Method, device, medium and equipment for identifying business district entities

Abstract

The present disclosure relates to the field of artificial intelligence technology, and provides a method, device, medium and equipment for identifying business district entities. The method includes: performing named entity recognition on the text to be recognized to determine the entity to be recognized; performing feature extraction on the entity to be recognized to obtain the name feature and attribute feature corresponding to the entity to be recognized; inputting the name feature and attribute feature into the trained entity recognition model, so that the entity recognition model predicts whether the entity to be recognized is a business district entity or a non-business district entity based on the name feature and attribute feature. The present technical solution realizes the automatic recognition of business district entities based on artificial intelligence technology, which is conducive to improving the efficiency and accuracy of entity recognition.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular, to a method for identifying a business district entity, a device for identifying a business district entity, and a computer-readable storage and electronic device for implementing the method. Background Art

[0002] In the related art, business district entities are usually screened manually. However, the screening rules that need to be paid attention to in the manual screening process are complicated, which is prone to errors and time-consuming. At the same time, the incomplete screening rules are prone to omissions, wrong selections, multiple selections, etc. In addition, the newly added screening rules also bring certain difficulties to the subsequent management and maintenance.

[0003] Therefore, there is an urgent need to provide an automatic recognition solution for business district entities to improve the accuracy and efficiency of entity recognition.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure. Summary of the invention

[0005] The purpose of the present disclosure is to provide a business district entity recognition method, a business district entity recognition device, and a computer-readable storage and electronic device for implementing the above method, thereby improving the accuracy and efficiency of entity recognition at least to a certain extent.

[0006] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0007] According to one aspect of the present disclosure, a method for identifying business district entities is provided, comprising: performing named entity recognition on a text to be identified to determine the entity to be identified; performing feature extraction on the entity to be identified to obtain name features and attribute features corresponding to the entity to be identified; inputting the name features and the attribute features into a trained entity recognition model, so that the entity recognition model predicts whether the entity to be identified is a business district entity or a non-business district entity based on the name features and the attribute features.

[0008] According to one aspect of the present disclosure, a device for identifying business district entities is provided, including: a module for determining an entity to be identified, configured to: perform named entity recognition on a text to be identified, so as to determine an entity to be identified; a module for processing an entity to be identified, configured to: perform feature extraction on the entity to be identified, so as to obtain name features and attribute features corresponding to the entity to be identified; and a recognition module, configured to: input the name features and the attribute features into a trained entity recognition model, so that the entity recognition model predicts that the entity to be identified is a business district entity or a non-business district entity based on the name features and the attribute features.

[0009] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned entity processing module to be identified is specifically configured to: extract the preset words of the business district contained in the entity to be identified or extract the preset words of the non-business district contained in the entity to be identified to obtain the name features corresponding to the entity to be identified; obtain the store types and the number of stores of each type contained in the entity to be identified to determine the store distribution characteristics, and obtain the attribute features corresponding to the entity to be identified.

[0010] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned device also includes: a model training module, and the model training module includes: an entity determination submodule, a feature extraction submodule, a loss determination submodule and an iterative calculation submodule.

[0011] Among them, the above-mentioned entity determination submodule is configured to: determine the entity set about the positive sample according to the identity information about multiple stores, and obtain the entity set about the negative sample; the above-mentioned feature extraction submodule is configured to: perform feature extraction on the above-mentioned entity set to obtain the name features and attribute features corresponding to the first category entity labels, and the name features and attribute features corresponding to the second category entity labels; the above-mentioned loss determination submodule is configured to: determine the model loss according to the name features and attribute features corresponding to the above-mentioned first category entity labels and the name features and attribute features corresponding to the above-mentioned second category entity labels; and the above-mentioned iterative calculation submodule is configured to: iteratively calculate the model parameters of the above-mentioned entity recognition model according to the above-mentioned model loss to optimize the above-mentioned model parameters and obtain the above-mentioned trained entity recognition model.

[0012] In some embodiments of the present disclosure, based on the aforementioned solution, the feature extraction submodule includes: an entity classification unit, a first extraction unit, a second extraction unit, a third extraction unit and a fourth extraction unit.

[0013] Among them, the above-mentioned entity classification unit is configured to: obtain the first category entity and the second category entity according to the label of the entity in the above-mentioned entity set; the above-mentioned first extraction unit is configured to: extract the first category preset words contained in the above-mentioned first category entity, and obtain the name features corresponding to the first category entity label; the above-mentioned second extraction unit is configured to: extract the second category preset words contained in the above-mentioned second category entity, and obtain the name features corresponding to the second category entity label; the above-mentioned third extraction unit is configured to: obtain the store types contained in the above-mentioned first category entity and the number of stores of each type to determine the store distribution characteristics, and obtain the attribute characteristics corresponding to the first category entity label; the above-mentioned fourth extraction unit is configured to: obtain the store types contained in the second category entity and the number of stores of each type to determine the store distribution characteristics, and obtain the attribute characteristics corresponding to the second category entity label.

[0014] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned first extraction unit is specifically configured to: determine the word at the end position in the above-mentioned first category entity as the first target word; if the similarity between the above-mentioned first target word and the above-mentioned first category preset word is greater than a preset threshold, then extract the above-mentioned first target word to obtain the name feature corresponding to the first category entity label.

[0015] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned second extraction unit is specifically configured to: determine the word at the end position in the above-mentioned second category entity as the second target word; if the similarity between the above-mentioned second target word and the above-mentioned second category preset word is greater than a preset threshold, then extract the above-mentioned second target word to obtain the name feature corresponding to the second category entity label.

[0016] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned third extraction unit is specifically configured to: group the first category entities containing the same words; respectively obtain the store keywords contained in each group of entities, and obtain the store distribution characteristics corresponding to each group of entities.

[0017] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned loss determination submodule is specifically configured as follows: for the ith decision tree, obtain the ith feature set, calculate the information entropy when any feature in the ith feature set is used as a splitting node, and use the feature corresponding to the maximum value of the information entropy difference as the root node of the ith decision tree, wherein the value of i is a positive integer not greater than the total number of preset decision trees, and the ith feature set includes the name feature and attribute feature corresponding to the first category entity label and the name feature and attribute feature corresponding to the second category entity label; after excluding the feature corresponding to the root node of the ith decision tree in the ith feature set, calculate the information entropy when any feature is used as a splitting node, and use the feature corresponding to the maximum value of the information entropy difference as the intermediate node of the ith decision tree until the preset splitting height or preset information entropy difference of the ith decision tree is reached; obtain the first predicted label corresponding to the leaf node determined when the decision is made with the root node, and obtain the second predicted label corresponding to the leaf node determined when the decision is made with the intermediate node, and determine the model loss corresponding to the ith decision tree according to the first predicted label, the second predicted label and the actual label.

[0018] In some embodiments of the present disclosure, based on the aforementioned scheme, after the loss determination submodule determines the model loss, the iterative calculation submodule is specifically configured to: determine the cross entropy loss function as the loss function of the entity recognition model based on the model loss; and use the stochastic gradient descent method to process the loss function to iteratively calculate the model parameters of the entity recognition model.

[0019] In some embodiments of the present disclosure, the above-mentioned entity determination submodule is specifically configured to: input the above-mentioned identity information about multiple stores into a pre-trained named entity recognition model; based on the above-mentioned pre-trained named entity recognition model, extract the target type of named entities from the unstructured identity information text to obtain the above-mentioned entity set about the positive sample.

[0020] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned identity information includes one or more of the following information: store name, store address, store category, store WiFi name, and longitude and latitude of the store location.

[0021] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned first category entity tags are business district tags, and the above-mentioned second category entity tags are non-business district tags, wherein the above-mentioned first category preset words include one or more of the following information: square, city, center, village and gathering; the above-mentioned second category preset words include one or more of the following information: town, block, courtyard, shop, lane and number.

[0022] According to one aspect of the present disclosure, a computer storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for identifying a business district entity described in the first aspect is implemented.

[0023] According to one aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method for identifying a business district entity described in the first aspect above by executing the executable instructions.

[0024] According to one aspect of the present disclosure, a computer program product or a computer program is provided, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the business district entity recognition method provided in each of the above embodiments.

[0025] It can be seen from the above technical solutions that the method for identifying a business district entity, the device for identifying a business district entity, the computer storage medium, and the electronic device in the exemplary embodiments of the present disclosure have at least the following advantages and positive effects:

[0026] In the technical solutions provided in some embodiments of the present disclosure, the entities to be identified are identified by performing named entity recognition on the text to be identified, thereby reducing the amount of data to be identified and improving the recognition efficiency of business district entities. Furthermore, in order to effectively distinguish whether the entity to be identified belongs to a business district entity, feature extraction is performed on the entity to be identified to obtain key features for distinguishing between business district entities and non-business district entities. The extracted key features are input into the trained entity recognition model to achieve automatic recognition of the entity to be identified, and automatic recognition of business district entities based on artificial intelligence technology can improve entity recognition efficiency and recognition accuracy.

[0027] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0029] Figure 1 A schematic diagram showing an exemplary system architecture to which the technical solution of the embodiments of the present disclosure can be applied.

[0030] Figure 2 A flowchart of a method for training an entity recognition model in an exemplary embodiment of the present disclosure is shown.

[0031] Figure 3 A schematic flow chart of a method for extracting features from an entity in an exemplary embodiment of the present disclosure is shown.

[0032] Figure 4 A pie chart containing the first category of preset words in an exemplary embodiment of the present disclosure is schematically shown.

[0033] Figure 5 A pie chart containing the second category of preset words in an exemplary embodiment of the present disclosure is schematically shown.

[0034] Figure 6 A flow chart of a method for determining model loss in an exemplary embodiment of the present disclosure is shown.

[0035] Figure 7 A schematic diagram of the process of a business district entity identification method in an exemplary embodiment of the present disclosure is shown.

[0036] Figure 8A schematic flow chart showing a method for processing an entity to be identified in an exemplary embodiment of the present disclosure.

[0037] Fig. 9 A schematic diagram of the structure of a business district entity identification device in an exemplary embodiment of the present disclosure is shown.

[0038] Fig.10 A schematic structural diagram of an electronic device in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0039] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more comprehensive and complete and will fully convey the concept of the example embodiments to those skilled in the art.

[0040] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the present disclosure.

[0041] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0042] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0043] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0044] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0045] Natural language processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field will involve natural language, that is, the language people use in daily life, so it is closely related to the study of linguistics. Natural language processing technology usually includes text processing, semantic understanding, machine translation, robot question answering, knowledge graph and other technologies.

[0046] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, transfer learning, inductive learning, and self-learning.

[0047] The solution provided in the embodiments of the present application involves technologies such as natural language processing and machine learning of artificial intelligence, which are specifically described by the following embodiments:

[0048] Figure 1A schematic diagram of a system architecture of an exemplary application environment in which a business district entity recognition solution according to an embodiment of the present disclosure can be applied is shown.

[0049] like Figure 1 As shown, the system architecture 100 may include a terminal 110 , a network 120 , and a server 130 . The terminal 110 and the server 130 are connected via the network 120 .

[0050] The terminal 110 can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. In the embodiment provided by the present technical solution, the terminal 110 can represent an application terminal and an approval terminal. The network 120 can be a communication medium of various connection types that can provide a communication link between the terminal 110 and the server 130, for example, it can be a wired communication link, a wireless communication link or an optical fiber cable, etc., which is not limited in this application. The server 130 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, and basic cloud computing services such as big data and artificial intelligence platforms.

[0051] Specifically, the server 130 can provide training for the entity recognition model, and can also store the trained project recommendation model entity recognition model to provide the terminal 110 with a business entity recognition algorithm.

[0052] The business district entity recognition method provided in the embodiment of the present disclosure can be executed by the server 130, and the business district entity recognition method in the embodiment of the present disclosure can also be applied to the terminal 110, and the present disclosure does not make any special limitation on this. The embodiment of the present disclosure mainly uses the business district entity recognition method applied to the server 130 as an example.

[0053] For example, in an exemplary embodiment, the server 130 completes the training of the entity recognition model. Specifically, the server 130 determines the entity set about the positive sample based on the identity information about the multiple stores, and also obtains the entity set about the negative sample. Then, the server 130 performs feature extraction on the entity set of the above positive and negative samples to obtain the name features and attribute features corresponding to the first category of entity tags, and the name features and attribute features corresponding to the second category of entity tags. Further, the server 130 determines the model loss based on the name features and attribute features corresponding to the first category of entity tags and the name features and attribute features corresponding to the second category of entity tags. The server 130 iteratively calculates the model parameters of the entity recognition model based on the model loss to optimize the model parameters and obtain the trained entity recognition model.

[0054] Furthermore, in an exemplary embodiment, named entity recognition is performed on the text to be recognized by the server 130 to determine the entity to be recognized. Specifically, the server 130 performs feature extraction on the entity to be recognized to obtain the name feature and attribute feature corresponding to the entity to be recognized. Furthermore, the name feature and attribute feature are input into the trained entity recognition model stored in the server 130 to determine whether the entity to be recognized is a business district entity or a non-business district entity according to the output of the entity recognition model.

[0055] In an exemplary embodiment, in order to facilitate the terminal to perform recognition operations, the above-mentioned trained entity recognition model can also be sent from the server 130 to the terminal 110, so that the terminal 110 can predict whether the entity to be recognized belongs to a business district entity. Specifically, the terminal 110 performs named entity recognition on the text to be recognized to determine the entity to be recognized, and extracts features of the entity to be recognized to obtain name features and attribute features of the entity to be recognized. Furthermore, the name features and attribute features are input into the trained entity recognition model stored in the terminal to determine whether the above-mentioned entity to be recognized is a business district entity or a non-business district entity according to the output of the entity recognition model.

[0056] Next, we will introduce an implementation example of entity recognition model training. Figure 2 A flow chart showing a method for training an entity recognition model in an exemplary embodiment of the present disclosure is shown. Figure 2 The training method of the entity recognition model provided in this embodiment includes:

[0057] Step S210, determining an entity set related to positive samples according to identity information about multiple stores, and obtaining an entity set related to negative samples;

[0058] Step S220, performing feature extraction on the entity set to obtain name features and attribute features corresponding to the first type of entity tags, and name features and attribute features corresponding to the second type of entity tags;

[0059] Step S230, determining a model loss according to the name features and attribute features corresponding to the first type of entity tags and the name features and attribute features corresponding to the second type of entity tags; and

[0060] Step S240, iteratively calculating the model parameters of the entity recognition model according to the model loss to optimize the model parameters and obtain a trained entity recognition model.

[0061] exist Figure 2In the technical solution provided by the illustrated embodiment, automatic identification of business district entities is achieved based on artificial intelligence technology, thereby avoiding the problems of low identification efficiency, low identification accuracy and difficult maintenance of screening rules caused by specifying screening rules and manually identifying business district entities based on the screening rules. At the same time, in the model training process provided by the present technical solution, in order to effectively distinguish whether the entity to be identified belongs to the business district entity, the acquired entity set is subjected to feature extraction to obtain the name features and attribute features corresponding to the first type of entity label and the name features and attribute features corresponding to the second type of entity label, and the model loss is determined according to the above-mentioned key features for distinguishing between business district entities and non-business district entities, so as to effectively distinguish between business district entities and non-business district entities through their respective characteristics, so as to effectively improve the accuracy of entity recognition.

[0062] The business cluster entity refers to a comprehensive place for leisure and shopping, such as Wanda Plaza, Tee Mall, Taikoo Hui, etc. Non-business cluster entities refer to entities other than business cluster entities. The above scheme is suitable for identifying and filtering data belonging to "business cluster entities" from complex data through the above entity recognition model.

[0063] The following Figure 2 The specific implementation methods of each step of the illustrated embodiment are described in detail:

[0064] Exemplarily, in step S210, raw data for training the entity recognition model is obtained. Identity information about multiple stores can be obtained. The identity information of the current store specifically refers to information used to distinguish it from other stores. Referring to Table 1, the identity information of the current store includes: the store name of the current store, the store address of the current store, the store category of the current store, the store WiFi name of the current store, and the longitude and latitude of the store location of the current store. For example, Xiao A Food (XX Plaza Store), the address is 6th Floor, XX Plaza, No. 228, XX Road, XX District, XX City, XX Province.

[0065] Table 1

[0066]

[0067] In an exemplary embodiment, on the one hand, the positive samples of the entity training model can be obtained through the above-mentioned identity information about the store (such as store address information), and on the other hand, the negative samples of the model are determined by obtaining non-business district entity-related information. For example, in Table 1, the row of "Grandma's House (Zhengjia Plaza Store)" can be used as a positive sample, and the row of "China Unicom" can be used as a negative sample. In view of the fact that the above-mentioned raw data is used to train the entity recognition model, it is necessary to perform named entity recognition on the above-mentioned raw data, that is, to identify the boundaries and categories of entity references in natural texts, specifically, to identify entities with specific meanings in the text, such as names of people, places, institutions, proper nouns, etc. In this way, the training set / test set for the above-mentioned entity recognition model can be determined.

[0068] Exemplarily, entity sets for positive and negative samples can be determined based on a pre-trained machine learning model. Specifically, by inputting the store address in the above raw data into a pre-trained named entity recognition model, it is possible to extract the target type of named entities from the text of the unstructured identity information, and obtain the entity set for the positive sample. The entity set for the negative sample can also be determined by a pre-trained machine learning model. Specifically, by inputting non-trade district entity related information into a pre-trained named entity recognition model, it is possible to extract named entities from the unstructured non-trade district entity related information text, and obtain the entity set for the negative sample.

[0069] The pre-trained named entity recognition model may adopt a pre-trained Hidden Markov Model (HMM) and a sequence labeling algorithm (CRF). Other machine learning models for named entity recognition in the art may also be adopted, which are not limited here.

[0070] Referring to Table 2, the entity obtained for the identity information shown in Table 1 is shown in the last column "Identification Entity" in the following table.

[0071] Table 2

[0072]

[0073] It should be noted that the above entity set includes business district entities and non-business district entities. By marking the entities in the entity set, positive samples marked as business district entities and negative samples marked as non-business district entities are obtained. Exemplarily, the above positive samples and negative samples can be divided into a training set and a test set. Exemplarily, the distribution ratio of training samples and test samples is shown in Table 3.

[0074] Table 3

[0075] About the number of entities in positive samples About the number of entities in negative samples Training set 6446 9656 Test Set 2762 4138

[0076] Among them, the above-mentioned training set is used for the training process of the entity recognition model, and the above-mentioned test set is used to test the entity training model in training. The specific test implementation examples will be introduced in the embodiments described later. The following first introduces the training process of the entity recognition model.

[0077] In an exemplary embodiment, when the model is trained according to the above training set, in order to enable the trained entity recognition model to distinguish between business district entities and non-business district entities with high accuracy, it is necessary to determine the key features that distinguish between the two. Therefore, in this technical solution, the above key features are extracted from the entities in the above training set, and the extracted key features are input into the entity recognition model for model training.

[0078] refer to Figure 2 In step S220, feature extraction is performed on the entity set to obtain name features and attribute features corresponding to the first type of entity tags, and name features and attribute features corresponding to the second type of entity tags.

[0079] For example, influenced by traditional culture and habits, the naming of business district entities can generally be defined according to certain customary standards. For example, business district entities generally end with words such as square, center, city, etc., indicating that they are large comprehensive shopping malls, while the naming of non-business district entities is generally messy, but there are still a small number of regularities to follow, such as XX compound, XX street, etc. in villages and towns. Since the naming of business district entities and non-business district entities can be distinguished by certain regularities, the name characteristics of business district entities and non-business district entities can be used as key features to distinguish the two.

[0080] Exemplary, reference Figure 3 A schematic flow chart of a method for extracting entity features in an exemplary embodiment of the present disclosure is shown. The method includes steps S310 to S330.

[0081] In step S310, the first category of preset words contained in the first category of entities are extracted to obtain name features corresponding to the first category of entity labels, wherein the first category of entities are determined according to the labels of the entities in the entity set.

[0082] Among them, the above-mentioned first-category entity tags are business district entity tags, and the above-mentioned first-category preset words include one or more of the following information: square, city, center, village and gathering.

[0083] In an exemplary embodiment, the specific implementation method of obtaining the name features corresponding to the first category entity tags from the first category entities (trading district entities) in the above step S310 may be: determining the word at the end position in the first category entity as the first target word; if the first target word is the same as the first category preset word, or the first target word contains the first category preset word, then extracting the first target word to obtain the name features corresponding to the above first category entity tags.

[0084] Exemplarily, the above-mentioned first-category preset words are generally located at the last position of the business district entity, such as mm square, nn big square, etc. Therefore, in this embodiment, the words at the end position in the first-category entity (i.e., the business district entity) are first obtained as the target words (recorded as "first target words"). And further determine whether the target word contains the above-mentioned first-category preset words: square, city, center, village or confluence. If the above-mentioned first target word contains any one of the first-category preset words, it means that the business district entity contains the name feature, and the target word is extracted as the name feature corresponding to the first-category entity label. For example, the extracted business district entity "mm square" contains the name feature "square" but does not contain the name feature "confluence", then the "square" dimension feature corresponding to the entity "mm square" can be expressed as "1", and at the same time, the "confluence" dimension feature corresponding to the feature can be expressed as "0". If the above-mentioned first target word does not contain any of the first-category preset words, it means that the business district entity does not contain the name feature, and the entity is not extracted.

[0085] Exemplary, reference Figure 4 The pie chart schematically shown in an exemplary embodiment of the present disclosure includes the first category of preset words, which includes a total of 6 dimensions of features. Specifically, 39% of the business district entities contain "square" in their names, 26% of the business district entities contain "city" in their names, 10% of the business district entities contain "center" in their names, 4% of the business district entities contain "li" in their names, and 3% of the business district entities contain "hui" in their names. It can be seen that 82% of the business district entities contain the above-mentioned first category of preset words, so the name feature can be used as a key feature to distinguish between business district entities and non-business district entities. In other words, extracting the name features of business district entities as key features and using them to train entity recognition models is conducive to improving the accuracy of entity recognition.

[0086] In an exemplary embodiment, not all entities with the above-mentioned first type of preset words are business district entities, and non-business district entities may also contain the above-mentioned first preset words. For example, although "Convention and Exhibition Center" contains "center", it does not belong to a business district entity. Therefore, in addition to extracting the name features corresponding to the above-mentioned business district entities, this technical solution also extracts the name features corresponding to non-business district entities to reduce the possibility of identifying non-business district entities as business district entities, thereby further improving the recognition accuracy of the entity recognition model.

[0087] Exemplarily, in step S320, the second category preset words contained in the second category entity are extracted to obtain name features corresponding to the second category entity label, wherein the second category entity is determined according to the label of the entity in the entity set.

[0088] Among them, the above-mentioned second-category entity tags are non-commercial district entity tags, and the above-mentioned second-category preset words include one or more of the following information: town, block, courtyard, shop, road, lane and number.

[0089] In an exemplary embodiment, the specific implementation method of obtaining the name features corresponding to the second category entity tags from the second category entities (non-trade district entities) in the above step S320 may be: determining the words at the end position in the second category entity as the second target words; if the second target words are the same as the second category preset words, or the second target words contain the second category preset words, then extracting the second target words to obtain the name features corresponding to the above second category entity tags.

[0090] Similar to the specific implementation of step S310, the above-mentioned second-category preset words are generally located at the last position of the non-trading district entity, such as mm town, nn square, etc. Therefore, in this embodiment, the words at the end position in the second-category entity (i.e., the non-trading district entity) are first obtained as the target words (referred to as "second target words"). And further determine whether the target word contains the above-mentioned second-category preset words: town, square, courtyard, shop, road, lane or number. If the above-mentioned second target word contains any one of the second-category preset words, it means that the non-trading district entity contains a name feature, and the target word is extracted as the name feature corresponding to the second-category entity label. If it does not contain any of the second-category preset words, it means that the non-trading district entity does not contain a name feature, and the entity will not be extracted. For example, the extracted non-commercial district entity "xx Road, No. yy" contains the name features "road" and "number", but does not contain the name feature "town". The "road" dimension feature corresponding to the entity "xx Road, No. yy" can be expressed as "1", and the "number" dimension feature corresponding to the feature can be expressed as "1". At the same time, the "town" dimension feature corresponding to the feature can be expressed as "0".

[0091] refer to Figure 5The pie chart containing the second type of preset words in an exemplary embodiment of the present disclosure schematically shown contains 11 dimensional features in total. Specifically, in addition to 59% of non-business district entities not containing the above second preset words, 59% of the names of non-business district entities contain keywords such as town, square, courtyard, shop, road, lane or number, and the second preset words generally do not appear in the names of business district entities. Therefore, in this technical solution, both the name features corresponding to the above business district entities and the name features corresponding to non-business district entities are extracted to improve the recognition accuracy of the entity recognition model.

[0092] In the exemplary embodiment, the name feature can effectively distinguish between business district entities and non-business district entities, but again, Figure 5 , the above non-trade zone entity name features can cover 59% of non-trade zone entities, but 41% of non-trade zone entities still do not contain the above second category of preset words. Figure 4 It can be seen that 18% of the business district entities do not contain the first type of preset words. Therefore, it is necessary to extract other key features to further distinguish business district entities from non-business district entities.

[0093] Since business district entities are usually shopping complexes, they generally gather a large number of different types of stores, such as beauty stores, food stores, clothing stores, shoe and bag stores, jewelry stores, digital home appliances and other electronic stores, and furniture stores. At the same time, there are also a large number of stores of the same type. Non-business district entities are usually some streets, markets or park squares, and generally do not have the above-mentioned multiple / multiple types of stores gathered at the same time. Therefore, the store distribution characteristics can be used as another key feature to distinguish business district entities from non-business district entities.

[0094] Then, in step S330, the store types and the number of stores of each type included in the first type of entity are obtained to determine the store distribution characteristics and obtain the attribute characteristics corresponding to the first type of entity label. And, in step S340, the store types and the number of stores of each type included in the second type of entity are obtained to determine the store distribution characteristics and obtain the attribute characteristics corresponding to the second type of entity label.

[0095] Exemplarily, the first-category entities containing the same words can be grouped first, and then the store keywords (such as store names) contained in each group of entities can be obtained respectively to obtain the store distribution characteristics corresponding to each group of entities. For example, the first-category entities (business district entities) containing "aa Plaza" are taken as the first group, the first-category entities (business district entities) containing "bb City" are taken as the second group, and the first-category entities (business district entities) containing "cc Center" are taken as the third group. In this way, the store keywords contained in the first group can be obtained to obtain the store distribution characteristics corresponding to the first group entity "aa Plaza", and so on, the store keywords contained in the second group are obtained to obtain the store distribution characteristics corresponding to the second group entity "bb City", and the store keywords contained in the third group are obtained to obtain the store distribution characteristics corresponding to the third group entity "cc Center".

[0096] Exemplarily, the dimensions of store attribute features are determined according to the store type. For example, in this embodiment, stores are divided into the following types: beauty type, food type, clothing type, shoes and bags type, jewelry type, digital appliances type and others. Therefore, the store attribute features in this embodiment are 7-dimensional.

[0097] It should be noted that the method for determining the attribute features corresponding to the second type of entity tags is similar to the exemplary embodiment of the attribute features corresponding to the first type of entity tags described above, and will not be repeated here. Figure 3 There is no restriction on the execution order of the steps.

[0098] For example, Table 4 shows the store distribution data of business district entities and non-business district entities in the food type, clothing type, and digital home appliance type, respectively.

[0099] Table 4

[0100]

[0101]

[0102] Referring to Table 4, it can be seen that there is indeed a significant difference in store distribution between business district entities and non-business district entities. Therefore, by extracting the attribute features corresponding to the first type of entity labels, that is, the store distribution features of business district entities and using them to train the entity recognition module, it is possible to more effectively distinguish between business district entities and non-business district entities. For example, the attribute features of "aa Plaza" can be expressed as "911,218,298".

[0103] In an exemplary embodiment, continue to refer to Figure 2 In step S230, the model loss is determined based on the name features and attribute features corresponding to the first category entity tags and the name features and attribute features corresponding to the second category entity tags.

[0104] Exemplarily, the name features and attribute features corresponding to the first type of entity tags extracted in the above embodiment and the name features and attribute features corresponding to the second type of entity tags are input into a binary classification model to classify and identify business district entities and non-business district entities. For example, a Light Gradient Boosting Machine (LightGBM) model is used, which is equivalent to an integrated model of multiple decision trees.

[0105] Specifically, Figure 6 A flow chart of a method for determining model loss in an exemplary embodiment of the present disclosure is shown, comprising:

[0106] Step S610, for the ith decision tree, obtain the ith feature set, calculate the information entropy when any feature in the ith feature set is used as a split node, and use the feature corresponding to the maximum information entropy difference as the root node of the ith decision tree, where i is a positive integer not greater than the total number of decision trees, and the ith feature set includes the name features and attribute features corresponding to the first category entity labels and the name features and attribute features corresponding to the second category entity labels.

[0107] The more chaotic the data is, the greater the information entropy is, and the more ordered the data is, the smaller the information entropy is.

[0108] Exemplarily, for the name features (such as including 6 dimensions) of the business district entity, the attribute features (such as including 7 dimensions) corresponding to the business district entity, and the name features (such as including 11 dimensions) corresponding to the non-business district entity, there are 24-dimensional features as a whole in this embodiment. For any one of the multiple decision trees, multiple dimensions (such as 15 dimensions) are obtained from the above 24-dimensional features as the feature set corresponding to the decision tree, and the information entropy is reduced the most after the split node in the feature set is calculated, and the node is used as the root node.

[0109] Step S620, after excluding the feature corresponding to the root node of the i-th decision tree from the i-th feature set, calculate the information entropy when any feature is used as a split node, and use the feature corresponding to the maximum value of the information entropy difference as the intermediate node of the i-th decision tree until the preset splitting height or preset information entropy difference of the i-th decision tree is reached.

[0110] Still taking the above embodiment as an example, after excluding the feature corresponding to the root node from the feature set, the information entropy of any feature as a split node is calculated again, and the feature corresponding to the maximum value of the information entropy difference is used as the intermediate node of the decision tree, and so on until the pre-required split height or information entropy difference corresponding to the decision tree is reached.

[0111] Step S630, obtain the first prediction label corresponding to the leaf node determined when the decision is made with the root node, and obtain the second prediction label corresponding to the leaf node determined when the decision is made with the intermediate node, and determine the model loss corresponding to the i-th decision tree based on the first prediction label, the second prediction label and the actual label.

[0112] Exemplarily, the difference between the first predicted label and the actual label corresponding to the leaf node is calculated to determine the first loss, and the difference between the second predicted label and the actual label corresponding to the relevant leaf node is calculated to determine the second loss. Furthermore, the model loss corresponding to the i-th decision tree is determined according to the first loss and the second loss. In an exemplary embodiment, the cross entropy loss function is used as the loss function of the entity recognition model, and the gradient descent method (StochasticGradient Descent, referred to as: SGD) is used to process the loss function to iteratively calculate the model parameters of the entity recognition model. Exemplarily, the specific parameters for training the above-mentioned entity recognition model can be as follows: learning rate 0.07, feature selection ratio 0.8.

[0113] Still taking the above embodiment as an example, further, the i+1th decision tree is constructed based on the probability loss of the i decision trees until the preset number of trees is reached. The root node and the intermediate node in each decision tree are determined as described in the above embodiment.

[0114] It should be noted that the number of decision trees, the split height or information entropy difference corresponding to each decision tree are determined by adjusting parameters.

[0115] In an exemplary embodiment, the entity recognition model after the above iterative optimization is evaluated by one or more of the following model evaluation indicators: accuracy, recall rate and receiver operating characteristic curve (ROC) area under the curve AUC (a model evaluation indicator, specifically used to evaluate the predictive value of the model; Area Under Curve is short for Area Under Curve). Specifically:

[0116] Exemplarily, after iterative optimization through training samples, the entity recognition model after iterative optimization (referred to as "model to be tested") is tested through samples in the test set (Table 3). Specifically, the following tests are performed for different parameters (including: the number of decision trees, the split height or information entropy difference corresponding to each decision tree, etc.), and then the optimal value corresponding to the parameter is obtained, and the entity recognition model with the optimal value of the parameter is used to identify and screen business district entities. For example, when the number of decision trees is 50 / 100 / 150, the samples in the test set are tested to obtain multiple test results (including the results corresponding to the following multiple test indicators: accuracy, recall, true positive TP, true negative TN, false negative FN and false positive FP, AUC, ROC curve).

[0117] In an exemplary embodiment, the testing process for a certain parameter (such as the number of decision trees) is as follows:

[0118] According to the test sample input to the model to be tested, the output data of the model is as follows: true positive TP, true negative TN, false negative FN and false positive FP. Among them, TP is the number of positive classes that are still positive after the positive class in the test set is judged by the model to be tested, TN is the number of negative classes after the positive class in the test set is judged by the model to be tested, FN is the number of negative classes after the negative class in the test set is judged by the model to be tested, and FP is the number of positive classes after the negative class in the test set is judged by the model to be tested. Positive and negative classes refer to the two categories manually annotated for the samples in the first part, that is, if a sample is manually annotated to belong to a specific class, then the sample belongs to the positive class, and samples that do not belong to the specific class belong to the negative class.

[0119] Secondly, the test results of the model to be tested are calculated according to the true positive TP, true negative TN, false negative FN and false positive FP.

[0120] The precision p and recall r are calculated according to the following two formulas respectively;

[0121] p=TP / (TP+FP)

[0122] r=TP / (TP+FN)

[0123] If the preset requirements corresponding to the test indicators are: if the accuracy test result is greater than p' (preset value), the preset requirements of the test indicator as accuracy are met, otherwise the preset requirements of the test indicator as accuracy are not met, and if the recall test result is greater than r' (preset value), the preset requirements of the test indicator as recall are met, otherwise the preset requirements of the test indicator as recall are not met.

[0124] In an exemplary embodiment, when the test indicator is AUC, the false positive rate FPR and the true positive rate TPR are determined using the following two formulas:

[0125] FPR=FP / (FP+TN)

[0126] TPR=TP / (TP+FN)

[0127] Furthermore, the receiver operating characteristic curve (ROC curve) is plotted with FPR as the horizontal axis and TPR as the vertical axis. The ROC curve is the characteristic curve of each indicator obtained, which is used to show the relationship between each indicator, and further calculates the area under the ROC curve AUC. The ROC curve is the characteristic curve of each indicator obtained, which is used to show the relationship between each indicator. AUC is the area under the ROC curve. The larger the AUC, the higher the predictive value of the model, and the model to be tested can be tested by AUC.

[0128] For example, if the number of decision trees is 100 and the test results meet the preset requirements of each test index, the model with 100 decision trees is used for predicting business district entities. For example, the model can eventually obtain an accuracy rate of nearly 89%.

[0129] In an exemplary embodiment, the entity recognition model that meets the prediction test index after testing can be used to identify and screen business district entities. Figure 7 FIG. 1 is a flow chart showing a method for identifying business district entities in an exemplary embodiment of the present disclosure. Figure 7 ,include:

[0130] Step S710, performing named entity recognition on the text to be recognized to determine the entity to be recognized; step S720, performing feature extraction on the entity to be recognized to obtain name features and attribute features corresponding to the entity to be recognized; and step S730, inputting the name features and the attribute features into a trained entity recognition model, so that the entity recognition model predicts whether the entity to be recognized is a business district entity or a non-business district entity based on the name features and the attribute features.

[0131] The exemplary implementation of performing named entity recognition on the text to be recognized to obtain the entity to be recognized in step S710 has been introduced in the aforementioned sample processing embodiment and will not be described in detail here.

[0132] In step S720, feature extraction is performed on the entity to be identified to obtain the name feature and attribute feature corresponding to the entity to be identified. Figure 8 As shown:

[0133] Step S721, extracting the preset words of the business district contained in the entity to be identified or extracting the preset words of the non-business district contained in the entity to be identified, and obtaining the name features corresponding to the entity to be identified; and step S722, obtaining the store types contained in the entity to be identified and the number of stores of each type to determine the store distribution characteristics, and obtaining the attribute features corresponding to the entity to be identified.

[0134] Among them, the exemplary implementation method of extracting the features of the entity to be identified according to the preset words in step S721 has been Figure 3 The corresponding embodiments have been introduced and will not be described again here.

[0135] Referring to Table 5, the prediction results for the entity to be identified are shown in the last column of Table 5. Among them, "1" indicates that the entity to be identified is a business district entity, and "0" indicates that the entity to be identified is a non-business district entity.

[0136] Table 5

[0137]

[0138] In the technical solution provided by the present disclosure, automatic recognition of business district entities is achieved based on artificial intelligence technology, thereby avoiding the problems of low recognition efficiency, low recognition accuracy and difficult maintenance of screening rules caused by specifying screening rules and manually identifying business district entities based on the screening rules. At the same time, in the model training process provided by the present technical solution, in order to effectively distinguish whether the entity to be identified belongs to the business district entity, the acquired entity set is subjected to feature extraction to obtain the name features and attribute features corresponding to the first type of entity labels and the name features and attribute features corresponding to the second type of entity labels. The model loss is determined according to the above-mentioned key features for distinguishing between business district entities and non-business district entities, so as to effectively distinguish between business district entities and non-business district entities through their respective characteristics, so as to effectively improve the accuracy of entity recognition.

[0139] Those skilled in the art will appreciate that all or part of the steps to implement the above embodiments are implemented as a computer program executed by a processor (including a GPU / CPU). When the computer program is executed by the GPU / CPU, the above functions defined by the above method provided in the present disclosure are performed. The program may be stored in a computer-readable storage medium, which may be a read-only memory, a disk or an optical disk, etc.

[0140] In addition, it should be noted that the above figures are only schematic illustrations of the processes included in the method according to the exemplary embodiment of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0141] The following Fig. 9 An embodiment of a business district entity recognition device disclosed in the present invention is introduced, which can be used to execute the business district entity recognition method disclosed above.

[0142] Fig. 9 A schematic diagram of the structure of a business district entity identification device in an exemplary embodiment of the present disclosure is shown.

[0143] like Fig. 9 As shown, the above-mentioned business district entity recognition device 900 includes: a module 901 for determining an entity to be recognized, which is configured to: perform named entity recognition on the text to be recognized to determine the entity to be recognized; a module 902 for processing an entity to be recognized, which is configured to: perform feature extraction on the above-mentioned entity to be recognized to obtain name features and attribute features about the above-mentioned entity to be recognized; and a recognition module 903, which is configured to: input the above-mentioned name features and the above-mentioned attribute features into a trained entity recognition model, so that the above-mentioned entity recognition model predicts that the above-mentioned entity to be recognized is a business district entity or a non-business district entity based on the above-mentioned name features and the above-mentioned attribute features.

[0144] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned entity processing module 902 to be identified is specifically configured to: extract the preset words of the business district contained in the above-mentioned entity to be identified or extract the preset words of the non-business district contained in the above-mentioned entity to be identified to obtain the name features corresponding to the above-mentioned entity to be identified; obtain the store types contained in the above-mentioned entity to be identified and the number of stores of each type to determine the store distribution characteristics, and obtain the attribute features corresponding to the above-mentioned entity to be identified.

[0145] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned device 900 also includes: a model training module 904, and the above-mentioned model training module 904 includes: an entity determination submodule 9041, a feature extraction submodule 9042, a loss determination submodule 9043 and an iterative calculation submodule 9044.

[0146] Among them, the above-mentioned entity determination submodule 9041 is configured to: determine the entity set about the positive sample according to the identity information about multiple stores, and obtain the entity set about the negative sample; the above-mentioned feature extraction submodule 9042 is configured to: extract features from the above-mentioned entity set to obtain the name features and attribute features corresponding to the first category entity tags, and the name features and attribute features corresponding to the second category entity tags; the above-mentioned loss determination submodule 9043 is configured to: determine the model loss according to the name features and attribute features corresponding to the above-mentioned first category entity tags and the name features and attribute features corresponding to the above-mentioned second category entity tags; and the above-mentioned iterative calculation submodule 9044 is configured to: iteratively calculate the model parameters of the above-mentioned entity recognition model according to the above-mentioned model loss to optimize the above-mentioned model parameters and obtain the above-mentioned trained entity recognition model.

[0147] In some embodiments of the present disclosure, based on the aforementioned solution, the feature extraction submodule 9042 includes: an entity classification unit, a first extraction unit, a second extraction unit, a third extraction unit and a fourth extraction unit.

[0148] Among them, the above-mentioned entity classification unit is configured to: obtain the first category entity and the second category entity according to the label of the entity in the above-mentioned entity set; the above-mentioned first extraction unit is configured to: extract the first category preset words contained in the above-mentioned first category entity, and obtain the name features corresponding to the first category entity label; the above-mentioned second extraction unit is configured to: extract the second category preset words contained in the above-mentioned second category entity, and obtain the name features corresponding to the second category entity label; the above-mentioned third extraction unit is configured to: obtain the store types contained in the above-mentioned first category entity and the number of stores of each type to determine the store distribution characteristics, and obtain the attribute characteristics corresponding to the first category entity label; the above-mentioned fourth extraction unit is configured to: obtain the store types contained in the above-mentioned second category entity and the number of stores of each type to determine the store distribution characteristics, and obtain the attribute characteristics corresponding to the second category entity label.

[0149] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned first extraction unit is specifically configured to: determine the word at the end position in the above-mentioned first category entity as the first target word; if the similarity between the above-mentioned first target word and the above-mentioned first category preset word is greater than a preset threshold, then extract the above-mentioned first target word to obtain the name feature corresponding to the first category entity label.

[0150] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned second extraction unit is specifically configured to: determine the word at the end position in the above-mentioned second category entity as the second target word; if the similarity between the above-mentioned second target word and the above-mentioned second category preset word is greater than a preset threshold, then extract the above-mentioned second target word to obtain the name feature corresponding to the second category entity label.

[0151] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned third extraction unit is specifically configured to: group the first category entities containing the same words; respectively obtain the store keywords contained in each group of entities, and obtain the store distribution characteristics corresponding to each group of entities.

[0152] In some embodiments of the present disclosure, based on the above-mentioned scheme, the above-mentioned loss determination submodule 9043 is specifically configured as follows: for the ith decision tree, obtain the ith feature set, calculate the information entropy when any feature in the above-mentioned ith feature set is used as a split node, and use the feature corresponding to the maximum value of the information entropy difference as the root node of the above-mentioned ith decision tree, wherein the value of i is a positive integer not greater than the total number of preset decision trees, and the above-mentioned ith feature set includes the name feature and attribute feature corresponding to the above-mentioned first category entity label and the name feature and attribute feature corresponding to the above-mentioned second category entity label; after excluding the feature corresponding to the root node of the above-mentioned ith decision tree in the above-mentioned ith feature set, calculate the information entropy when any feature is used as a split node, and use the feature corresponding to the maximum value of the information entropy difference as the intermediate node of the above-mentioned ith decision tree until the preset split height or preset information entropy difference of the above-mentioned ith decision tree is reached; obtain the first predicted label corresponding to the leaf node determined when the decision is made with the above-mentioned root node, and obtain the second predicted label corresponding to the leaf node determined when the decision is made with the above-mentioned intermediate node, and determine the model loss corresponding to the above-mentioned ith decision tree according to the above-mentioned first predicted label, the above-mentioned second predicted label and the actual label.

[0153] In some embodiments of the present disclosure, based on the aforementioned scheme, after determining the above-mentioned model loss, the above-mentioned iterative calculation submodule 9044 is specifically configured as: determining the cross entropy loss function as the loss function of the above-mentioned entity recognition model according to the above-mentioned model loss; and using the stochastic gradient descent method to process the above-mentioned loss function to iteratively calculate the model parameters of the above-mentioned entity recognition model.

[0154] In some embodiments of the present disclosure, the above-mentioned entity determination submodule 9041 is specifically configured as follows: inputting the above-mentioned identity information about multiple stores into a pre-trained named entity recognition model; based on the above-mentioned pre-trained named entity recognition model, extracting target type named entities from the unstructured identity information text to obtain the above-mentioned entity set about the positive sample.

[0155] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned identity information includes one or more of the following information: store name, store address, store category, store WiFi name, and longitude and latitude of the store location.

[0156] In some embodiments of the present disclosure, based on the aforementioned scheme, the above-mentioned first category entity tags are business district tags, and the above-mentioned second category entity tags are non-business district tags, wherein the above-mentioned first category preset words include one or more of the following information: square, city, center, village and gathering; the above-mentioned second category preset words include one or more of the following information: town, block, courtyard, shop, lane and number.

[0157] The specific details of each unit in the above-mentioned business district entity identification device have been described in detail in the above-mentioned business district entity identification method embodiment, so they will not be repeated here.

[0158] Fig.10 A schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present disclosure is shown.

[0159] It should be noted that Fig.10 The computer system 1000 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0160] like Fig.10 As shown, the computer system 1000 includes a processor 1001, wherein the processor 1001 may specifically include: a graphics processing unit (Graphics Processing Unit, referred to as: GPU) and a central processing unit (Central Processing Unit, referred to as: CPU), which can perform various appropriate actions and processes according to the program stored in the read-only memory (Read-Only Memory, referred to as: ROM) 1002 or the program loaded from the storage part 1008 to the random access memory (Random Access Memory, referred to as: RAM) 1003. In RAM 1003, various programs and data required for system operation are also stored. The processor 1001, ROM 1002 and RAM 1003 are connected to each other through a bus 1004. An input / output (Input / Output, referred to as: I / O) interface 1005 is also connected to the bus 1004.

[0161] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a local area network (LAN) card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed so that a computer program read therefrom is installed into the storage section 1008 as needed.

[0162] In particular, according to an embodiment of the present disclosure, the process described below with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 1009, and / or installed from a removable medium 1011. When the computer program is executed by the processor 1001, various functions defined in the system of the present application are executed.

[0163] It should be noted that the computer-readable medium shown in the embodiment of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0164] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0165] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0166] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the method described in the above embodiment.

[0167] For example, the electronic device can implement Figure 7 As shown in: Step S710, performing named entity recognition on the text to be recognized to determine the entity to be recognized; Step S720, performing feature extraction on the entity to be recognized to obtain name features and attribute features corresponding to the entity to be recognized; and Step S730, inputting the name features and the attribute features into the trained entity recognition model, so that the entity recognition model predicts that the entity to be recognized is a business district entity or a non-business district entity based on the name features and the attribute features.

[0168] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0169] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0170] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0171] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for identifying a business district entity, characterized in that: The method comprises: Perform named entity recognition on the text to be recognized to determine the entity to be recognized; Extracting preset words of the business district type contained in the entity to be identified or extracting preset words of the non-business district type contained in the entity to be identified to obtain a name feature corresponding to the entity to be identified; Obtaining the store types and the number of stores of each type contained in the entity to be identified, so as to determine the store distribution characteristics and obtain the attribute characteristics corresponding to the entity to be identified; The name feature and the attribute feature are input into a trained entity recognition model, so that the entity recognition model predicts whether the entity to be recognized is a business district entity or a non-business district entity according to the name feature and the attribute feature.

2. The method for identifying a business district entity according to claim 1, characterized in that: The method further comprises: Determine an entity set about positive samples according to identity information about multiple stores, and obtain an entity set about negative samples; Performing feature extraction on the entity set to obtain name features and attribute features corresponding to the first type of entity tags, and name features and attribute features corresponding to the second type of entity tags; Determining a model loss according to the name features and attribute features corresponding to the first category entity labels and the name features and attribute features corresponding to the second category entity labels; The model parameters of the entity recognition model are iteratively calculated according to the model loss to optimize the model parameters and obtain the trained entity recognition model.

3. The method for identifying a business district entity according to claim 2, characterized in that: Performing feature extraction on the entity set to obtain name features and attribute features corresponding to the first type of entity tags, and name features and attribute features corresponding to the second type of entity tags, including: Extracting first-category preset words contained in the first-category entity to obtain name features corresponding to the first-category entity label, wherein the first-category entity is determined according to the label of the entity in the entity set; Extracting the second category of preset words contained in the second category of entities to obtain name features corresponding to the second category of entity labels, wherein the second category of entities is determined according to the labels of entities in the entity set; Obtaining the store types and the number of stores of each type included in the first type of entities to determine store distribution characteristics and obtain attribute characteristics corresponding to the first type of entity labels; The store types and the number of stores of each type included in the second type of entity are obtained to determine the store distribution characteristics and obtain the attribute characteristics corresponding to the second type of entity label.

4. The method for identifying a business district entity according to claim 3, characterized in that: The first type of preset words contained in the first type of entity are extracted to obtain the name features corresponding to the first type of entity label, including: Determine the word at the end position of the first type of entity as the first target word; If the first target word is the same as the first category of preset words, or the first target word includes the first category of preset words, the first target word is extracted to obtain a name feature corresponding to the first category of entity tags.

5. The method for identifying a business district entity according to claim 3, characterized in that: The second category of preset words contained in the second category of entities are extracted to obtain name features corresponding to the second category of entity labels, including: Determine the word at the end position in the second type of entity as the second target word; If the second target word is the same as the second type of preset word, or the second target word includes the second type of preset word, the second target word is extracted to obtain a name feature corresponding to the second type of entity tag.

6. The method for identifying a business district entity according to claim 3, characterized in that: The store types and the number of stores of each type included in the first type of entities are obtained to determine store distribution characteristics, including: Group the first-category entities that contain the same words; The store keywords contained in each group of entities are obtained respectively, and the store distribution characteristics corresponding to each group of entities are obtained.

7. The method for identifying a business district entity according to claim 2, characterized in that: Determining a model loss according to the name features and attribute features corresponding to the first category entity labels and the name features and attribute features corresponding to the second category entity labels, including: For the i-th decision tree, obtain the i-th feature set, calculate the information entropy when any feature in the i-th feature set is used as a split node, and use the feature corresponding to the maximum value of the information entropy difference as the root node of the i-th decision tree, wherein the value of i is a positive integer not greater than the total number of preset decision trees, and the i-th feature set includes the name features and attribute features corresponding to the first type of entity labels and the name features and attribute features corresponding to the second type of entity labels; After excluding the feature corresponding to the root node of the i-th decision tree from the i-th feature set, calculate the information entropy when any feature is used as a split node, and use the feature corresponding to the maximum value of the information entropy difference as the intermediate node of the i-th decision tree until the preset splitting height or the preset information entropy difference of the i-th decision tree is reached; Obtain a first prediction label corresponding to a leaf node determined when making a decision with the root node, and obtain a second prediction label corresponding to a leaf node determined when making a decision with the intermediate node, and determine the model loss corresponding to the i-th decision tree according to the first prediction label, the second prediction label and the actual label.

8. The method for identifying a business district entity according to claim 2, characterized in that: After determining the model loss, the method further comprises: Determining a cross entropy loss function as a loss function of the entity recognition model according to the model loss; Iteratively calculating the model parameters of the entity recognition model according to the model loss, including: The loss function is processed by a stochastic gradient descent method to iteratively calculate the model parameters of the entity recognition model.

9. The method for identifying a business district entity according to any one of claims 2 to 8, characterized in that: The entity set of the positive samples is determined based on the identity information of multiple stores, including: Inputting the identity information about the plurality of stores into a pre-trained named entity recognition model; Based on the pre-trained named entity recognition model, named entities of target types are extracted from unstructured identity information text to obtain a set of entities related to positive samples.

10. The method for identifying a business district entity according to any one of claims 2 to 8, characterized in that: The identity information includes one or more of the following information: store name, store address, store category, store WiFi name, and store location latitude and longitude.

11. The method for identifying a business district entity according to any one of claims 3 to 5, characterized in that: The first type of entity tags are business district tags, and the second type of entity tags are non-business district tags, wherein the first type of preset words include one or more of the following information: square, city, center, village and gathering; the second type of preset words include one or more of the following information: town, block, courtyard, shop, road, lane and number.

12. A business district entity recognition device, characterized in that: The device comprises: The to-be-recognized entity determination module is configured to: perform named entity recognition on the to-be-recognized text to determine the to-be-recognized entity; The entity processing module to be identified is configured to: extract the preset words of the business district contained in the entity to be identified or extract the preset words of the non-business district contained in the entity to be identified to obtain the name features corresponding to the entity to be identified; obtain the store types and the number of stores of each type contained in the entity to be identified to determine the store distribution features, and obtain the attribute features corresponding to the entity to be identified; The recognition module is configured to: input the name feature and the attribute feature into a trained entity recognition model, so that the entity recognition model predicts whether the entity to be recognized is a business district entity or a non-business district entity based on the name feature and the attribute feature.

13. A computer-readable storage medium, characterized in that: A computer program is stored thereon; When the computer program is executed by a processor, the method for identifying business district entities as described in any one of claims 1 to 11 is implemented.

14. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device, used to store one or more programs, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the business district entity recognition method as described in any one of 1 to 11.

15. A computer program product, characterized in that The computer program product includes computer instructions, which are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the business district entity recognition method as described in any one of 1 to 11.

Images