CN114021576A - Text-based natural language understanding decision method - Google Patents
Text-based natural language understanding decision method Download PDFInfo
- Publication number
- CN114021576A CN114021576A CN202111257879.4A CN202111257879A CN114021576A CN 114021576 A CN114021576 A CN 114021576A CN 202111257879 A CN202111257879 A CN 202111257879A CN 114021576 A CN114021576 A CN 114021576A
- Authority
- CN
- China
- Prior art keywords
- semantic
- cache
- semantic analysis
- analysis
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 19
- 238000004458 analytical method Methods 0.000 claims abstract description 89
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000004141 dimensional analysis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- 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
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/216—Parsing using statistical methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Machine Translation (AREA)
Abstract
The invention provides a text-based natural language understanding decision method, which comprises the following steps of: establishing semantic cache; acquiring voice information of a user, and performing semantic analysis; and feeding back the optimal semantic analysis result to the user, and writing the semantic analysis result irrelevant to the context into a semantic cache. And obtaining cache semantics according to the received requested text statement, and returning the cache semantics if corresponding cache is matched. And if no matched cache exists, performing semantic analysis. And after the semantic analysis result is obtained, returning the optimal semantic analysis result on one side, and judging the write cache operation on the other side. The semantic meaning is clear, the semantic meaning irrelevant to the context can be written into the cache, so that the analysis result in the cache can be directly obtained and returned when the same semantic request comes next time, and the analysis efficiency is improved.
Description
Technical Field
The invention relates to the technical field of self-language understanding, in particular to a text-based natural language understanding decision method.
Background
With the rise of artificial intelligence, natural language understanding is an important direction in the field of artificial intelligence, and the theory and method of human and computer communication through natural language are mainly studied. At present, a smart television is popular when being operated by voice, and users can meet different requirements such as weather checking, video watching, music listening and the like through voice interaction. The process is to recognize the voice data to obtain corresponding text data, understand the natural language according to the text and feed back the understood result. Because Chinese is complex, a single semantic understanding mode cannot meet requirements, so that semantic understanding can be performed in multiple modes, semantic understanding results in different modes may be different, and analysis needs to be performed through a certain strategy if understanding which best meets user requirements is selected from multiple semantic understandings.
Disclosure of Invention
The invention aims to provide a text-based natural language understanding decision method. So as to solve the technical problems existing in the background technology.
In order to achieve the purpose, the invention adopts the following technical scheme:
a text-based natural language understanding decision method, comprising the steps of:
establishing semantic cache;
acquiring voice information of a user, and performing semantic analysis;
and feeding back the optimal semantic analysis result to the user, and writing the semantic analysis result irrelevant to the context into a semantic cache.
In some embodiments, the semantic cache includes a fixed semantic cache and a real-time semantic cache, where the fixed semantic cache is a semantic analysis result write-in cache of some commonly used statements counted in advance; the real-time semantic cache is the semantic analyzed after the user requests in real time, and the semantic analysis which is high in semantic analysis confidence and not influenced by context is stored in the cache.
In some embodiments, the obtaining voice information of a user and performing semantic parsing includes: semantic analysis is carried out by adopting three modes, wherein a, semantic analysis is carried out by adopting a grammar rule; b. performing semantic analysis by adopting an RNN algorithm; c. semantic parsing is carried out by using an Elastic Search (ES) intelligent search, and the confidence of a parsing result is calculated in each mode.
In some embodiments of the present invention, the,
semantic analysis results are screened, and the results have three conditions: a. the semantic analysis of the three modes is the same; b. the semantic analysis of the two modes is the same; c. the semantic analysis of the three modes is different; a and b directly select the result with the same semantic resolution, and c needs to be selected through three dimensions.
In some embodiments, the dimensions are, respectively, the current application of the television, the confidence level, the user profile.
Advantageous effects
Through the text-based natural language understanding decision method provided by the invention, the user experience is improved.
And obtaining cache semantics according to the received requested text statement, and returning the cache semantics if corresponding cache is matched. And if no matched cache exists, performing semantic analysis. And after the semantic analysis result is obtained, returning the optimal semantic analysis result on one side, and judging the write cache operation on the other side. The semantic meaning is clear, the semantic meaning irrelevant to the context can be written into the cache, so that the analysis result in the cache can be directly obtained and returned when the same semantic request comes next time, and the analysis efficiency is improved.
Drawings
FIG. 1 is a flow chart of obtaining semantic parsing.
Fig. 2 is a flow of semantic parsing result selection.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
On the contrary, this application is intended to cover any alternatives, modifications, equivalents, and alternatives that may be included within the spirit and scope of the application as defined by the appended claims. Furthermore, in the following detailed description of the present application, certain specific details are set forth in order to provide a better understanding of the present application. It will be apparent to one skilled in the art that the present application may be practiced without these specific details.
The text-based natural language understanding decision method according to the embodiment of the present application will be described in detail below with reference to fig. 1-2. It is to be noted that the following examples are only for explaining the present application and do not constitute a limitation to the present application.
As shown in fig. 1, the text-based natural language understanding decision method includes the following steps:
establishing semantic cache; the semantic cache is divided into fixed semantic cache and real-time semantic cache. The fixed semantic cache is a cache into which semantic parsing results of some commonly used sentences counted in advance are written. The real-time semantic cache is the semantic analyzed after the user requests in real time, and the semantic analysis which has high semantic analysis confidence and is not influenced by the context is stored in the cache, so that the same statement request can be fed back very quickly. And if the requested statement is matched with the statement in the cache, directly outputting the cache result.
Acquiring voice information of a user, and performing semantic analysis; semantic analysis is carried out by adopting three modes, wherein a, semantic analysis is carried out by adopting a grammar rule; b. performing semantic analysis by adopting an RNN algorithm; c. semantic parsing is carried out by using an Elastic Search (ES) intelligent search, and the confidence of a parsing result is calculated in each mode. The present invention does not describe the method of semantic parsing in detail.
Because different modes are adopted for semantic analysis, analysis results may be inconsistent, semantic analysis results need to be screened, and the results have three conditions: a. the semantic analysis of the three modes is the same; b. the semantic analysis of the two modes is the same; c. the semantic analysis of the three modes is different; a and b directly select the result with the same semantic resolution, and c needs to be selected through three dimensions. For the case of c we can choose through three dimensions.
The dimensions are the current application of the television, the confidence level, and the user profile, respectively. The semantic analysis gives the domain to which the given semantic belongs, and if the domain after the semantic analysis is the same as the currently applied domain, the result is closer to the real semantic. If the request of "play you through the ocean" is that you can be both a tv show and a song, then it can be distinguished according to the current application, if the current application is a video application, the field is selected to be the parsing of the video, if the current application is a music application, the field is selected to be the parsing of the music.
If the analyzed fields are the same, the judgment can be carried out according to the confidence level of semantic analysis, and the preference with high confidence level is given. At this time, if the optimum analysis result is not selected, it can be judged by combining the user image. The user can be portrayed according to historical request data of the user, the request intention of the user is analyzed according to usual habit characteristics of the user, and semantic analysis is selected. At this time, if the result of semantic analysis can not be selected, the result is selected according to the order of grammar, algorithm and search, thereby obtaining the final analysis result.
And feeding back the optimal semantic analysis result to the user, and writing the semantic analysis result irrelevant to the context into a semantic cache. Some semantics of the final result after the semantic analysis are clear, and the semantic analysis irrelevant to the context can be written into the cache, so that the analysis result can be directly obtained and returned to the user when the same semantic request comes next time.
As shown in fig. 1, after receiving a requested text statement, cache semantics are obtained, and if there is a corresponding cache match, the cache semantics are returned. And if no matched cache exists, performing semantic analysis.
Semantic analysis is divided into three modes, namely grammar rule analysis, RNN algorithm analysis and elastic search analysis. And the grammar rule analysis is to define grammar rules through a grammar network and analyze the matching degree of the request text and the grammar rules to carry out semantic analysis. The RNN algorithm analysis is semantic analysis by using a recurrent neural network algorithm. The elastic search parsing is semantic parsing through an elastic search (es) algorithm. The present invention does not describe these three semantic parsing methods in detail. The semantic analysis of the three modes is carried out simultaneously.
After the semantic parsing of the three modes is finished, the parsing result which best accords with the request text is selected. As a result, there are three cases: a. the semantic analysis of the three modes is the same; b. the semantic analysis of the two modes is the same; c. the semantic parsing of the three modes is different. a and b directly select the result with the same semantic resolution, and the case of c needs to be selected through three dimensions. The three dimensions are the current application of the television, the confidence level, and the user profile, respectively. The semantic analysis in the three modes can provide the domain, intention, related attribute, attribute value and confidence degree to which the semantic belongs.
As shown in fig. 2, the flow of selecting the semantic parsing result is as follows:
firstly, judging whether the analyzed field is the same as the field currently applied by the television. If the request of 'playing the Drift sea to see you' is the request, the Drift sea to see you are both TV shows and songs in our database, the method can be distinguished according to the current application, if the current application is a video application, the selected field is the analysis result of the video, and if the current application is a music application, the selected field is the analysis result of the music. And judging the value of the confidence coefficient if the analysis result cannot be distinguished by the current application. And selecting the analysis result with the highest confidence coefficient.
If the analysis result cannot be distinguished by the confidence, the user image is combined for judgment. The user portrait is used for analyzing the usual habit characteristics of the user according to the historical request data of the user and portraying the user. If "play floating ocean and see you", the current television application is neither a video application nor a music application, and the confidence degrees of the two analysis modes are the same, then see the user portrait. If the user's image is favorite music, the analysis result of the music field is selected.
And if the best analysis result cannot be distinguished through three-dimensional analysis, selecting the optimal analysis result according to the grammar rule, the RNN algorithm and the elastic search sequence, and obtaining the final analysis result. Since the accuracy of semantic parsing is, from the top to the bottom, grammar rules, RNN algorithms, elastic search searches, according to a large number of test data analyses.
And after the semantic analysis result is obtained, returning the optimal semantic analysis result on one side, and judging the write cache operation on the other side. The semantic meaning is clear, the semantic meaning irrelevant to the context can be written into the cache, so that the analysis result in the cache can be directly obtained and returned when the same semantic request comes next time, and the analysis efficiency is improved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (5)
1. A text-based natural language understanding decision method is characterized by comprising the following steps:
establishing semantic cache;
acquiring voice information of a user, and performing semantic analysis;
and feeding back the optimal semantic analysis result to the user, and writing the semantic analysis result irrelevant to the context into a semantic cache.
2. The text-based natural language understanding decision method according to claim 1, wherein the semantic cache comprises a fixed semantic cache and a real-time semantic cache, wherein the fixed semantic cache is a semantic parsing result write cache of some commonly used sentences counted in advance; the real-time semantic cache is the semantic analyzed after the user requests in real time, and the semantic analysis which is high in semantic analysis confidence and not influenced by context is stored in the cache.
3. The text-based natural language understanding decision method according to claim 1, wherein the obtaining of the speech information of the user and the semantic parsing comprise: semantic analysis is carried out by adopting three modes, wherein a, semantic analysis is carried out by adopting a grammar rule; b. performing semantic analysis by adopting an RNN algorithm; c. semantic parsing is carried out by using an Elastic Search (ES) intelligent search, and the confidence of a parsing result is calculated in each mode.
4. The text-based natural language understanding decision method according to claim 3, wherein semantic parsing results need to be filtered, and the results have three conditions: a. the semantic analysis of the three modes is the same; b. the semantic analysis of the two modes is the same; c. the semantic analysis of the three modes is different; a and b directly select the result with the same semantic resolution, and c needs to be selected through three dimensions.
5. The method of claim 4, wherein the dimensions are current application, confidence level, and user profile of the TV.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111257879.4A CN114021576A (en) | 2021-10-27 | 2021-10-27 | Text-based natural language understanding decision method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111257879.4A CN114021576A (en) | 2021-10-27 | 2021-10-27 | Text-based natural language understanding decision method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114021576A true CN114021576A (en) | 2022-02-08 |
Family
ID=80058286
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111257879.4A Pending CN114021576A (en) | 2021-10-27 | 2021-10-27 | Text-based natural language understanding decision method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114021576A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104360994A (en) * | 2014-12-04 | 2015-02-18 | 科大讯飞股份有限公司 | Natural language understanding method and natural language understanding system |
| 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 |
| CN109933773A (en) * | 2017-12-15 | 2019-06-25 | 上海擎语信息科技有限公司 | A kind of multiple semantic sentence analysis system and method |
| CN110473540A (en) * | 2019-08-29 | 2019-11-19 | 京东方科技集团股份有限公司 | Voice interactive method and system, terminal device, computer equipment and medium |
-
2021
- 2021-10-27 CN CN202111257879.4A patent/CN114021576A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104360994A (en) * | 2014-12-04 | 2015-02-18 | 科大讯飞股份有限公司 | Natural language understanding method and natural language understanding system |
| CN109933773A (en) * | 2017-12-15 | 2019-06-25 | 上海擎语信息科技有限公司 | A kind of multiple semantic sentence analysis system and method |
| 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 |
| CN110473540A (en) * | 2019-08-29 | 2019-11-19 | 京东方科技集团股份有限公司 | Voice interactive method and system, terminal device, computer equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108460396B (en) | Negative sampling method and device | |
| CN102831177B (en) | Statement error correction and system thereof | |
| US8392414B2 (en) | Hybrid audio-visual categorization system and method | |
| US9384196B2 (en) | Signature generation for multimedia deep-content-classification by a large-scale matching system and method thereof | |
| CN109241524A (en) | Semantic analysis method and device, computer readable storage medium, electronic equipment | |
| US20230031591A1 (en) | Methods and apparatus to facilitate generation of database queries | |
| CN109902289B (en) | News video theme segmentation method oriented to fuzzy text mining | |
| KR100676863B1 (en) | System and method for providing music search service | |
| CN107247768A (en) | Method for ordering song by voice, device, terminal and storage medium | |
| CN114154570A (en) | Sample screening method and system and neural network model training method | |
| CN108153875B (en) | Corpus processing method and device, intelligent sound box and storage medium | |
| Li et al. | Learning visual styles from audio-visual associations | |
| CN113506553A (en) | Audio automatic labeling method based on transfer learning | |
| CN113345407A (en) | Style speech synthesis method and device, electronic equipment and storage medium | |
| CN111523430B (en) | Customizable interactive video production method and device based on UCL | |
| CN117035078A (en) | Multi-mode knowledge graph unified representation learning framework | |
| CN114970538A (en) | Text error correction method and device | |
| CN114021576A (en) | Text-based natural language understanding decision method | |
| CN112445902A (en) | Method for identifying user intention in multi-turn conversation and related equipment | |
| US20230027078A1 (en) | Systems and methods for generation and deployment of a human-personified virtual agent using pre-trained machine learning-based language models and a video response corpus | |
| CN116070020A (en) | Food material recommendation method, equipment and storage medium based on knowledge graph | |
| CN114239545A (en) | Natural language generation method, device, equipment, medium and product | |
| CN111767377B (en) | Efficient spoken language understanding and identifying method oriented to low-resource environment | |
| CN112820274B (en) | Voice information recognition correction method and system | |
| CN118093791B (en) | AI knowledge base generation method and system combined with cloud computing |
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 |