CN113935309A - Skill optimization processing method and system based on semantic platform - Google Patents
Skill optimization processing method and system based on semantic platform Download PDFInfo
- Publication number
- CN113935309A CN113935309A CN202111066358.0A CN202111066358A CN113935309A CN 113935309 A CN113935309 A CN 113935309A CN 202111066358 A CN202111066358 A CN 202111066358A CN 113935309 A CN113935309 A CN 113935309A
- Authority
- CN
- China
- Prior art keywords
- semantic
- skill
- score
- platform
- service
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 26
- 238000003672 processing method Methods 0.000 title claims abstract description 18
- 238000011156 evaluation Methods 0.000 claims abstract description 75
- 238000012549 training Methods 0.000 claims abstract description 49
- 238000004458 analytical method Methods 0.000 claims abstract description 46
- 238000013136 deep learning model Methods 0.000 claims abstract description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 12
- 238000012163 sequencing technique Methods 0.000 claims description 8
- 238000007619 statistical method Methods 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 4
- 238000009826 distribution Methods 0.000 claims description 2
- 238000000034 method Methods 0.000 abstract description 8
- 230000000694 effects Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013550 semantic technology Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- 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
- G06F16/3343—Query execution using phonetics
-
- 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
- G06N5/04—Inference or reasoning models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Acoustics & Sound (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Machine Translation (AREA)
Abstract
The invention relates to the technical field of intelligent voice, and provides a skill optimization processing method and a system based on a semantic platform, wherein a corresponding evaluation template is manufactured in advance according to the analysis rule of each skill service, historical semantic result data of each skill service is acquired, word analysis is carried out to obtain training corpora, a training model is obtained after deep learning model training is carried out, so that after semantic result information generated from the semantic platform is acquired, the semantic result information is automatically and respectively matched with the evaluation template and standard key information of each skill service to obtain a first score and a second score, the actual score of each skill service is further calculated (namely the correctness of a semantic analysis result is automatically detected), and finally, the execution strategies in sequence are generated according to the scores of the actual scores of all the skill services, so that the purposes of automatic evaluation, automatic strategy generation and automatic optimization of the system are realized, the intelligent voice recognition method has the advantages of improving the recognition accuracy of intelligent voice, saving a large amount of labor and time, optimizing economic benefits and reducing operation cost.
Description
Technical Field
The invention relates to the technical field of intelligent voice, in particular to a skill optimization processing method and system based on a semantic platform.
Background
At present, the voice assistant is widely popularized and popularized in various industries, and the voice semantic technology also achieves a better effect through the development of machine learning and deep learning technologies in recent years. Technology companies related to the large voice semantics establish self semantic analysis platforms, and the semantics in each sentence of the user are analyzed and understood through the platforms, so that a correct semantic result is calculated and inferred. Such as:
the user says ' how today's weather ' to the voice assistant, after the semantic platform receives the string of text requests, the Domain (Domain) of the sentence or the skill belonging to the weather Domain is identified through reasoning of various algorithms and models, the intention (Intent) is the query, the slot value (slot) has main time (today), and the like. When the system identifies the semantic content, the system searches for the weather content of the current day. And finally, the semantic platform combines the Domain, Intent, slot, the inquired weather information and other necessary contents together according to a self-defined JSON format and sends the combined weather information and other necessary contents to the client, so that a primary analysis task is completed.
A more sophisticated semantic platform will support tens or hundreds of skills or domains, each representing a different area of real life such as weather, alarm, calendar, music, booking tickets, etc. When a user sends a voice request, the semantic platform analyzes the skills supported by the platform and detailed semantic content in the skill range, which the voice request belongs to.
However, the existing semantic platform does not have an effective means to automatically detect the correctness of the semantic analysis result in the production environment after analyzing and reasoning the semantic information of the voice request through an algorithm or a model. In addition, because the algorithm or the model is designed or trained in advance, and the input request is unlimited, the semantic analysis result is often asked in real application.
Therefore, in order to improve the accuracy of semantic analysis results, if the analysis results recorded by the semantic platform are analyzed manually one by one, the following problems still exist:
the semantic analysis platform with huge user quantity can lead the voice request to reach the quantity of thousands of millions, and because the generated semantic result data quantity is huge, the manual analysis consumes time and labor and has ultralow efficiency.
Disclosure of Invention
The invention provides a skill optimization processing method and system based on a semantic platform, and solves the technical problems that the accuracy of skill selection is low and user experience is poor due to the fact that an effective technical means is lacked in an existing intelligent voice conversation platform to improve the accuracy of semantic analysis results.
In order to solve the technical problems, the invention provides a skill optimization processing method based on a semantic platform, which comprises the following steps:
s1, obtaining semantic result information generated by the semantic platform according to the voice request;
s2, manufacturing a corresponding evaluation template according to the analysis rule of each skill service, matching the evaluation template with the semantic result information, and determining a first score according to the matching degree;
s3, establishing a training model according to historical semantic result data of each skill service, predicting key information corresponding to the semantic result information according to the training model, matching with key information of each skill service standard, and determining a second score according to the matching degree;
s4, calculating the actual score of each skill service according to the first score and the second score;
and S5, generating an execution strategy according to the actual scores of all the skill services.
The basic scheme is characterized in that corresponding evaluation templates are manufactured in advance according to analysis rules of each skill service, historical semantic result data of each skill service are obtained, word analysis is carried out to obtain training corpora, after semantic result information generated from a semantic platform is obtained, the training corpora is automatically matched with the evaluation templates (matched with the format of the semantic result information) and standard key information (matched with predicted key information) of each skill service respectively to obtain a first score and a second score, the actual score of each skill service is calculated (namely the correctness of the semantic analysis result is automatically detected), and finally, execution strategies in sequence are generated according to the score of the actual scores of all the skill services, so that the purposes of automatic evaluation, automatic strategy generation and automatic optimization of a system are achieved, the recognition accuracy of intelligent voice is improved, and a large amount of manpower and time are saved, optimizing economic benefits and reducing operating costs.
In further embodiments, the step S1 includes:
s11, acquiring a voice request of a user and sending the voice request to a semantic platform;
and S12, the semantic platform analyzes or infers the voice request according to a default algorithm and a default model to generate semantic result information and sends the semantic result information to the client.
According to the technical scheme, the semantic platform is improved to recognize the voice request of the user, standardized semantic result information is obtained, and comparison and analysis of subsequent skill service evaluation are facilitated.
In further embodiments, the step S2 includes:
s21, obtaining semantic analysis cases obtained by each skill service according to the analysis rules;
s22, counting the fixed format in the semantic analysis case, and integrating to obtain an evaluation template;
s23, calculating the matching degree of the semantic result information and each skill service evaluation template, and substituting the matching degree into a first preset rule to determine a first score.
According to the scheme, based on the uniformity of output data of actual skill services, a fixed format is counted and integrated into an evaluation template from a semantic analysis case obtained according to an analysis rule, the matching degree of the calculated semantic result information and the evaluation template (namely the matching degree of the format of the calculated semantic result information and the evaluation template) is directly substituted into a first preset rule to obtain a first score, whether a voice request of a user belongs to the corresponding skill service can be quickly judged according to the first score, the processing logic is clear, the data processing efficiency is high, and the method is a powerful basis for detecting the accuracy of the semantic analysis result.
In further embodiments, the step S3 includes:
s31, acquiring historical semantic result data of each skill service, and summarizing corresponding key information according to the corresponding field;
s32, marking all the key information, integrating to obtain a training corpus, and performing deep learning model training to obtain a training model;
and S33, substituting the semantic result information into the training model to obtain predicted key information, calculating the matching degree of the predicted key information and the standard key information of each skill service, and further substituting a second preset rule to determine a second score.
According to the technical scheme, starting from the actual content of a semantic analysis result, historical semantic result data of each skill service is obtained, corresponding key information is summarized according to the corresponding field, training corpora are obtained through labeling and integration, a training model is obtained after deep learning model training is carried out, the key information represents the essential characteristics of the whole skill service, a second preset rule is entered after the matching degree of the predicted key information and the standard key information is obtained through calculation to determine a second score, whether the execution result of the voice request of the user belongs to the corresponding skill service can be quickly judged according to the second score, the processing logic is clear, the data processing efficiency is high, and the method is a powerful basis for detecting the accuracy rate of the semantic analysis result.
In further embodiments, the step S4 includes the steps of:
s41, substituting the first score and the second score of each skill service into a preset algorithm, and calculating corresponding stage scores;
and S42, obtaining the stage scores in each skill service preset time period for statistical analysis, and calculating the actual scores.
According to the scheme, after the first score and the second score are integrated and substituted into the preset algorithm to calculate the corresponding stage score, the stage score in each skill service preset time period is obtained for statistical analysis, the actual score is calculated, continuous evaluation and analysis can be performed in long-term operation, the evaluation accuracy is improved, and the voice optimization execution effect is further improved.
In further embodiments, the step S5 includes:
s51, acquiring actual scores of all skill services, and sequencing;
and S52, determining the priority of each corresponding skill service according to the sequencing sequence, obtaining a priority list and generating a corresponding execution strategy.
According to the technical scheme, the semantic result information and the actual scores of all skill services are calculated, sequencing and priority setting are carried out to generate the corresponding execution strategy, the skill service with the highest accuracy can be distributed to the semantic result information, so that the skill scheduling efficiency is improved, delay is reduced, and the user experience is improved.
In further embodiments, in said step S41: the preset algorithm is an averaging algorithm or a weight distribution algorithm.
In a further embodiment, the first preset rule is: the matching degree of the semantic result information and the evaluation template is in direct proportion to the score of the first score; the second preset rule is as follows: the matching degree of the semantic result information and the training corpus is in direct proportion to the score of the second score.
The invention also provides a skill optimization processing system based on the semantic platform, which comprises the semantic platform, a template evaluation module and a model evaluation module which are connected with the semantic platform, and a strategy generation module which is connected with the template evaluation module and the model evaluation module;
the semantic platform is used for analyzing or reasoning the voice request of the user according to the algorithm and the model, generating semantic result information and sending the semantic result information to the template evaluation module and the model evaluation module;
the template evaluation module is used for manufacturing a corresponding evaluation template according to the analysis rule of each skill service, matching the evaluation template with the semantic result information and determining a first score according to the matching degree;
the model evaluation module is used for establishing a training model according to historical semantic result data of each skill service, predicting the prediction key information corresponding to the semantic result information according to the training model, matching with each skill service standard key information and determining a second score according to the matching degree;
and the strategy generation module calculates the actual score of each skill service according to the first score and the second score, and further generates an execution strategy.
In a further embodiment, the evaluation template is in a fixed format extracted from semantic parsing cases corresponding to the skill service; the training corpus comprises key information related to the skill service field and summarized from historical semantic result data corresponding to the skill service.
Drawings
Fig. 1 is a work flow chart of a skill optimization processing method based on a semantic platform according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of semantic result information provided by an embodiment of the invention;
FIG. 3 is a schematic diagram of an evaluation template provided by an embodiment of the present invention;
fig. 4 is a skill optimization processing system based on a semantic platform according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention will be described in detail below with reference to the accompanying drawings, which are given solely for the purpose of illustration and are not to be construed as limitations of the invention, including the drawings which are incorporated herein by reference and for illustration only and are not to be construed as limitations of the invention, since many variations thereof are possible without departing from the spirit and scope of the invention.
Example 1
As shown in fig. 1, the skill optimization processing method based on the semantic platform according to the embodiment of the present invention includes steps S1 to S5:
s1, obtaining semantic result information generated by the semantic platform according to the voice request, including steps S11-S12:
s11, acquiring a voice request of a user and sending the voice request to a semantic platform;
and S12, the semantic platform analyzes or infers the voice request according to a default algorithm and a default model to generate semantic result information and sends the semantic result information to the client.
The semantic result information output by the semantic platform in this embodiment is combined together in the JSON format and sent to the client, and the result generated by taking "how much the weather is today" as an example semantic platform is shown in fig. 2.
According to the technical scheme, the semantic platform is improved to recognize the voice request of the user, standardized semantic result information is obtained, and comparison and analysis of subsequent skill service evaluation are facilitated.
S2, according to the analysis rule of each skill service, making a corresponding evaluation template, matching the evaluation template with semantic result information, and determining a first score according to the matching degree, wherein the method comprises the following steps of S21-S23:
s21, obtaining semantic analysis cases obtained by each skill service according to the analysis rules;
s22, counting the fixed format in the semantic parsing case, and integrating to obtain an evaluation template;
and S23, calculating the matching degree of the semantic result information and each skill service evaluation template, and further substituting the matching degree into a first preset rule to determine a first score.
In this embodiment, the first preset rule is: the matching degree of the semantic result information and the evaluation template is in direct proportion to the score of the first score.
Specifically, the semantic result information of the same domain and the same intent has substantially the same format (e.g., JSON format).
As can be seen from FIG. 2, in this example, the semantic information domain is weather, intent is query, datatime is today, city is metropolis. I.e., the user entered the intent to query the weather of the metropolis today. The semantic result information is embedded with a data [ ] field for storing weather results. Therefore, even if the user's utterance is transformed but the parsed semantic domain and intent are not changed, the generated result will be substantially the same as the above example. For example, "i want to know weather of the day in Beijing", though the semantic result nature analyzed finally is still to inquire weather conditions, only time, place, and weather conditions are changed, but the format of the result is not changed. By analogy, the formats of the final generated results are similar as long as other domains meet the above conditions.
Therefore, referring to fig. 3, after the specific content in the semantic result information is removed, the remaining format is the evaluation template.
In this embodiment, based on the uniformity of output data of actual skill services, a fixed format is counted and integrated as an evaluation template in a semantic analysis case obtained according to an analysis rule, the calculated matching degree of semantic result information and the evaluation template (i.e., the matching degree of the format of the calculated semantic result information and the evaluation template) is directly substituted into a first preset rule to obtain a first score, whether a voice request of a user belongs to a corresponding skill service can be quickly judged according to the first score, processing logic is clear, data processing efficiency is high, and the method is a powerful basis for detecting the accuracy of a semantic analysis result.
S3, establishing a training model according to historical semantic result data of each skill service, predicting key information corresponding to the semantic result information according to the training model, matching the predicted key information with key information of each skill service standard, and determining a second score according to the matching degree, wherein the method comprises the following steps of S31-S33:
s31, acquiring historical semantic result data of each skill service, and summarizing corresponding key information according to the corresponding field;
s32, marking all key information, integrating to obtain a training corpus, and performing deep learning model training to obtain a training model;
and S33, substituting the semantic result information into the training model to obtain predicted key information, calculating the matching degree of the predicted key information and the standard key information of each skill service, and further substituting the predicted key information into a second preset rule to determine a second score.
In this embodiment, the second preset rule is: the matching degree of the semantic result information and the training corpus is in direct proportion to the score of the second score.
Specifically, taking a weather query as an example, the result of querying weather, which can be analyzed from semantic result information, generally contains some key information, such as "Chengdu" indicating a place, because the weather query must have an explicit location. "temperature" directly reflects an important indicator of weather conditions. And so on, including wind direction, PM2.5, and so on, which are similarly weather-related information, while these are very few important information that may be present in other domains. Therefore, the place, the temperature, the wind direction, the PM2.5 and the like belong to the key information of the weather skill, the vocabulary set obtained by integrating the place, the temperature, the wind direction, the PM2.5 and the like is the training corpus of the weather skill, and the training corpus is subjected to deep learning model training to obtain the training model.
Before evaluation, corresponding standard key information is defined in advance according to output data (such as semantic result information data) of each skill service.
During evaluation, the semantic result information is substituted into the training model to obtain the predicted key information, the predicted key information is compared with the standard key information (namely, the matching degree is obtained), and the second score is determined according to the similarity of the predicted key information and the standard key information.
Starting from the actual content of the semantic analysis result, the embodiment obtains the historical semantic result data of each skill service, induces corresponding key information according to the corresponding field, labels and integrates to obtain a training corpus, performs deep learning model training to obtain a training model, represents the essential characteristics of the whole skill service by using the key information, enters a second preset rule to determine a second score after calculating the matching degree of the predicted key information and the standard key information, and can quickly judge whether the execution result of the voice request of the user belongs to the corresponding skill service according to the second score.
S4, calculating the actual score of each skill service according to the first score and the second score, comprising the steps of S41-S42:
and S41, substituting the first score and the second score of each skill service into a preset algorithm, and calculating the corresponding stage score.
In the present embodiment, the preset algorithm includes, but is not limited to, an averaging and weight assignment algorithm.
And S42, obtaining the stage scores in each skill service preset time period for statistical analysis, and calculating the actual scores.
Similarly, when performing statistical analysis, the stage scores within a preset time period may also be processed in a manner of averaging or weight assignment to calculate the actual scores.
In the embodiment, after the first score and the second score are integrated and substituted into the preset algorithm to calculate the corresponding stage score, the stage score in each skill service preset time period is obtained for statistical analysis, the actual score is calculated, continuous evaluation and analysis can be performed in long-term operation, the evaluation accuracy is improved, and the voice optimization execution effect is further improved.
S5, generating an execution strategy according to the actual scores of all skill services, comprising the steps S51-S52:
s51, acquiring actual scores of all skill services, and sequencing;
and S52, determining the priority of each skill service according to the sequencing sequence, obtaining a priority list and generating a corresponding execution strategy.
Specifically, all skill services are sorted in the order of the actual scores from large to small to obtain a priority list, and the skill services are executed according to the priority order of the priority list, that is, an execution strategy.
According to the skill scheduling method and the skill scheduling system, the actual scores of the semantic result information and all skill services are calculated, sequencing and priority setting are carried out to generate the corresponding execution strategy, the skill service with the highest accuracy can be distributed to the semantic result information, the skill scheduling efficiency is improved, delay is reduced, and user experience is improved.
The embodiment of the invention prepares a corresponding evaluation template in advance according to the analysis rule of each skill service, acquires historical semantic result data of each skill service, performs word analysis to obtain a training corpus, performs deep learning model training to obtain a training model, so that after semantic result information generated from a semantic platform is acquired, the semantic result information is automatically and respectively matched with the evaluation template (matched with the format of the semantic result information) and standard key information (matched with prediction key information) of each skill service to obtain a first score and a second score, further calculates the actual score of each skill service (namely, automatically detects the correctness of the semantic analysis result), finally generates a sequential execution strategy according to the scores of the actual scores of all the skill services, realizes the purposes of automatic evaluation, automatic strategy generation and automatic optimization of a system, and improves the recognition accuracy of intelligent voice, saving a large amount of manpower and time, optimizing economic benefits and reducing the operation cost.
Example 2
The reference numerals included in the drawings of the present embodiment include: the system comprises a semantic platform 1, a template evaluation module 2, a model evaluation module 3 and a strategy generation module 4.
The embodiment of the invention also provides a skill optimization processing system based on the semantic platform 1, which is shown in fig. 4 and comprises the semantic platform 1, a template evaluation module 2 and a model evaluation module 3 which are connected with the semantic platform 1, and a strategy generation module 4 which is connected with the template evaluation module 2 and the model evaluation module 3;
the semantic platform 1 is used for analyzing or reasoning the voice request of the user according to the algorithm and the model, generating semantic result information and sending the semantic result information to the template evaluation module 2 and the model evaluation module 3 of the client;
the template evaluation module 2 is used for making a corresponding evaluation template according to the analysis rule of each skill service, matching the evaluation template with semantic result information and determining a first score according to the matching degree;
the model evaluation module 3 is used for establishing a training model according to historical semantic result data of each skill service, predicting key information corresponding to the semantic result information according to the training model, matching the predicted key information with key information of each skill service standard, and determining a second score according to the matching degree;
the strategy generation module 4 calculates the actual score of each skill service according to the first score and the second score, and further generates an execution strategy.
In a further embodiment, the evaluation template is in a fixed format extracted from semantic parsing cases corresponding to the skill service; the training corpus comprises key information related to the skill service field and summarized from historical semantic result data corresponding to the skill service.
The processing system provided by this embodiment adopts each module to implement each step in the processing method, and provides a hardware basis for the processing method, thereby facilitating implementation of the method.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (10)
1. A skill optimization processing method based on a semantic platform is characterized by comprising the following steps:
s1, obtaining semantic result information generated by the semantic platform according to the voice request;
s2, manufacturing a corresponding evaluation template according to the analysis rule of each skill service, matching the evaluation template with the semantic result information, and determining a first score according to the matching degree;
s3, establishing a training model according to historical semantic result data of each skill service, predicting key information corresponding to the semantic result information according to the training model, matching with key information of each skill service standard, and determining a second score according to the matching degree;
s4, calculating the actual score of each skill service according to the first score and the second score;
and S5, generating an execution strategy according to the actual scores of all the skill services.
2. The skill optimization processing method based on semantic platform as claimed in claim 1, wherein said step S1 includes:
s11, acquiring a voice request of a user and sending the voice request to a semantic platform;
and S12, the semantic platform analyzes or infers the voice request according to a default algorithm and a default model to generate semantic result information and sends the semantic result information to the client.
3. The skill optimization processing method based on semantic platform as claimed in claim 1, wherein said step S2 includes:
s21, obtaining semantic analysis cases obtained by each skill service according to the analysis rules;
s22, counting the fixed format in the semantic analysis case, and integrating to obtain an evaluation template;
s23, calculating the matching degree of the semantic result information and each skill service evaluation template, and substituting the matching degree into a first preset rule to determine a first score.
4. The skill optimization processing method based on semantic platform as claimed in claim 3, wherein the step S3 includes:
s31, acquiring historical semantic result data of each skill service, and summarizing corresponding key information according to the corresponding field;
s32, marking all the key information, integrating to obtain a training corpus, and performing deep learning model training to obtain a training model;
and S33, substituting the semantic result information into the training model to obtain predicted key information, calculating the matching degree of the predicted key information and the standard key information of each skill service, and further substituting a second preset rule to determine a second score.
5. The skill optimization processing method based on semantic platform as claimed in claim 1, wherein said step S4 includes the steps of:
s41, substituting the first score and the second score of each skill service into a preset algorithm, and calculating corresponding stage scores;
and S42, obtaining the stage scores in each skill service preset time period for statistical analysis, and calculating the actual scores.
6. The skill optimization processing method based on semantic platform as claimed in claim 1, wherein said step S5 includes:
s51, acquiring actual scores of all skill services, and sequencing;
and S52, determining the priority of each corresponding skill service according to the sequencing sequence, obtaining a priority list and generating a corresponding execution strategy.
7. The semantic platform-based skill optimization processing method according to claim 5, wherein in the step S41: the preset algorithm is an averaging algorithm or a weight distribution algorithm.
8. The skill optimization processing method based on the semantic platform as claimed in claim 4, wherein the first preset rule is: the matching degree of the semantic result information and the evaluation template is in direct proportion to the score of the first score; the second preset rule is as follows: the matching degree of the semantic result information and the training corpus is in direct proportion to the score of the second score.
9. A skill optimization processing system based on a semantic platform is characterized in that: the system comprises a semantic platform, a template evaluation module and a model evaluation module which are connected with the semantic platform, and a strategy generation module which is connected with the template evaluation module and the model evaluation module;
the semantic platform is used for analyzing or reasoning the voice request of the user according to the algorithm and the model, generating semantic result information and sending the semantic result information to the template evaluation module and the model evaluation module;
the template evaluation module is used for manufacturing a corresponding evaluation template according to the analysis rule of each skill service, matching the evaluation template with the semantic result information and determining a first score according to the matching degree;
the model evaluation module establishes a training model according to historical semantic result data of each skill service, predicts prediction key information corresponding to the semantic result information according to the training model, matches the prediction key information with each skill service standard key information, and determines a second score according to the matching degree;
and the strategy generation module calculates the actual score of each skill service according to the first score and the second score, and further generates an execution strategy.
10. The semantic platform based skill optimization processing system of claim 9, wherein: the evaluation template is in a fixed format extracted from the semantic analysis case corresponding to the skill service; the training corpus comprises key information related to the skill service field and summarized from historical semantic result data corresponding to the skill service.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111066358.0A CN113935309A (en) | 2021-09-13 | 2021-09-13 | Skill optimization processing method and system based on semantic platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111066358.0A CN113935309A (en) | 2021-09-13 | 2021-09-13 | Skill optimization processing method and system based on semantic platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113935309A true CN113935309A (en) | 2022-01-14 |
Family
ID=79275499
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111066358.0A Pending CN113935309A (en) | 2021-09-13 | 2021-09-13 | Skill optimization processing method and system based on semantic platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113935309A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116303978A (en) * | 2023-05-17 | 2023-06-23 | 福建博士通信息股份有限公司 | Potential user mining method based on voice analysis |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108829757A (en) * | 2018-05-28 | 2018-11-16 | 广州麦优网络科技有限公司 | A kind of intelligent Service method, server and the storage medium of chat robots |
| CN110209791A (en) * | 2019-06-12 | 2019-09-06 | 百融云创科技股份有限公司 | It is a kind of to take turns dialogue intelligent speech interactive system and device more |
| CN110334201A (en) * | 2019-07-18 | 2019-10-15 | 中国工商银行股份有限公司 | A kind of intension recognizing method, apparatus and system |
| CN111090730A (en) * | 2019-12-05 | 2020-05-01 | 中科数智(北京)科技有限公司 | Intelligent voice scheduling system and method |
| CN112052316A (en) * | 2020-08-12 | 2020-12-08 | 深圳市欢太科技有限公司 | Model evaluation method, model evaluation device, storage medium and electronic equipment |
| CN112131890A (en) * | 2020-09-15 | 2020-12-25 | 北京慧辰资道资讯股份有限公司 | Method, device and equipment for constructing intelligent recognition model of conversation intention |
| WO2021129240A1 (en) * | 2019-12-26 | 2021-07-01 | 思必驰科技股份有限公司 | Skill scheduling method and apparatus for voice conversation platform |
| CN113205817A (en) * | 2021-07-06 | 2021-08-03 | 明品云(北京)数据科技有限公司 | Speech semantic recognition method, system, device and medium |
-
2021
- 2021-09-13 CN CN202111066358.0A patent/CN113935309A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108829757A (en) * | 2018-05-28 | 2018-11-16 | 广州麦优网络科技有限公司 | A kind of intelligent Service method, server and the storage medium of chat robots |
| CN110209791A (en) * | 2019-06-12 | 2019-09-06 | 百融云创科技股份有限公司 | It is a kind of to take turns dialogue intelligent speech interactive system and device more |
| CN110334201A (en) * | 2019-07-18 | 2019-10-15 | 中国工商银行股份有限公司 | A kind of intension recognizing method, apparatus and system |
| CN111090730A (en) * | 2019-12-05 | 2020-05-01 | 中科数智(北京)科技有限公司 | Intelligent voice scheduling system and method |
| WO2021129240A1 (en) * | 2019-12-26 | 2021-07-01 | 思必驰科技股份有限公司 | Skill scheduling method and apparatus for voice conversation platform |
| CN112052316A (en) * | 2020-08-12 | 2020-12-08 | 深圳市欢太科技有限公司 | Model evaluation method, model evaluation device, storage medium and electronic equipment |
| CN112131890A (en) * | 2020-09-15 | 2020-12-25 | 北京慧辰资道资讯股份有限公司 | Method, device and equipment for constructing intelligent recognition model of conversation intention |
| CN113205817A (en) * | 2021-07-06 | 2021-08-03 | 明品云(北京)数据科技有限公司 | Speech semantic recognition method, system, device and medium |
Non-Patent Citations (2)
| Title |
|---|
| 李潇;闵华松;林云汉;: "一种用于CBR推理机的案例学习算法研究", 计算机应用研究, no. 12, 12 December 2017 (2017-12-12) * |
| 王堃;林民;李艳玲;: "端到端对话系统意图语义槽联合识别研究综述", 计算机工程与应用, no. 14 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116303978A (en) * | 2023-05-17 | 2023-06-23 | 福建博士通信息股份有限公司 | Potential user mining method based on voice analysis |
| CN116303978B (en) * | 2023-05-17 | 2023-08-15 | 福建博士通信息股份有限公司 | Potential user mining method based on voice analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108304372B (en) | Entity extraction method and device, computer equipment and storage medium | |
| CN111341305B (en) | Audio data labeling method, device and system | |
| WO2013080406A1 (en) | Dialog system, redundant message removal method and redundant message removal program | |
| WO2021109690A1 (en) | Multi-type question smart answering method, system and device, and readable storage medium | |
| CN108549628B (en) | Sentence-breaking device and method for stream type natural language information | |
| US20160284344A1 (en) | Speech data recognition method, apparatus, and server for distinguishing regional accent | |
| CN105354199B (en) | A kind of recognition methods of entity meaning and system based on scene information | |
| CN105027198A (en) | Speech recognition system and speech recognition device | |
| CN112163424A (en) | Data labeling method, device, equipment and medium | |
| CN114840671A (en) | Dialogue generation method, model training method, device, equipment and medium | |
| CN111554276B (en) | Speech recognition method, device, equipment and computer readable storage medium | |
| CN112151014A (en) | Method, device and equipment for evaluating voice recognition result and storage medium | |
| US20210104235A1 (en) | Arbitration of Natural Language Understanding Applications | |
| CN111611355A (en) | Dialog reply method, device, server and storage medium | |
| CN116244410A (en) | Index data analysis method and system based on knowledge graph and natural language | |
| CN111554275A (en) | Speech recognition method, device, equipment and computer readable storage medium | |
| CN113935309A (en) | Skill optimization processing method and system based on semantic platform | |
| CN111369294A (en) | Software cost estimation method and device | |
| CN110750626B (en) | Scene-based task-driven multi-turn dialogue method and system | |
| CN113591463A (en) | Intention recognition method and device, electronic equipment and storage medium | |
| CN112463967A (en) | Emotion early warning method, system, equipment and storage medium | |
| CN112669850A (en) | Voice quality detection method and device, computer equipment and storage medium | |
| CN111177331A (en) | Dialog intention recognition method and device | |
| JP2020187282A (en) | Information processing device, information processing method, and program | |
| CN115691503A (en) | Voice recognition method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |