CN112380356A - Method, device, electronic equipment and medium for constructing catering knowledge graph - Google Patents
Method, device, electronic equipment and medium for constructing catering knowledge graph Download PDFInfo
- Publication number
- CN112380356A CN112380356A CN202011377278.2A CN202011377278A CN112380356A CN 112380356 A CN112380356 A CN 112380356A CN 202011377278 A CN202011377278 A CN 202011377278A CN 112380356 A CN112380356 A CN 112380356A
- Authority
- CN
- China
- Prior art keywords
- recipe
- data
- entity
- predicate
- knowledge graph
- 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 61
- 235000013305 food Nutrition 0.000 claims abstract description 120
- 235000015097 nutrients Nutrition 0.000 claims abstract description 85
- 239000000463 material Substances 0.000 claims abstract description 71
- 238000003860 storage Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims description 46
- 235000012054 meals Nutrition 0.000 claims description 38
- 230000004927 fusion Effects 0.000 claims description 19
- 230000015654 memory Effects 0.000 claims description 19
- 238000000605 extraction Methods 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 11
- 238000005065 mining Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 9
- 238000000265 homogenisation Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 abstract description 9
- 238000003058 natural language processing Methods 0.000 abstract description 4
- 235000016709 nutrition Nutrition 0.000 description 9
- 238000002360 preparation method Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 230000035764 nutrition Effects 0.000 description 6
- 241000227653 Lycopersicon Species 0.000 description 4
- 235000007688 Lycopersicon esculentum Nutrition 0.000 description 4
- 238000004590 computer program Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 235000013601 eggs Nutrition 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000005520 cutting process Methods 0.000 description 3
- 238000010411 cooking Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000004615 ingredient Substances 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 241001347978 Major minor Species 0.000 description 1
- 244000061456 Solanum tuberosum Species 0.000 description 1
- 235000002595 Solanum tuberosum Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 235000015278 beef Nutrition 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000004280 healthy diet Nutrition 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 235000005955 light diet Nutrition 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 235000004213 low-fat Nutrition 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 235000021048 nutrient requirements Nutrition 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- 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
-
- 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
- G06F16/338—Presentation of query results
-
- 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
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The embodiment of the application discloses a method and a device for constructing a catering knowledge graph, electronic equipment and a computer-readable storage medium, and relates to the technical field of knowledge graphs, natural language processing and cloud services. One embodiment of the method comprises: acquiring recipe data and nutrient data corresponding to food materials contained in the recipe data; the method comprises the steps that unstructured recipe data are segmented into a plurality of statements, an Ernie-MRC model is utilized to extract a main and predicate triple taking a recipe name as an entity from at least one statement, and the entity is used as a subject of the main and predicate triple; and constructing an initial food matching knowledge graph according to the main and predicate triple, and establishing connection between each food material in the initial food matching knowledge graph and corresponding nutrient data to obtain a target food matching knowledge graph. The embodiment properly processes unstructured recipe data, and combines nutrient data to construct a more comprehensive catering knowledge graph, thereby providing a good data base for subsequent query operation.
Description
Technical Field
The present application relates to the technical field of artificial intelligence, and in particular, to the technical field of a knowledge graph, natural language processing, and cloud service, and in particular, to a method, an apparatus, an electronic device, and a computer-readable storage medium for constructing a catering knowledge graph.
Background
With the rapid development of productivity, people are no longer in the times of extreme scarcity of substances, and are turning to continuously pursue healthy lifestyles, especially healthy diets, i.e. the demand of people for nutritional knowledge and nutritional meals is increasing at a rapid pace.
Disclosure of Invention
The embodiment of the application provides a method and a device for constructing a catering knowledge graph, electronic equipment and a computer-readable storage medium.
In a first aspect, an embodiment of the present application provides a method for constructing a meal distribution knowledge graph, including: acquiring recipe data and nutrient data corresponding to food materials contained in the recipe data; the method comprises the steps that unstructured recipe data are segmented into a plurality of statements, and an Ernie-MRC model is utilized to extract a principal and predicate object triple taking a recipe name as an entity from at least one statement; wherein, the entity is used as a subject of the main predicate element triple; and constructing an initial food matching knowledge graph according to the main and predicate triple, and establishing connection between each food material in the initial food matching knowledge graph and corresponding nutrient data to obtain a target food matching knowledge graph.
In a second aspect, an embodiment of the present application provides an apparatus for constructing a meal assembly knowledge graph, including: a recipe data and nutrient data acquisition unit configured to acquire recipe data and nutrient data corresponding to food materials included in the recipe data; the unstructured recipe data processing unit is configured to divide unstructured recipe data into a plurality of statements, and extract a predicate triplet taking a recipe name as an entity from at least one statement by using an Ernie-MRC model; wherein, the entity is used as a subject of the main predicate element triple; and the food matching knowledge map construction unit is configured to construct an initial food matching knowledge map according to the main-predicate triple, and establish connection between each food material in the initial food matching knowledge map and corresponding nutrient data to obtain a target food matching knowledge map.
In a third aspect, an embodiment of the present application provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to perform the method for constructing a recipe knowledge graph as described in any one of the implementations of the first aspect when executed.
In a fourth aspect, the present application provides a non-transitory computer-readable storage medium storing computer instructions for enabling a computer to implement the method for constructing a meal knowledge graph as described in any implementation manner of the first aspect when executed.
According to the method, the device, the electronic equipment and the computer-readable storage medium for constructing the catering knowledge graph, the recipe data and the nutrient data corresponding to the food materials contained in the recipe data are firstly obtained; then, cutting unstructured recipe data into a plurality of sentences, and extracting a main predicate triple taking a recipe name as an entity from at least one sentence by utilizing an Ernie-MRC model, wherein the entity is used as a subject of the main predicate triple; and finally, constructing an initial food matching knowledge graph according to the main and predicate triple, and establishing connection between each food material in the initial food matching knowledge graph and corresponding nutrient data to obtain a target food matching knowledge graph.
In order to meet the current demand on professional nutrition catering knowledge, the scheme for constructing the catering knowledge map is provided, an implementation mode that accurate subject-predicate-object triples are extracted from most of recipe data existing in an unstructured form by segmenting the recipe data and then using an Ernie-MRC model is provided, unstructured recipe data can be well processed, and meanwhile, the food material and nutrient data are constructed to enable the finally constructed target catering knowledge map to simultaneously contain the recipe, the food material and the nutrient data, so that the scheme is more comprehensive and provides a comprehensive and accurate data base for subsequent query operation of a user.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
FIG. 1 is an exemplary system architecture to which the present application may be applied;
FIG. 2 is a flow chart of a method for constructing a recipe knowledge graph provided by an embodiment of the present application;
FIG. 3 is a flowchart of a method for processing entities according to an embodiment of the present application;
FIG. 4 is a schematic flow chart diagram of a method for constructing and using a catering knowledge graph in an application scenario provided by an embodiment of the application;
FIG. 5 is a block diagram of an apparatus for constructing a recipe knowledge graph according to an embodiment of the present disclosure;
fig. 6 is a schematic structural diagram of an electronic device suitable for executing the method for constructing a recipe knowledge graph according to an embodiment of the present application.
Detailed Description
The present application will be described in further detail with reference to the following drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not restrictive of the invention. It should be noted that, for convenience of description, only the portions related to the related invention are shown in the drawings.
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.
Fig. 1 illustrates an exemplary system architecture 100 to which embodiments of the present methods, apparatuses, electronic devices and computer-readable storage media for building a recipe knowledge graph may be applied.
As shown in fig. 1, the system architecture 100 may include a user terminal 101, a server 102, data sources 103, 104, 105. Data communication between the user terminal 101 and the server 102, and between the server 102 and the data sources 103, 104, 105 can be implemented in various wireless or wired manners according to actual situations.
The data sources 103, 104, 105 are respectively and independently recorded with recipe data and/or nutrient data, specifically, the data sources 103, 104, 105 may be a certain vertical website for recording the recipe data and/or nutrient data, or a paper or electronic tool book for recording the recipe data and/or nutrient data, etc.; the server 102 is configured to collect basic data required for constructing a meal matching knowledge graph from the data sources 103, 104, and 105, arrange the basic data into a meal matching knowledge graph containing recipe data and nutrient data of each food material contained in each recipe according to preset data processing logic, and respond to a received query request for a specified meal matching requirement sent by a user through the user terminal 101 after constructing and obtaining an available meal matching knowledge graph.
The user terminal 101 and the server 102 may be hardware or software. When the user terminal 101 is hardware, it may be various electronic devices including, but not limited to, a smart phone, a tablet computer, a laptop portable computer, a desktop computer, and the like; when the user terminal 101 is software, it may be installed in the electronic device listed above, and it may be implemented as multiple pieces of software or software modules, or may be implemented as a single piece of software or software modules, and is not limited herein. When the server 102 is hardware, it may be implemented as a distributed server cluster formed by multiple servers, or may be implemented as a single server; when the server 102 is software, it may be implemented as a plurality of software or software modules, or may be implemented as a single software or software module, and is not limited in this respect. The data sources 103, 104, and 105 may include a software representation of a site class according to an actual situation, and may also include a hardware representation of a hard disk, a database, and the like, which is not limited herein.
The user terminal 101, the server 102, and the data sources 103, 104, and 105 may each have a corresponding application installed or have a corresponding entity function component added to achieve their respective purposes according to actual situations. For example, the user terminal 101 may be installed with a meal matching demand query application for initiating a query request to the server 102, the server 102 may be installed with a knowledge data processing application for acquiring basic data from the data sources 103, 104, and 105, performing data processing, and constructing a meal matching knowledge graph, and may also be installed with a request processing application for processing the received user query request.
Taking the server 102 as an example, the following effects can be achieved when the knowledge data processing application is run: firstly, acquiring recipe data and nutrient data corresponding to food materials contained in the recipe data from data sources 103, 104 and 105 respectively; then, cutting unstructured recipe data into a plurality of sentences, and extracting a main predicate triple taking a recipe name as an entity from at least one sentence by utilizing an Ernie-MRC model, wherein the entity is used as a subject of the main predicate triple; and finally, constructing an initial food matching knowledge graph according to the main and predicate triple, and establishing connection between each food material in the initial food matching knowledge graph and corresponding nutrient data to obtain a target food matching knowledge graph. That is, the server 102 operates the knowledge data processing application according to the above steps to construct a food preparation knowledge graph, which satisfies the premise of supporting the user to perform query. Next, the server 102 can achieve the following effects when running the request processing application: firstly, receiving a specified catering demand query request transmitted by a user through a user terminal 101; then, calling a pre-constructed catering knowledge graph to inquire out a matching recipe corresponding to the specified catering demand inquiry request; finally, the matching recipe is returned to the user terminal 101 for presentation to the user.
It should be understood that, in the above examples, the knowledge data processing application is used to construct the food preparation knowledge graph, and the request processing application is used to process the user query request, so to ensure that the request processing application can successfully call the food preparation knowledge graph constructed by the knowledge data processing application, the two types of applications should make the setting of the call authority or the setting of the communication interface in advance. Or in practical application, the knowledge data processing application and the request processing application are integrated into a large application as a functional component respectively, so that the data calling problem between different applications is converted into the data calling problem in the same application.
Since the process of constructing the food matching knowledge graph requires complex processing on huge data, the scheme for constructing the food matching knowledge graph is generally executed by the server 102 with strong computing power and more computing resources, and accordingly, the device for constructing the food matching knowledge graph is generally arranged in the server 102. However, it should be understood that once the recipe knowledge graph is constructed, it will not occupy more computing resources when called, so that when the finished recipe knowledge graph is allowed to be stored locally in the user terminal 101, the subsequent query operation can be directly and completely completed locally in the user terminal 101.
It should be understood that the number of user terminals, servers, data sources in fig. 1 is merely illustrative. There may be any number of user terminals, servers, data sources, as desired for implementation.
Referring to fig. 2, fig. 2 is a flowchart of a method for constructing a recipe knowledge graph according to an embodiment of the present application, where the process 200 includes the following steps:
step 201: acquiring recipe data and nutrient data corresponding to food materials contained in the recipe data;
this step is intended to acquire recipe data and nutrient data corresponding to food materials contained in the recipe data by an execution subject (e.g., server 102 shown in fig. 1) of the method for constructing a catering intellectual graph.
Wherein the recipe data and corresponding nutrient data can be obtained from a plurality of data sources (e.g., data sources 103, 104, 105 shown in fig. 1), such as a vertical website, a paper or electronic version of a workbook, a database, an authoritative third party, etc., that specifically records the recipe data and/or nutrient data. Specifically, the recipe data and the corresponding nutrient data may be obtained from the same data source, or may be obtained from different data sources. Further, when the recipe data and the nutrient data are respectively obtained from different data sources, the food materials used in the recipe data are used as connection points matched with the corresponding nutrient data.
Further, the quality of the recipe data and the nutrient data acquired by different data sources is different, in order to avoid increasing difficulty of the low-quality data on subsequent data processing as much as possible, the low-quality information may be screened and removed after the original data is acquired, the criteria for determining the low-quality information may include at least one of discordance of sentences, more wrongly written words, unreasonable typesetting, less total words, no drawings or drawings not corresponding to the text information, and the details of how many wrongly written words are determined to be more, how unreasonable typesetting is performed, and how many words are determined to be less total words (for example, the total words are less than 50) should be flexibly formulated in combination with all possible requirements or influence factors in practical application scenarios, and no specific limitation is made here.
Step 202: the method comprises the steps that unstructured recipe data are segmented into a plurality of statements, and an Ernie-MRC model is utilized to extract a principal and predicate object triple taking a recipe name as an entity from at least one statement;
on the basis of step 201, this step is intended to perform appropriate processing on the acquired unstructured recipe data by the execution subject, so as to successfully extract a predicate-object triple taking the recipe name as an entity from the unstructured recipe data.
It should be understood that the data may be divided into three types, structured, semi-structured, and unstructured, which respectively correspond to whether complete structure information is contained therein, and the structured data, because it contains complete structure information, may be implemented by referring to the structure information when extracting a Predicate triple (also referred to as SPO triple for short, SPO is an abbreviation of a Subject prefix Object in english); the semi-structured data only contains partial structural information, and the partial structural information can generally guide the extraction of the SPO triples, but the accuracy may be slightly low; the unstructured data does not contain structural information at all or cannot identify contained structural information, so that the extraction of the SPO triples cannot be completed by referring to the structural information like the structured information and the semi-structured information, and the recipe data describing the recipe information belongs to a text type only for remembering, and the accurate SPO triples cannot be easily extracted from the text data like a normal article.
In order to solve the problem that an SPO triple used for constructing a knowledge graph cannot be accurately extracted from unstructured recipe data, in the step, firstly, an unstructured recipe data is segmented into a plurality of sentences (or short sentences) through a Chinese word segmentation tool, then, an entity in the sentence (or short sentence) and a main predicate triple (SPO triple) taking the entity as a subject are identified by using an Ernie-MRC model, for example, an example of the SPO triple is that "tomato fried eggs are obtained by cooking tomatoes and eggs", wherein "tomato fried eggs" are recipe names and are used as subjects, "tomatoes" and "eggs" are food materials and are used as objects, "cooking" is action and is used as a predicate, and the SPO is formed jointly.
The Ernie-MRC model is a pre-training model (generally, all models need to be trained in advance through training samples, and a mainstream deep learning network belongs to the pre-training model) suitable for NLP (Natural Language Processing) in a chinese Language environment, and is used for performing feature recognition and Processing on segmented sentences or short sentences in the step, so that entity and SPO triples are recognized, and good Processing on unstructured recipe data is realized together.
Furthermore, in order to improve the accuracy of the extracted SPO triples as much as possible, after the unstructured recipe data is segmented into a plurality of sentences and before the reniere triples are extracted from at least one sentence by using the Ernie-MRC model, dimension reduction processing (e.g., Embedding processing) may be performed on each sentence, so as to obtain an Embedding corresponding to each sentence, and in the subsequent steps, the reniere triples with the recipe name as an entity may be extracted from the Embedding of at least one sentence by using the Ernie-MRC model. It should be understood that compared with the method of directly processing the Chinese text by the model, the model is generally more sensitive to the low-dimensional features after dimension reduction, and can also avoid misleading caused by some Chinese texts, especially when the model is trained based on training samples after the same dimension reduction. Of course, in addition to the dimension reduction effect achieved by the Embedding process, other processing modes that can achieve the same or similar effect may also be used, and are not specifically limited herein.
Step 203: and constructing an initial food matching knowledge graph according to the main and predicate triple, and establishing connection between each food material in the initial food matching knowledge graph and corresponding nutrient data to obtain a target food matching knowledge graph.
On the basis of step 202, in this step, an initial food matching knowledge graph is first constructed according to the SPO triples obtained by processing the recipe data, that is, only food material information and related information such as a preparation method or a preparation process included in each recipe are currently recorded in the initial food matching knowledge graph, and then a target food matching knowledge graph including the recipe name, the food material and nutrients corresponding to the food material is perfected by establishing a corresponding relationship between the food material and the previously acquired nutrient data.
The corresponding relation between the food materials and the corresponding nutrient data can be established in various ways, for example, the nutrient data can be added to the initial food matching knowledge graph as a lower node of the food materials, sub-graphs of the food materials and the nutrients can be independently added to the initial food matching knowledge graph, and the association query between the food materials sub-graph and the food materials and nutrient sub-graph is realized by taking the food materials as connection points, and the like.
In order to meet the current requirements on professional nutrition catering knowledge, the method for constructing the catering knowledge map provided by the application provides an implementation mode of firstly segmenting most of recipe data existing in an unstructured form and then extracting accurate subject-predicate-guest triples from the segmented recipe data by using an Ernie-MRC model, so that unstructured recipe data can be well processed, and meanwhile, the target catering knowledge map finally constructed by a mode of constructing food material and nutrient data simultaneously contains the recipe, food material and nutrient data, so that the method is more comprehensive and provides a comprehensive and accurate data base for subsequent query operation of a user.
In order to enable the constructed catering intellectual map to contain the recipe data and the nutrient data which are as comprehensive as possible, the recipe data and the nutrient data are obtained through a plurality of data sources, but the multi-source recipe data and the nutrient data can also cause the conditions of different contents (such as food materials) with the same (dish) name and different (dish) names with the same (dish) name, and the like, so that adverse effects are brought to the follow-up extraction of the SPO triples, and in order to improve the accuracy of the information recorded in the constructed catering intellectual map as much as possible, the processing links of disambiguation, normalization, preferential fusion among similar entities and the like can be increased, and the accuracy of the SPO triples is ensured in a manner of improving the accuracy of the entities. One implementation, including but not limited to, may be seen in the flowchart shown in fig. 3, where the flow 300 may include the following steps:
step 301: inputting recipe data of the same entity respectively acquired from a plurality of recipe data sources into a preset LightGBM model;
the LightGBM model described in this embodiment is actually a classifier trained by using a framework of the LightGBM model to obtain a result of determining whether two entities with the same name are exactly the same entity according to differences of recipe data, and the LightGBM model is directly used to refer to the classifier for convenience of calling.
Step 302: controlling the LightGBM model to normalize at least two entities having the same food material with a ratio exceeding a preset ratio into the same entity;
based on step 301, the light gbm model is used to normalize at least two entities having the same food material with a ratio exceeding a preset ratio into the same entity, that is, the classifier trained based on the light gbm model determines that the inputted recipe data of the two same entities also have high consistency and should belong to the same entity, so that a unique entity can be obtained through entity normalization.
Step 303: controlling the LightGBM model to adjust at least two entities with the same food materials which do not exceed a preset proportion into different entities in a renaming manner;
on the basis of step 301, this step is intended to adjust at least two entities having the same food material that does not exceed the preset ratio to different entities by the above-mentioned execution subject control LightGBM model in a renaming manner, i.e. the classifier trained based on the LightGBM model determines that the inputted recipe data of two identical entities does not have high consistency, and therefore the two entities should not have the same entity name, and therefore the entity will be adjusted to different entities by renaming the names of other entities.
Steps 301-303 provide a solution for disambiguating and normalizing at least two entities with the same entity name, and the classifier trained by the LightGBM model can improve the classification accuracy as much as possible.
Step 304: respectively calculating the comprehensive scores of the recipe data corresponding to at least two similar entities according to at least one of the field integrity, the consumption of food materials containing the food materials and the collection amount of the user;
in this step, the executing entity calculates the respective composite scores of the recipe data corresponding to the at least two similar entities according to at least one of the field integrity, the consumption of the food material including the food material, and the user collection amount. The field integrity terminal can be attributed to the influence parameters of the data source, and the user collection can be attributed to the influence parameters caused by the user access. The more parameters are used at the same time, the more comprehensive scores closer to the actual situation are brought. Furthermore, when the three influence parameters are used simultaneously, different weights can be given according to actual conditions, so that the comprehensive score is closer to the actual conditions through weighting of the weights.
Step 305: performing entity content fusion on the entity with the lower comprehensive score and the entity with the higher comprehensive score to obtain a fused entity;
on the basis of step 304, this step is intended to perform entity content fusion between the entity with lower comprehensive score and the entity with higher comprehensive score by the executing entity, that is, fusing the entity content with lower comprehensive score into the entity content with higher comprehensive score, so as to achieve the fullest possible.
Step 306: constructing an initial catering knowledge map according to the main-predicate-guest triples with the fusion entity as the subject;
on the basis of the step 305, the step aims to construct an initial catering knowledge map according to the main-predicate triple taking the fusion entity as the subject, so as to bring the improvement of content accuracy for the formed initial catering knowledge map by means of the processing process of the fusion entity.
Step 307: and establishing connection between each food material and corresponding nutrient data in the initial catering knowledge graph to obtain a target catering knowledge graph.
This step is the same as the second half of step 203 shown in fig. 2, and for the same contents, please refer to the corresponding parts in the previous embodiment, which is not described herein again.
Steps 301-307 in this embodiment do not include some of the preprocessing steps in the previous embodiment, but rather directly perform additional processing (i.e., disambiguation, normalization, and preferential fusion) on the identified entities, it being understood that the embodiment may also be a complete embodiment by adding preprocessing steps. Meanwhile, in the embodiment, there is no cause and dependency relationship between the entity disambiguation and normalization scheme provided in steps 301 to 303 and the similar entity preferential fusion scheme provided in steps 304 to 305, and the execution sequence can also be changed.
On the basis of any embodiment, the existing meal matching label can be extracted from the recipe data and/or the nutrient data, and then the homogenization mining of the meal matching label is carried out on part of the recipe data and/or part of the nutrient data without the existing meal matching label by using the graph neural network, so that the supplement attachment of the meal matching label is realized, and the labeling of all meal matching data is further realized. The existing meal matching labels can be personalized labels applied to different angles of the recipe characteristics, the nutrient characteristics and the like, such as light diet, body building, low fat and the like. The homogeneous mining of the missing tags is realized by using the graph neural network, because the graph relation contained in the graph neural network can more clearly show the corresponding relation between the tags and the data compared with the conventional non-graph neural network.
On the basis that the available recipe knowledge graph is constructed in any embodiment, the execution main body can call the recipe knowledge graph to respond to the query request of the user according to the following scheme, so that the requirement of the user on a certain type of dish and/or the intake of certain nutrients is met:
receiving an incoming request for inquiring the specified catering requirements;
determining and returning a matching recipe corresponding to the specified catering demand query request by using the catering knowledge map; and the supply amount of the nutrients of the food materials contained in the matching recipe is not less than the required amount of the nutrients of the specified catering demand query request.
The query request for the specified catering requirement can include at least one of a specified dish name, a specified nutrient requirement and a specified catering label, and an identifiable query type can be flexibly added according to actual conditions, and is not specifically limited here.
For further understanding, the present application further provides a specific implementation scheme in combination with a specific application scenario, please refer to the flowchart shown in fig. 4:
under the scene, the three links of data acquisition, food preparation knowledge map construction and food preparation knowledge map use can be divided according to the execution sequence, wherein the food preparation knowledge map construction link can be subdivided into 5 sub-links of knowledge extraction, entity disambiguation/normalization, similar entity preferential fusion, nutrient data edge construction and label mining/supplement, and each link is introduced respectively as follows:
1) data acquisition:
respectively capturing recipe and food nutrition ingredient table data and the like from an analysis recipe, a recipe verticality site and a nutrition verticality site, wherein the recipe data can comprise main information such as food materials, consumption and dish pictures and auxiliary information such as type labels, scores, favorite people numbers and adopted people numbers; the food nutrient composition table comprises food materials and nutrient components thereof;
2) filtering low-quality information:
and performing low-quality filtering on the captured recipe data. The low quality judgment basis may include that the description information is less than 50 words, that a picture is missing, etc.;
3) and (3) knowledge extraction:
and aiming at some unstructured recipe long texts, extracting entity information by using an Ernie-MRC model, wherein the types of entities to be extracted comprise food materials, consumption, labels and the like. The extraction module cuts the long text information into a plurality of short sentences through a Chinese word cutting tool, then calculates short sentence Embedding by using a pre-training model, and then transmits Embedding data into an Ernie-MRC model to extract SPO triples of entity information;
4) entity disambiguation/normalization:
due to the fact that data are captured from multiple source sites, a large number of repeated entity situations exist, for example, different sites have potato and beef, and entity disambiguation needs to be carried out on the situation to enable the multiple source data to be normalized. And respectively modeling because the attributes of the recipe data and the nutrient composition table data are different. And selecting a recipe name, food materials and the like as characteristics according to the recipe data. Aiming at the nutrient component data, selecting food materials and nutrient components as characteristics. Inputting the characteristics into a pre-training model for vectorization, and then training a classifier by using a LightGBM model to judge whether the entities are the same or not. Because the similarity of the whole data is compared pairwise in the comparison process, the entity names can be screened in a mode that the editing distance is smaller than a specified threshold value for accelerating the operation;
5) preferential fusion of similar entities:
after entity disambiguation and normalization, preferential fusion of multi-source data corresponding to the same entity is needed. Attributes of which one site exists but other sites do not exist are directly supplemented into the map; attribute values are preferred by the need for multiple sources and inconsistent values. And (4) counting the field integrity of the source data of each station, taking the integrity as the station weight characteristic, and considering the characteristics of the consumption proportion, the number of favorite people and the like of the food materials in the recipe data. After the 3 characteristics are uniformly weighted and calculated, the source attribute data with the highest score is selected and recorded into the map. Calculating the preference of the nutritional ingredient data in a similar way, and only selecting site weight as the characteristic;
6) building a border with nutrients:
since the recipe data does not contain nutrient data, it needs to be set aside with nutrient component data. Establishing indexes for the nutrient data, recalling corresponding nutrient lists in a retrieval mode, and then taking the total content of the nutrients of the food materials in the recipe as the nutrient attributes of the recipe, thereby establishing a side relationship;
7) catering tag mining/replenishment
The recipe data captured from partial sites may have recipe tags attached to them, but the coverage is not complete, and the part of the recipe data lacking tags is mined by using a graph neural network to obtain homogenous tags, and the recipe tags are supplemented. The atlas entity may be vectorized using a pre-trained model. Then, semi-supervised learning is carried out by using graph neural network modeling map structure information, mining of homogeneous food matching labels is realized, and a food matching knowledge map with complemented labels is finally obtained;
8) use of a catering knowledge map:
and the server providing the food matching demand query service for the user receives a specified food matching demand query request transmitted by the user, then queries in the constructed food matching knowledge graph according to the actual demand of the user, and returns a query result to the user.
Further, the server can receive long-term subscription of the user, or can regularly push matching recipes with corresponding requirements to the user, and can even provide a heat requirement result for the user according to the heat information (calorie information) obtained through further conversion.
With further reference to fig. 5, as an implementation of the method shown in the above figures, the present application provides an embodiment of an apparatus for constructing a catering knowledge graph, which corresponds to the embodiment of the method shown in fig. 2, and which is particularly applicable to various electronic devices.
As shown in fig. 5, the apparatus 500 for constructing a meal knowledge graph of the present embodiment may include: the system comprises a recipe data and nutrient data acquisition unit 501, an unstructured recipe data processing unit 502 and a catering knowledge map construction unit 503. The recipe data and nutrient data acquiring unit 501 is configured to acquire recipe data and nutrient data corresponding to food materials included in the recipe data; an unstructured recipe data processing unit 502 configured to segment unstructured recipe data into a plurality of statements, and extract a predicate triplet with a recipe name as an entity from at least one statement using an Ernie-MRC model; wherein, the entity is used as a subject of the main predicate element triple; the meal matching knowledge map building unit 503 is configured to build an initial meal matching knowledge map according to the three major-minor triples, and build connection between each food material in the initial meal matching knowledge map and corresponding nutrient data to obtain a target meal matching knowledge map.
In the present embodiment, in the apparatus 500 for constructing a meal recipe knowledge map: the detailed processing of the recipe data and nutrient data obtaining unit 501, the unstructured recipe data processing unit 502, and the meal matching knowledge graph constructing unit 503 and the technical effects brought by the processing can be referred to the related description of step 201 and step 203 in the corresponding embodiment of fig. 2, and are not described herein again.
In some optional implementations of the present embodiment, the apparatus 500 for constructing a recipe knowledge graph may further include:
the dimension reduction processing unit is configured to respectively perform dimension reduction processing on each statement after the unstructured recipe data is segmented into a plurality of statements and before the reniere triple is extracted from at least one statement by using an Ernie-MRC model, so as to respectively obtain Embedding corresponding to each statement; and the unstructured recipe data processing unit comprises a predicate element triple extraction subunit configured to extract a predicate element triple from the at least one statement using an Ernie-MRC model, the predicate element triple extraction subunit being further configured to:
and extracting the predicate-object triple from the Embedding of at least one statement by utilizing an Ernie-MRC model.
In some optional implementations of the present embodiment, the apparatus 500 for constructing a recipe knowledge graph may further include:
a data input unit configured to input recipe data of the same entity respectively acquired from a plurality of recipe data sources into a preset LightGBM model;
a multi-entity normalization unit configured to control the LightGBM model to normalize at least two entities having the same food material in excess of a preset ratio to the same entity;
a multi-entity disambiguation unit configured to control the LightGBM model to rename at least two entities having the same food material that does not exceed a preset ratio to different entities.
In some optional implementations of the present embodiment, the apparatus 500 for constructing a recipe knowledge graph may further include:
the similar entity comprehensive score calculating unit is configured to calculate respective comprehensive scores of the recipe data corresponding to the at least two similar entities according to at least one of field integrity, the consumption of food materials containing food materials and the collection amount of users;
the similar entity fusion unit is configured to perform entity content fusion on the entity with the lower comprehensive score and the entity with the higher comprehensive score to obtain a fused entity; and
the unstructured recipe data processing unit comprises an initial recipe knowledge graph construction subunit configured to construct an initial recipe knowledge graph according to the subject-predicate triplets, the initial recipe knowledge graph construction subunit configured to:
and constructing an initial catering knowledge map according to the main-predicate-object triple taking the fusion entity as the subject.
In some optional implementations of the present embodiment, the apparatus 500 for constructing a recipe knowledge graph may further include:
an existing recipe tag extraction unit configured to extract an existing recipe tag from the recipe data and/or the nutrient data;
and the homogenization label mining unit is configured to utilize the graph neural network to perform homogenization mining on the meal matching labels on part of the recipe data and/or part of the nutrient data to which the existing meal matching labels are not attached.
In some optional implementations of the present embodiment, the apparatus 500 for constructing a recipe knowledge graph may further include:
the query request receiving unit is configured to receive an incoming specified catering requirement query request;
the matching recipe determining and returning unit is configured to determine and return a matching recipe corresponding to the specified catering demand query request by using the catering knowledge map; and the supply amount of the nutrients of the food materials contained in the matching recipe is not less than the required amount of the nutrients of the specified catering demand query request.
The embodiment exists as an embodiment of an apparatus corresponding to the embodiment of the method, and in order to meet the current demand on professional nutrition catering knowledge, the apparatus for constructing a catering knowledge map provided by the application provides an implementation manner of segmenting most of recipe data existing in an unstructured form and extracting accurate subject-predicate-guest triples from the recipe data by using an Ernie-MRC model, so that unstructured recipe data can be better processed, and meanwhile, a target catering knowledge map finally constructed by establishing edges of food materials and nutrient data simultaneously contains the recipe, food materials and nutrient data, so that the method is more comprehensive and provides a comprehensive and accurate data basis for subsequent query operations of a user.
According to an embodiment of the present application, an electronic device and a computer-readable storage medium are also provided.
FIG. 6 shows a block diagram of an electronic device suitable for use in implementing the method for building a recipe knowledge graph of an embodiment of the application. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present application that are described and/or claimed herein.
As shown in fig. 6, the electronic apparatus includes: one or more processors 601, memory 602, and interfaces for connecting the various components, including a high-speed interface and a low-speed interface. The various components are interconnected using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions for execution within the electronic device, including instructions stored in or on the memory to display graphical information of a GUI on an external input/output apparatus (such as a display device coupled to the interface). In other embodiments, multiple processors and/or multiple buses may be used, along with multiple memories and multiple memories, as desired. Also, multiple electronic devices may be connected, with each device providing portions of the necessary operations (e.g., as a server array, a group of blade servers, or a multi-processor system). In fig. 6, one processor 601 is taken as an example.
The memory 602 is a non-transitory computer readable storage medium as provided herein. Wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to perform the method for constructing a recipe knowledge graph as provided herein. The non-transitory computer readable storage medium of the present application stores computer instructions for causing a computer to perform the method for constructing a catering intellectual graph provided by the present application.
The memory 602, as a non-transitory computer readable storage medium, may be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the method for constructing a meal knowledge graph in the embodiments of the present application (e.g., the meal data and nutrient data acquisition unit 501, the unstructured meal data processing unit 502, and the meal knowledge graph construction unit 503 shown in fig. 5). The processor 601 executes various functional applications of the server and data processing by running non-transitory software programs, instructions and modules stored in the memory 602, namely, implements the method for constructing a recipe knowledge graph in the above method embodiments.
The memory 602 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area can store various types of data and the like created by the electronic equipment in the process of executing the method for constructing the catering knowledge graph. Further, the memory 602 may include high speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, memory 602 optionally includes memory remotely located from processor 601, which may be connected via a network to an electronic device adapted to perform a method for constructing a recipe knowledge graph. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
An electronic device adapted to perform the method for constructing a recipe knowledge graph may further comprise: an input device 603 and an output device 604. The processor 601, the memory 602, the input device 603 and the output device 604 may be connected by a bus or other means, and fig. 6 illustrates the connection by a bus as an example.
The input device 603 may receive input numeric or character information and generate key signal inputs related to user settings and function controls of an electronic apparatus suitable for performing the method for constructing a recipe knowledge graph, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, a pointer, one or more mouse buttons, a track ball, a joystick, or the like. The output devices 604 may include a display device, auxiliary lighting devices (e.g., LEDs), and tactile feedback devices (e.g., vibrating motors), among others. The display device may include, but is not limited to, a Liquid Crystal Display (LCD), a Light Emitting Diode (LED) display, and a plasma display. In some implementations, the display device can be a touch screen.
Various implementations of the systems and techniques described here can be realized in digital electronic circuitry, integrated circuitry, application specific ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
These computer programs (also known as programs, software applications, or code) include machine instructions for a programmable processor, and may be implemented using high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.
The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The Server may be a cloud Server, which is also called a cloud computing Server or a cloud host, and is a host product in a cloud computing service system, so as to solve the defects of high management difficulty and weak service extensibility in the conventional physical host and Virtual Private Server (VPS) service.
In order to meet the current requirement on professional nutrition catering knowledge, the embodiment of the application provides an implementation mode of firstly segmenting most of recipe data existing in an unstructured form and then extracting accurate subject-predicate-object triplets from the segmented recipe data by using an Ernie-MRC model, so that unstructured recipe data can be well processed, and meanwhile, the finally constructed target catering knowledge graph simultaneously contains recipe, food and nutrient data in a mode of establishing edges of the food and the nutrient data, so that the method is more comprehensive and provides a comprehensive and accurate data base for subsequent query operation of a user.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present application may be executed in parallel, sequentially, or in different orders, and the present invention is not limited thereto as long as the desired results of the technical solutions disclosed in the present application can be achieved.
The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (14)
1. A method for constructing a meal recipe knowledge graph, comprising:
acquiring recipe data and nutrient data corresponding to food materials contained in the recipe data;
the method comprises the steps that unstructured recipe data are segmented into a plurality of statements, and an Ernie-MRC model is utilized to extract a principal and predicate object triple taking a recipe name as an entity from at least one statement; wherein the entity is used as a subject of the three-tuple of the subject and the predicate object;
and constructing an initial food matching knowledge graph according to the main and predicate triple, and establishing connection between each food material in the initial food matching knowledge graph and corresponding nutrient data to obtain a target food matching knowledge graph.
2. The method of claim 1, wherein after segmenting the unstructured recipe data into a plurality of statements, before extracting a predicate-bin triple from at least one of the statements using an Ernie-MRC model, further comprising:
performing dimensionality reduction on each statement respectively to obtain an Embedding corresponding to each statement respectively; and
the extracting of the predicate-element triplet taking the recipe name as the entity from at least one statement by using the Ernie-MRC model comprises the following steps:
and extracting a main predicate object triple taking the recipe name as an entity from Embedding of at least one statement by using the Ernie-MRC model.
3. The method of claim 1, further comprising:
inputting recipe data of the same entity respectively acquired from a plurality of recipe data sources into a preset LightGBM model;
controlling the LightGBM model to normalize at least two entities having the same food material with a ratio exceeding a preset ratio into the same entity;
controlling the LightGBM model to rename at least two entities having the same food material not exceeding the preset proportion into different entities.
4. The method of claim 1, further comprising:
respectively calculating the comprehensive scores of the recipe data corresponding to at least two similar entities according to at least one of the field integrity, the consumption of food materials containing the food materials and the collection amount of the user;
performing entity content fusion on the entity with the lower comprehensive score and the entity with the higher comprehensive score to obtain a fused entity; and
the method for constructing the initial catering knowledge map according to the main and predicate element triples comprises the following steps:
and constructing the initial catering knowledge map according to the main-predicate-object triple taking the fusion entity as the subject.
5. The method of claim 1, further comprising:
extracting existing meal tags from the recipe data and/or the nutrient data;
and carrying out homogenization mining on the meal matching labels on part of the recipe data and/or part of the nutrient data without the existing meal matching labels by using a graph neural network.
6. The method of any of claims 1-5, further comprising:
receiving an incoming request for inquiring the specified catering requirements;
determining and returning a matching recipe corresponding to the specified catering demand query request by using the catering knowledge map; and the supply amount of the nutrients of the food materials contained in the matching recipe is not less than the required amount of the nutrients of the specified catering demand query request.
7. An apparatus for constructing a meal recipe knowledge graph, comprising:
a recipe data and nutrient data acquisition unit configured to acquire recipe data and nutrient data corresponding to food materials included in the recipe data;
the unstructured recipe data processing unit is configured to segment unstructured recipe data into a plurality of statements, and extract a predicate triplet taking a recipe name as an entity from at least one statement by using an Ernie-MRC model; wherein the entity is used as a subject of the three-tuple of the subject and the predicate object;
and the food matching knowledge map construction unit is configured to construct an initial food matching knowledge map according to the main-predicate triple group, and establish connection between each food material in the initial food matching knowledge map and corresponding nutrient data to obtain a target food matching knowledge map.
8. The apparatus of claim 7, further comprising:
the dimension reduction processing unit is configured to respectively perform dimension reduction processing on each statement after the unstructured recipe data is segmented into a plurality of statements and before a predicate-predicate triple is extracted from at least one statement by using an Ernie-MRC model, so as to respectively obtain Embedding corresponding to each statement; and
the unstructured recipe data processing unit comprises a predicate element triple extraction subunit configured to extract a predicate element triple from at least one of the statements using an Ernie-MRC model, the predicate element triple extraction subunit being further configured to:
and extracting the main predicate element triple from the Embedding of at least one statement by using the Ernie-MRC model.
9. The apparatus of claim 7, further comprising:
a data input unit configured to input recipe data of the same entity respectively acquired from a plurality of recipe data sources into a preset LightGBM model;
a multi-entity normalization unit configured to control the LightGBM model to normalize at least two entities having the same food material in excess of a preset ratio to the same entity;
a multi-entity disambiguation unit configured to control the LightGBM model to rename at least two entities having the same food material that does not exceed the preset ratio to different entities.
10. The apparatus of claim 7, further comprising:
the similar entity comprehensive score calculating unit is configured to calculate respective comprehensive scores of the recipe data corresponding to the at least two similar entities according to at least one of field integrity, the consumption of food materials containing food materials and the collection amount of users;
the similar entity fusion unit is configured to perform entity content fusion on the entity with the lower comprehensive score and the entity with the higher comprehensive score to obtain a fused entity; and
the unstructured recipe data processing unit comprises an initial recipe knowledge graph construction subunit configured to construct an initial recipe knowledge graph according to the subject-predicate triplets, the initial recipe knowledge graph construction subunit configured to:
and constructing the initial catering knowledge map according to the main-predicate-object triple taking the fusion entity as the subject.
11. The apparatus of claim 7, further comprising:
an existing recipe tag extraction unit configured to extract an existing recipe tag from the recipe data and/or the nutrient data;
and the homogenization label mining unit is configured to utilize the graph neural network to perform homogenization mining on the meal matching labels on part of the recipe data and/or part of the nutrient data to which the existing meal matching labels are not attached.
12. The apparatus of any of claims 7-11, further comprising:
the query request receiving unit is configured to receive an incoming specified catering requirement query request;
the matching recipe determining and returning unit is configured to determine and return a matching recipe corresponding to the specified catering demand query request by using the catering knowledge graph; and the supply amount of the nutrients of the food materials contained in the matching recipe is not less than the required amount of the nutrients of the specified catering demand query request.
13. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method for building a meal knowledge graph of any one of claims 1-6.
14. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method for constructing a meal knowledge graph of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011377278.2A CN112380356A (en) | 2020-11-30 | 2020-11-30 | Method, device, electronic equipment and medium for constructing catering knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011377278.2A CN112380356A (en) | 2020-11-30 | 2020-11-30 | Method, device, electronic equipment and medium for constructing catering knowledge graph |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112380356A true CN112380356A (en) | 2021-02-19 |
Family
ID=74590302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011377278.2A Pending CN112380356A (en) | 2020-11-30 | 2020-11-30 | Method, device, electronic equipment and medium for constructing catering knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112380356A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113656590A (en) * | 2021-07-16 | 2021-11-16 | 北京百度网讯科技有限公司 | Industry map construction method and device, electronic equipment and storage medium |
| CN113886779A (en) * | 2021-09-30 | 2022-01-04 | 北京房江湖科技有限公司 | Method for identifying person identity, storage medium and computer program product |
| CN113918677A (en) * | 2021-10-18 | 2022-01-11 | 智联(无锡)信息技术有限公司 | Data processing method and device based on knowledge graph automation link layering and computer readable medium |
| CN114661918A (en) * | 2022-03-17 | 2022-06-24 | 北京百度网讯科技有限公司 | Knowledge graph construction method and device, storage medium and electronic equipment |
| CN117725237A (en) * | 2024-02-07 | 2024-03-19 | 深圳鸿博智成科技有限公司 | Recipe generation method, device, equipment and storage medium based on big data |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480183A (en) * | 2017-07-06 | 2017-12-15 | 广州索答信息科技有限公司 | A kind of kitchen domain knowledge collection of illustrative plates querying method, device and storage medium |
| US20190034780A1 (en) * | 2017-07-31 | 2019-01-31 | Microsoft Technology Licensing, Llc | Knowledge Graph For Conversational Semantic Search |
| US10242320B1 (en) * | 2018-04-19 | 2019-03-26 | Maana, Inc. | Machine assisted learning of entities |
| CN110504019A (en) * | 2019-08-30 | 2019-11-26 | 北京妙医佳健康科技集团有限公司 | User individual dietary recommendations continued method, apparatus, electronic equipment and storage medium |
| CN111261260A (en) * | 2020-01-13 | 2020-06-09 | 深圳市华膳科技有限公司 | Diet recommendation system |
| CN111291185A (en) * | 2020-01-21 | 2020-06-16 | 京东方科技集团股份有限公司 | Information extraction method and device, electronic equipment and storage medium |
| CN111625715A (en) * | 2020-05-09 | 2020-09-04 | 北京达佳互联信息技术有限公司 | Information extraction method and device, electronic equipment and storage medium |
| CN111897969A (en) * | 2020-07-27 | 2020-11-06 | 武汉大学 | Method and system for analyzing correlation between food components and nutritional health based on knowledge graph |
-
2020
- 2020-11-30 CN CN202011377278.2A patent/CN112380356A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480183A (en) * | 2017-07-06 | 2017-12-15 | 广州索答信息科技有限公司 | A kind of kitchen domain knowledge collection of illustrative plates querying method, device and storage medium |
| US20190034780A1 (en) * | 2017-07-31 | 2019-01-31 | Microsoft Technology Licensing, Llc | Knowledge Graph For Conversational Semantic Search |
| US10242320B1 (en) * | 2018-04-19 | 2019-03-26 | Maana, Inc. | Machine assisted learning of entities |
| CN110504019A (en) * | 2019-08-30 | 2019-11-26 | 北京妙医佳健康科技集团有限公司 | User individual dietary recommendations continued method, apparatus, electronic equipment and storage medium |
| CN111261260A (en) * | 2020-01-13 | 2020-06-09 | 深圳市华膳科技有限公司 | Diet recommendation system |
| CN111291185A (en) * | 2020-01-21 | 2020-06-16 | 京东方科技集团股份有限公司 | Information extraction method and device, electronic equipment and storage medium |
| CN111625715A (en) * | 2020-05-09 | 2020-09-04 | 北京达佳互联信息技术有限公司 | Information extraction method and device, electronic equipment and storage medium |
| CN111897969A (en) * | 2020-07-27 | 2020-11-06 | 武汉大学 | Method and system for analyzing correlation between food components and nutritional health based on knowledge graph |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113656590A (en) * | 2021-07-16 | 2021-11-16 | 北京百度网讯科技有限公司 | Industry map construction method and device, electronic equipment and storage medium |
| CN113656590B (en) * | 2021-07-16 | 2023-12-15 | 北京百度网讯科技有限公司 | Industry map construction method and device, electronic equipment and storage medium |
| CN113886779A (en) * | 2021-09-30 | 2022-01-04 | 北京房江湖科技有限公司 | Method for identifying person identity, storage medium and computer program product |
| CN113918677A (en) * | 2021-10-18 | 2022-01-11 | 智联(无锡)信息技术有限公司 | Data processing method and device based on knowledge graph automation link layering and computer readable medium |
| CN114661918A (en) * | 2022-03-17 | 2022-06-24 | 北京百度网讯科技有限公司 | Knowledge graph construction method and device, storage medium and electronic equipment |
| CN117725237A (en) * | 2024-02-07 | 2024-03-19 | 深圳鸿博智成科技有限公司 | Recipe generation method, device, equipment and storage medium based on big data |
| CN117725237B (en) * | 2024-02-07 | 2024-04-30 | 深圳鸿博智成科技有限公司 | Recipe generation method, device, equipment and storage medium based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112380356A (en) | Method, device, electronic equipment and medium for constructing catering knowledge graph | |
| CN110543574B (en) | Knowledge graph construction method, device, equipment and medium | |
| CN112115299B (en) | Video searching method, video searching device, video recommending method, electronic equipment and storage medium | |
| CN111753099B (en) | Method and system for enhancing relevance of archive entity based on knowledge graph | |
| CN108701161B (en) | Providing images for search queries | |
| CN111625635A (en) | Question-answer processing method, language model training method, device, equipment and storage medium | |
| CN111190522A (en) | Generating three-dimensional digital content from natural language requests | |
| JP7222040B2 (en) | Model training, image processing method and device, storage medium, program product | |
| CN112541359B (en) | Document content identification method, device, electronic equipment and medium | |
| US9606975B2 (en) | Apparatus and method for automatically generating visual annotation based on visual language | |
| US9418058B2 (en) | Processing method for social media issue and server device supporting the same | |
| CN113627797B (en) | Method, device, computer equipment and storage medium for generating staff member portrait | |
| CN111580947A (en) | Online collaborative remote sensing image annotation system based on artificial intelligence | |
| CN114595686B (en) | Knowledge extraction method, and training method and device of knowledge extraction model | |
| CN116012481A (en) | Image generation processing method and device, electronic equipment and storage medium | |
| CN111767381A (en) | Automatic question answering method and device | |
| CN113836314B (en) | Knowledge graph construction method, device, equipment and storage medium | |
| CN111538815A (en) | Text query method, device, equipment and storage medium | |
| CN111767334A (en) | Information extraction method and device, electronic equipment and storage medium | |
| CN113988795A (en) | Task flow processing method, device, equipment, storage medium and program product | |
| CN114861677A (en) | Information extraction method, information extraction device, electronic equipment and storage medium | |
| CN113836316B (en) | Processing method, training method, device, equipment and medium for ternary group data | |
| JP2023010805A (en) | Method for training document information extraction model and extracting document information, device, electronic apparatus, storage medium and computer program | |
| CN111666417A (en) | Method and device for generating synonyms, electronic equipment and readable storage medium | |
| CN113408290A (en) | Intelligent marking method and system for Chinese text |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210219 |