by inputting the historical resource data and the current resource data into the first trend function, the system can obtain a predicted trend value of the resource utilization, which helps the system to know whether the resource utilization is increasing, decreasing or keeping stable, the first standard trend evaluation interval is determined according to the requirements and characteristics of the system, the interval can help the system to judge whether the change of the resource utilization is within an acceptable range, and the system can judge the trend of the first resource according to the relation between the first trend value and the first standard trend evaluation interval:
if the first trend value belongs to the first standard trend evaluation interval, indicating that the resource trend is stable;
if the first trend value is larger than the first standard trend evaluation interval, indicating that the resource trend is enhanced;
if the first trend value is smaller than the first standard trend evaluation interval, indicating that the resource trend is reduced;
Through establishing trend models and functions, the system can predict the trend of the resource utilization rate, so that corresponding adjustment can be made in advance, resources can be reasonably allocated by the system according to evaluation of the resource trend, the requirement of a dialogue system can be met, the condition of resource waste or deficiency is avoided, and real-time monitoring and evaluation of the resource trend can help the system to timely adjust parameters such as dialogue quality, response speed and the like, so that the system performance is optimized, and the user experience is improved.
The step of obtaining first resource data of a current artificial intelligence dialog comprises the steps of:
S1011, acquiring the acquisition frequency of the resource data;
s1012, constructing acquisition nodes according to the acquisition frequency;
S1013, acquiring resource data corresponding to each acquisition node, namely the first resource data of the current artificial intelligent dialogue.
In steps S1011 to S1013, the collection frequency refers to a time interval during which the system collects resource data, which may be determined according to requirements and performance requirements of the system, and usually is in units of seconds, milliseconds, and microseconds, the collection node refers to a time point determined according to the collection frequency on a time axis, for example, if the collection frequency is once every millisecond, there is one collection node every millisecond, on each collection node, the system may obtain corresponding resource data, including one or more of CPU utilization, memory utilization, network bandwidth utilization, or algorithm resource utilization, etc., and these data may be obtained through a system monitoring tool or API, by setting the collection frequency and constructing the collection node, the system may monitor usage conditions of resources in real time, discover changes in resource utilization in time, frequently obtain resource data may provide more accurate resource utilization information, the system may perform finer tuning according to these information to improve performance and efficiency of the system, by continuously collecting resource data, the system may perform trend analysis and prediction to help the system predict future resource requirements, and accordingly adjust and plan future resource utilization, and schedule future resource utilization, and may be able to solve the system failure or may have a problem or may be solved by timely planning the future resource utilization.
The method comprises the steps of obtaining a first dialogue level of the artificial intelligence according to a first resource trend, wherein the steps comprise:
s201, acquiring a first trend value of a first resource trend;
S202, acquiring a first dialogue evaluation level interval;
s203, judging a first dialogue grade of the artificial intelligence according to the first trend value and the first dialogue evaluation grade interval;
if the first trend value belongs to a first-stage evaluation interval in the first dialogue evaluation grade interval, the first dialogue grade is shown to be the first stage, and the occupied resource rate is the highest;
If the first trend value belongs to a second-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is indicated to be second-level, and the occupied resource rate is high;
If the first trend value belongs to a third-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is three-level, and the occupied resource rate belongs to the standard occupancy rate;
if the first trend value belongs to a four-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be four-level, and the occupied resource rate is low;
if the first trend value belongs to a five-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be five-level, and the occupied resource rate is the lowest.
As in steps S201 to S203, the first resource trend value is a result obtained according to the trend analysis described above, reflects a trend of change of the system resource utilization, may be stable, enhanced or reduced, and the first session evaluation level interval is set according to the performance requirement and the resource management policy of the system, and may be classified into different levels according to the specific situation of the system, so as to reflect different requirements on the resource utilization, and the system may determine the level of the first session of the artificial intelligence according to the first trend value and the first session evaluation level interval:
if the first trend value belongs to a first-stage evaluation interval in the first dialogue evaluation grade interval, the first dialogue grade is shown to be the first stage, and the occupied resource rate is the highest;
If the first trend value belongs to a second-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is indicated to be second-level, and the occupied resource rate is high;
If the first trend value belongs to a third-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is three-level, and the occupied resource rate belongs to the standard occupancy rate;
if the first trend value belongs to a four-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be four-level, and the occupied resource rate is low;
If the first trend value belongs to a five-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be five-level, and the occupied resource rate is the lowest;
By determining the conversation level according to the resource trend, the system can more effectively manage resources, avoid excessive use or waste of the resources, dynamically adjust parameters such as conversation quality, response speed and the like according to different conversation levels, so as to adapt to the current resource situation, improve system performance and user experience, allocate priorities to the demands of the resources according to the conversations of different levels, ensure that the system can meet the demands of various conversations to the greatest extent, and improve the overall efficiency and stability of the system.
The method comprises the steps of obtaining system resource retention rate data and obtaining second resource data of an artificial intelligence dialogue according to a first dialogue level of the artificial intelligence, and comprises the following steps:
S301, acquiring system resource reservation rate data;
s302, acquiring first resource data and a first dialogue grade of all artificial intelligence dialogues;
s303, acquiring a second resource function;
s304, inputting the system resource retention rate data, the first resource data and the first dialogue grade into a second resource function to obtain a second resource value, namely the second resource data of the artificial intelligent dialogue.
In steps S301 to S304, the system resource reservation rate data reflects the spare resources remaining when the current system resource of the system reaches a peak value, the reservation rate data indicates how much spare resources remain for the system to operate normally, the first resource data is the previously acquired resource data of the current artificial intelligence session, the first session level of all the artificial intelligence sessions is the previously determined level according to the first resource trend, the second resource function is a mathematical model, the second resource value can be calculated according to the current resource reservation rate of the system, the first resource data, the first session level, and other factors, the function can be defined and adjusted according to the requirements and performance requirements of the system, and the second resource function isThe system takes the system resource retention rate data, the first resource data and the first dialogue grade as input, inputs the system resource retention rate data, the first resource data and the first dialogue grade into a second resource function for calculation, and obtains the second resource value, namely the second resource data of the artificial intelligent dialogue.
The step of obtaining dialogue data and converting the dialogue data into first vector data comprises the steps of:
S401, acquiring dialogue data;
S402, extracting text information from dialogue data;
s403, splitting the text information into a plurality of characteristic information blocks;
s404, converting the dialogue data into a vector information set according to the characteristic information blocks, namely the first vector data.
In the steps S401 to S404, the dialogue data may be text or speech data collected from interactions between the user and the artificial intelligence dialogue system, where the dialogue data includes questions posed by the user, answers of the system, and possible context information, the system extracts text information from the dialogue data, and for the text dialogue, this may involve a natural language processing technique such as word segmentation, stop word removal, word part-of-speech tagging, etc., so as to further process, the system splits the extracted text information into a plurality of feature information blocks, where each feature information block may be a part of a feature vector, and finally, the system represents each feature information block as a vector, and combines all feature information blocks into a vector information set, that is, first vector data, where the vectors may be forms such as word embedding vector, TF-IDF vector, etc., so as to characterize semantic information and features of the dialogue data, and convert the dialogue data into a vector form, so as to better represent the semantic information of the dialogue data, thereby facilitating the subsequent processing of the extracted text information and splitting the text information into a plurality of feature information blocks, each feature information block may be used as a part of a feature vector, and finally, each feature information block may be represented as a vector, and all feature information blocks may be combined into a vector information, and all feature information sets may be combined into a vector information, and may be combined into a vector information set, and may be used to characterize the semantic information, and feature information, and may be more convenient to implement a feature analysis, and may be more accurate, and a feature-based on the feature analysis system may be performed.
Judging first precision data of the knowledge base vector search according to the first vector data, wherein the first precision data comprises the following steps:
s501, acquiring a standard precision evaluation interval;
S502, matching the first vector data with a standard precision evaluation interval, and judging first precision data of the knowledge base vector search according to a matched interval result;
If the first vector data belongs to a first-level precision interval in the standard precision evaluation interval, the first precision data indicating that the knowledge base vector search is fine search, namely, the occupied system resource is large;
If the first vector data belongs to a secondary precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is standard search, namely the occupied system resources are normal;
if the first vector data belongs to a three-level precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is rough search, namely, the occupied system resource is small.
In the steps S501 to S502, the standard precision evaluation interval is set according to the performance requirement and the search requirement of the system, and the interval can be divided into different precision grades according to the specific condition of the system so as to reflect different requirements on the search precision, the system matches the first vector data with the standard precision evaluation interval, and according to the matching result, the system can judge the first precision data of the knowledge base vector search:
If the first vector data belongs to a first-level precision interval in the standard precision evaluation interval, the first precision data indicating that the knowledge base vector search is fine search, namely, the occupied system resource is large;
If the first vector data belongs to a secondary precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is standard search, namely the occupied system resources are normal;
if the first vector data belongs to a three-level precision interval in the standard precision evaluation interval, the first precision data indicating that the knowledge base vector search is rough search, namely, the occupied system resources are small;
According to the precision requirement of the first vector data, the system can reasonably distribute resources, the condition of resource waste or deficiency is avoided, the system can more effectively search through matching the search precision with the resource utilization rate, the search efficiency and the search accuracy are improved, according to different search precision requirements, the system can provide search results of different levels, different requirements of users on the search results are met, and the user experience is improved.
The step of obtaining the knowledge base vector to search the second precision data according to the second resource data and the first precision data comprises the following steps:
s601, respectively extracting first precision compensation values from all first precision data;
s602, acquiring a second compensation precision resource model, and extracting a second compensation precision resource function from the second compensation precision resource model;
s603, inputting the second resource value and the first precision compensation value into a second compensation precision resource function to obtain a second compensation precision resource value;
S604, judging whether the second compensation precision resource value is larger than a second resource value;
If the second compensation precision resource value is smaller than or equal to the second resource value, the second compensation precision resource value is indicated to be the second precision data;
if the second compensation precision resource value is larger than the second resource data, the first precision data is the second precision data;
In the above steps S601 to S604, for each first precision data, the system needs to calculate corresponding compensation values according to the characteristics and requirements of the first precision data and the actual conditions of the system resources, or by classifying the first precision data, the second compensation precision resource model is a model for calculating a second compensation precision resource value according to the compensation value of the first precision data and the second resource value, the system needs to extract corresponding functions from the model, the second compensation precision resource function is that Wherein E is expressed as a second resource value, T is expressed as a first precision compensation value, a is expressed as the number of all first precision compensation values, b is expressed as the total number of all first precision compensation values, C is expressed as the current first precision compensation value, the system inputs the second resource value and the compensation value of each first precision data into a second compensation precision resource function for calculation, a corresponding second compensation precision resource value is obtained, and for each second compensation precision resource value, the system needs to judge the relation between the second compensation precision resource value and the second resource value:
If the second compensation precision resource value is smaller than or equal to the second resource value, the second compensation precision resource value is indicated to be the second precision data;
If the second compensation precision resource value is larger than the second resource value, the first precision data is the second precision data;
The system can adjust the searching precision according to the actual resource condition by calculating the first precision compensation value and the second compensation precision resource value so as to balance the searching effect and the resource utilization rate, the second compensation precision resource model can be adjusted according to the requirement of the first precision data and the system resource condition so as to optimize the use efficiency of the system resource, and the system can improve the searching precision and efficiency on the premise of ensuring the resource utilization rate by adjusting the searching precision and improve the satisfaction degree of a user on the searching result.
The step of obtaining artificial intelligence dialogue answers from the knowledge base based on the second accuracy data, comprising:
S701, acquiring the resource occupancy rate of vector search from second precision data;
s702, selecting corresponding time of vector search according to the resource occupancy rate, and acquiring answers of the artificial intelligence dialogue from a knowledge base according to the corresponding time of vector search.
In the steps S701 to S702, the information of the resource utilization rate is extracted from the second precision data, the information can be selected according to the comparison between the precision data and the resource data of the preset value of the system, the system needs to extract the resource occupancy rate of the vector search from the data to determine the resource amount required by the search, the system determines the required search time according to the resource occupancy rate of the vector search extracted from the second precision data, the higher the resource occupancy rate is, the more time is likely to be required to complete the search, after the corresponding time of the vector search is determined, the system can search the corresponding content from the knowledge base and provide answers of the artificial intelligent dialogue, the answers can be ordered and selected according to the matching degree of the search result, the system can balance the resource amount required by the search and the resource utilization rate of the system according to the resource occupancy rate in the second precision data, ensure that the search process does not have an excessive influence on the system performance, the system can optimize the search efficiency on the premise of ensuring the resource utilization rate by determining the corresponding search time according to the resource occupancy rate, the system can enhance the accuracy and speed of the search, the answer is satisfied by the intelligent dialogue from the obtained from the knowledge base, and the user experience is satisfied by providing the user.
And a terminal for active dialogue and knowledge base vector search based on artificial intelligence, comprising:
one or more processors;
A storage device having one or more programs stored thereon;
The one or more programs, when executed by the one or more processors, cause the one or more processors to implement a method for active dialogue and knowledge base vector search based on artificial intelligence.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention. Structures, devices and methods of operation not specifically described and illustrated herein, unless otherwise indicated and limited, are implemented according to conventional means in the art.
Claims (10)
1. A method for active dialogue and knowledge base vector search based on artificial intelligence, comprising:
acquiring first resource data of an artificial intelligent dialogue, and determining a first resource trend according to the first resource data;
According to the first resource trend, acquiring a first dialogue level of the artificial intelligence;
Acquiring system resource retention rate data, and acquiring second resource data of an artificial intelligent dialogue according to the first dialogue level of the artificial intelligent;
acquiring dialogue data and converting the dialogue data into first vector data;
Judging first precision data of the knowledge base vector search according to the first vector data;
Acquiring a knowledge base vector to search the second precision data according to the second resource data and the first precision data;
and obtaining answers of the artificial intelligence dialogue from the knowledge base according to the second precision data.
2. The method of claim 1, wherein the steps of obtaining first resource data of the artificial intelligence dialogue and determining a first resource trend based on the first resource data comprise:
Acquiring first resource data of a current artificial intelligence dialogue;
acquiring a first trend model, and extracting a first trend function from the first trend model;
inputting historical resource data and current resource data into a first trend function to obtain a first trend value;
Acquiring a first standard trend evaluation interval, and judging a first resource trend according to the first standard trend evaluation interval and a first trend value;
if the first trend value belongs to the first standard trend evaluation interval, the first resource trend is indicated to be stable;
if the first trend value is greater than the first standard trend evaluation interval, indicating that the first resource trend is enhanced;
if the first trend value is smaller than the first standard trend evaluation interval, the first resource trend is indicated to be reduced.
3. The method of active dialogue and knowledge base vector search based on artificial intelligence of claim 2, wherein the step of obtaining the first resource data of the current artificial intelligence dialogue comprises:
Acquiring the acquisition frequency of resource data;
Constructing acquisition nodes according to the acquisition frequency;
And acquiring the resource data corresponding to each acquisition node, namely the first resource data of the current artificial intelligence dialogue.
4. The method of claim 2, wherein the step of obtaining a first dialogue level of the artificial intelligence based on the first resource trend comprises:
acquiring a first trend value of a first resource trend;
acquiring a first dialogue evaluation level interval;
Judging a first dialogue grade of the artificial intelligence according to the first trend value and the first dialogue evaluation grade interval;
if the first trend value belongs to a first-stage evaluation interval in the first dialogue evaluation grade interval, the first dialogue grade is shown to be the first stage, and the occupied resource rate is the highest;
If the first trend value belongs to a second-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is indicated to be second-level, and the occupied resource rate is high;
If the first trend value belongs to a third-level evaluation interval in the first dialogue evaluation level interval, the first dialogue evaluation level is three-level, and the occupied resource rate belongs to the standard occupancy rate;
if the first trend value belongs to a four-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be four-level, and the occupied resource rate is low;
if the first trend value belongs to a five-level evaluation interval in the first dialogue evaluation level interval, the first dialogue level is indicated to be five-level, and the occupied resource rate is the lowest.
5. The method of claim 4, wherein the step of obtaining system resource retention data and obtaining artificial intelligence session second resource data according to an artificial intelligence first session level comprises:
acquiring system resource retention rate data;
acquiring first resource data and a first dialogue grade of all artificial intelligence dialogues;
Acquiring a second resource function;
And inputting the system resource retention rate data, the first resource data and the first dialogue grade into a second resource function to obtain a second resource value, namely the second resource data of the artificial intelligent dialogue.
6. The method of claim 1, wherein the step of obtaining dialogue data and converting the dialogue data into first vector data comprises:
Acquiring dialogue data;
Extracting text information from the dialogue data;
splitting the text information into a plurality of characteristic information blocks;
and converting the dialogue data into a vector information set, namely the first vector data, according to the plurality of characteristic information blocks.
7. The method of claim 1, wherein the step of determining first accuracy data of the knowledge base vector search based on the first vector data comprises:
Acquiring a standard precision evaluation interval;
Matching the first vector data with a standard precision evaluation interval, and judging first precision data of the knowledge base vector search according to a matched interval result;
If the first vector data belongs to a first-level precision interval in the standard precision evaluation interval, the first precision data indicating that the knowledge base vector search is fine search, namely, the occupied system resource is large;
If the first vector data belongs to a secondary precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is standard search, namely the occupied system resources are normal;
if the first vector data belongs to a three-level precision interval in the standard precision evaluation interval, the first precision data indicating the knowledge base vector search is rough search, namely, the occupied system resource is small.
8. The method of claim 5, wherein the step of obtaining knowledge base vectors to search for second accuracy data based on second resource data and first accuracy data comprises:
respectively extracting first precision compensation values from all the first precision data;
Acquiring a second compensation precision resource model, and extracting a second compensation precision resource function from the second compensation precision resource model;
inputting the second resource value and the first precision compensation value into a second compensation precision resource function to obtain a second compensation precision resource value;
judging whether the second compensation precision resource value is larger than a second resource value or not;
If the second compensation precision resource value is smaller than or equal to the second resource value, the second compensation precision resource value is indicated to be the second precision data;
and if the second compensation precision resource value is larger than the second resource data, indicating that the first precision data is the second precision data.
9. The method of claim 1, wherein the step of obtaining artificial intelligence dialogue answers from a knowledge base based on second accuracy data comprises:
Acquiring the resource occupancy rate of vector search from the second precision data;
And selecting corresponding time of vector search according to the resource occupancy rate, and acquiring answers of the artificial intelligent dialogue from a knowledge base according to the corresponding time of vector search.
10. A terminal for active dialogue and knowledge base vector search based on artificial intelligence, comprising:
one or more processors;
A storage device having one or more programs stored thereon;
The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of performing artificial intelligence based active dialogue and knowledge base vector search of any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410381555.9A CN117972069B (en) | 2024-04-01 | 2024-04-01 | Method for carrying out active dialogue and knowledge base vector search based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410381555.9A CN117972069B (en) | 2024-04-01 | 2024-04-01 | Method for carrying out active dialogue and knowledge base vector search based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117972069A true CN117972069A (en) | 2024-05-03 |
| CN117972069B CN117972069B (en) | 2024-05-28 |
Family
ID=90854997
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410381555.9A Active CN117972069B (en) | 2024-04-01 | 2024-04-01 | Method for carrying out active dialogue and knowledge base vector search based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117972069B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020152274A1 (en) * | 2001-04-12 | 2002-10-17 | Matsushita Graphic Communication Systems, Inc. | Apparatus and method for receiving e-mail |
| JP2007102685A (en) * | 2005-10-07 | 2007-04-19 | Quality Kk | Personal information search program |
| US20090228528A1 (en) * | 2008-03-06 | 2009-09-10 | International Business Machines Corporation | Supporting sub-document updates and queries in an inverted index |
| KR101357804B1 (en) * | 2012-10-24 | 2014-02-03 | 이동균 | System and methdo for question and answer |
| US20180081628A1 (en) * | 2016-09-21 | 2018-03-22 | International Business Machines Corporation | Preserving Temporal Relevance in a Response to a Query |
| US20180276525A1 (en) * | 2015-12-03 | 2018-09-27 | Huawei Technologies Co., Ltd. | Method and neural network system for human-computer interaction, and user equipment |
| CN108829865A (en) * | 2018-06-22 | 2018-11-16 | 海信集团有限公司 | Information retrieval method and device |
| CN111030534A (en) * | 2019-12-20 | 2020-04-17 | 东南大学 | Parameter identification method for permanent magnet synchronous motor in steady-state operation mode |
| CN111506394A (en) * | 2020-04-15 | 2020-08-07 | 广东工业大学 | Docker Swarm container scheduling method and system |
| CN116049362A (en) * | 2022-12-26 | 2023-05-02 | 合肥照隼互联网科技有限公司 | Big data optimization method and server applied to digital cloud service |
| CN116089640A (en) * | 2022-11-18 | 2023-05-09 | 西安电子科技大学青岛计算技术研究院 | Self-adaptive face searching method, system, medium, equipment and terminal |
| CN116226338A (en) * | 2022-11-16 | 2023-06-06 | 四川封面传媒科技有限责任公司 | Multi-round dialogue system and method based on searching and generating fusion |
| CN117009469A (en) * | 2022-08-24 | 2023-11-07 | 腾讯科技（深圳）有限公司 | Intelligent dialogue method, device, equipment and storage medium |
| CN117472511A (en) * | 2023-11-09 | 2024-01-30 | 中国工商银行股份有限公司 | Container resource monitoring method, device, computer equipment and storage medium |
-
2024
- 2024-04-01 CN CN202410381555.9A patent/CN117972069B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020152274A1 (en) * | 2001-04-12 | 2002-10-17 | Matsushita Graphic Communication Systems, Inc. | Apparatus and method for receiving e-mail |
| JP2007102685A (en) * | 2005-10-07 | 2007-04-19 | Quality Kk | Personal information search program |
| US20090228528A1 (en) * | 2008-03-06 | 2009-09-10 | International Business Machines Corporation | Supporting sub-document updates and queries in an inverted index |
| KR101357804B1 (en) * | 2012-10-24 | 2014-02-03 | 이동균 | System and methdo for question and answer |
| US20180276525A1 (en) * | 2015-12-03 | 2018-09-27 | Huawei Technologies Co., Ltd. | Method and neural network system for human-computer interaction, and user equipment |
| US20180081628A1 (en) * | 2016-09-21 | 2018-03-22 | International Business Machines Corporation | Preserving Temporal Relevance in a Response to a Query |
| CN108829865A (en) * | 2018-06-22 | 2018-11-16 | 海信集团有限公司 | Information retrieval method and device |
| CN111030534A (en) * | 2019-12-20 | 2020-04-17 | 东南大学 | Parameter identification method for permanent magnet synchronous motor in steady-state operation mode |
| CN111506394A (en) * | 2020-04-15 | 2020-08-07 | 广东工业大学 | Docker Swarm container scheduling method and system |
| CN117009469A (en) * | 2022-08-24 | 2023-11-07 | 腾讯科技（深圳）有限公司 | Intelligent dialogue method, device, equipment and storage medium |
| CN116226338A (en) * | 2022-11-16 | 2023-06-06 | 四川封面传媒科技有限责任公司 | Multi-round dialogue system and method based on searching and generating fusion |
| CN116089640A (en) * | 2022-11-18 | 2023-05-09 | 西安电子科技大学青岛计算技术研究院 | Self-adaptive face searching method, system, medium, equipment and terminal |
| CN116049362A (en) * | 2022-12-26 | 2023-05-02 | 合肥照隼互联网科技有限公司 | Big data optimization method and server applied to digital cloud service |
| CN117472511A (en) * | 2023-11-09 | 2024-01-30 | 中国工商银行股份有限公司 | Container resource monitoring method, device, computer equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| QINGQING CAO等: "DeQA: On-Device Question Answering", 《MOBISYS \'19: PROCEEDINGS OF THE 17TH ANNUAL INTERNATIONAL CONFERENCE ON MOBILE SYSTEMS, APPLICATIONS, AND SERVICES》, 12 June 2019 (2019-06-12), pages 27 - 40 * |
| 汪晨曦: "基于BERT的法律问答系统", 《中国优秀硕士学位论文全文数据库社会科学Ⅰ辑》, no. 1, 15 January 2023 (2023-01-15), pages 120 - 783 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117972069B (en) | 2024-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110263144A (en) | A kind of answer acquisition methods and device | |
| CN110162775B (en) | Method and device for determining intention recognition accuracy and computer equipment | |
| CN113254620B (en) | Response method, device and equipment based on graph neural network and storage medium | |
| CN111489095A (en) | Risk user management method and device, computer equipment and storage medium | |
| CN117951270A (en) | Document retrieval method and device and related equipment | |
| CN117972069B (en) | Method for carrying out active dialogue and knowledge base vector search based on artificial intelligence | |
| CN110830654B (en) | Method and equipment for playing voice message | |
| CN112925895A (en) | Natural language software operation and maintenance method and device | |
| CN111310072B (en) | Keyword extraction method, keyword extraction device and computer-readable storage medium | |
| CN112925912A (en) | Text processing method, and synonymous text recall method and device | |
| EP4365754A1 (en) | Common question answering method and system, device and medium | |
| CN116186259A (en) | Session cue scoring method, device, equipment and storage medium | |
| CN109002498B (en) | Man-machine conversation method, device, equipment and storage medium | |
| CN111899728B (en) | Training method and device for intelligent voice assistant decision strategy | |
| CN114282606A (en) | Object identification method and device, computer readable storage medium and computer equipment | |
| CN118317047B (en) | Doctor-patient video communication intelligent allocation method and system based on remote consultation scene | |
| CN111984789B (en) | Corpus classification method, corpus classification device and server | |
| CN116127172B (en) | Self-adaptive learning system and method | |
| CN116843203B (en) | Service access processing method, device, equipment, medium and product | |
| CN116431779B (en) | FAQ question-answering matching method and device in legal field, storage medium and electronic device | |
| CN118312601B (en) | Intelligent AI conversation method and device based on AI natural language understanding | |
| CN111930921B (en) | Intention prediction method and device | |
| CN115658283A (en) | Resource allocation method, device, equipment, medium and program product | |
| CN118551849A (en) | Intelligent language response recognition system and method based on attention mechanism | |
| CN114707592A (en) | Model structure obtaining method and device |
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 |