CN115858695A - Information processing method and device and storage medium - Google Patents
Information processing method and device and storage medium Download PDFInfo
- Publication number
- CN115858695A CN115858695A CN202211565452.5A CN202211565452A CN115858695A CN 115858695 A CN115858695 A CN 115858695A CN 202211565452 A CN202211565452 A CN 202211565452A CN 115858695 A CN115858695 A CN 115858695A
- Authority
- CN
- China
- Prior art keywords
- attribute
- information
- initial
- target
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the application discloses an information processing method, an information processing device and a storage medium, wherein the information processing method comprises the following steps: under the condition that the target object is received, acquiring an object detail graph corresponding to the target object; inputting the object detail graph into an attribute detection model to obtain initial attribute information of a target object; classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values; and determining a plurality of corresponding relations between the plurality of attribute names and the plurality of attribute values by using the relation recognition model, and using the plurality of corresponding relations as target attribute information of the target object.
Description
Technical Field
The present application relates to the field of internet technologies, and in particular, to an information processing method and apparatus, and a storage medium.
Background
Due to the special application scene of the industrial goods, the attribute parameters of the goods are always strictly required to be consistent with the requirements, so that the completeness of the attribute parameters of the goods is very important for manual identification and intelligent matching.
In the prior art, the attribute information of the commodity is manually acquired from the detailed quotient graph and written into the database, and because the detailed quotient graphs are large in quantity, the attribute information acquired from the detailed quotient graphs manually often has wrong parameters, namely, the accuracy of acquiring the attribute information is reduced.
Disclosure of Invention
In order to solve the above technical problem, embodiments of the present application are expected to provide an information processing method and apparatus, and a storage medium, which can improve accuracy in acquiring target attribute information.
The technical scheme of the application is realized as follows:
an embodiment of the present application provides an information processing method, including:
under the condition that a target object is received, acquiring an object detail graph corresponding to the target object;
inputting the object detail graph into an attribute detection model to obtain initial attribute information of the target object;
classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values;
and determining a plurality of corresponding relations between the plurality of attribute names and the plurality of attribute values by using a relation recognition model, and using the plurality of corresponding relations as target attribute information of the target object.
An embodiment of the present application provides an information processing apparatus, the device comprises:
the device comprises an acquisition unit, a processing unit and a display unit, wherein the acquisition unit is used for acquiring an object detail graph corresponding to a target object under the condition of receiving the target object;
the input unit is used for inputting the object detail graph into an attribute detection model to obtain initial attribute information of the target object;
the classification unit is used for classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values;
a determining unit configured to determine a plurality of correspondence relationships between the plurality of attribute names and the plurality of attribute values using a relationship recognition model, and to use the plurality of correspondence relationships as target attribute information of the target object.
An embodiment of the present application provides an information processing apparatus, the apparatus including:
the information processing system includes a memory, a processor, and a communication bus, the memory communicating with the processor through the communication bus, the memory storing an information processing program executable by the processor, and the processor executing the information processing method when the information processing program is executed.
The embodiment of the application provides a storage medium, which stores a computer program thereon and is applied to an information processing device, wherein the computer program is used for realizing the information processing method when being executed by a processor.
The embodiment of the application provides an information processing method, an information processing device and a storage medium, wherein the information processing method comprises the following steps: under the condition that the target object is received, acquiring an object detail graph corresponding to the target object; inputting the object detail graph into an attribute detection model to obtain initial attribute information of a target object; classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values; and determining a plurality of corresponding relations between the plurality of attribute names and the plurality of attribute values by using the relation recognition model, and using the plurality of corresponding relations as target attribute information of the target object. By adopting the method implementation scheme, the information processing device acquires the initial attribute information of the target object from the object detail map by using the attribute detection model, establishes a plurality of corresponding relations between a plurality of attribute names and a plurality of attribute values for a plurality of attribute names and a plurality of attribute values determined from the initial attribute information by using the attribute detection model, the attribute classification model and the relation identification model, does not need to acquire the target attribute information of the target object from the object detail map manually, and can acquire the target attribute information from the object detail map accurately by using the information processing method under the condition that the number of the object detail maps is large, namely, the accuracy in acquiring the target attribute information is improved.
Drawings
Fig. 1 is a flowchart of an information processing method according to an embodiment of the present application;
fig. 2 is a first schematic diagram of an exemplary information processing flow provided in an embodiment of the present application;
fig. 3 is a schematic diagram of an exemplary information processing flow provided in an embodiment of the present application;
fig. 4 is a first schematic structural diagram of an information processing apparatus according to an embodiment of the present disclosure;
fig. 5 is a schematic diagram illustrating a structure of an information processing apparatus according to an embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
An information processing method is provided in an embodiment of the present application, and an information processing method is applied to an information processing apparatus, and fig. 1 is a flowchart of the information processing method provided in the embodiment of the present application, and as shown in fig. 1, the information processing method may include:
s101, under the condition that the target object is received, an object detail graph corresponding to the target object is obtained.
The information processing method provided by the embodiment of the application is suitable for a scene of acquiring the target attribute information from the object detail drawing of the target object.
In the embodiment of the present application, the information processing apparatus may be implemented in various forms. For example, the information processing apparatus described in the present application may include apparatuses such as a mobile phone, a camera, a tablet computer, a notebook computer, a palmtop computer, a Personal Digital Assistant (PDA), a Portable Media Player (PMP), a navigation apparatus, a wearable device, a smart band, a pedometer, and the like, and apparatuses such as a Digital TV, a desktop computer, a server, and the like.
In the embodiment of the application, the target object can be a commodity on an e-commerce platform. The object detail view may be a merchant detail view of the item.
In this embodiment of the present application, the number of the object detail diagrams may be multiple, and the specific number of the object detail diagrams may be determined according to an actual situation, which is not limited in this embodiment of the present application.
In the embodiment of the application, a display screen is arranged in the information processing device, and the information processing device can receive the target object transmitted by the user from the display screen. The information processing device is provided with an information transmission interface, so that the information processing device receives a target object transmitted by other equipment from the information transmission interface; the information processing apparatus may also receive the target object in other manners; the specific manner in which the information processing apparatus receives the target object may be determined according to actual situations, which is not limited in this embodiment of the present application.
In the embodiment of the present application, the information processing apparatus may acquire an object detail view of the target object from a merchant detail page of the commodity; the information processing apparatus may also acquire an object detail map of the target object from another device; the specific manner in which the information processing apparatus acquires the object detail map of the target object may be determined according to actual situations, and is not limited in the embodiment of the present application.
It should be noted that, in order to fully display each commodity, the merchant may equip each commodity with a plurality of detailed merchant drawings, that is, a group of detailed merchant drawings, so that one commodity corresponds to a group of detailed merchant drawings. In other words, one target object corresponds to one set of object detail graphs.
For example, a group of detailed quotient figures can number from 30 to 50; the number of the quotient detail graphs can be other numbers; the number of the specific group of quotient detail drawings can be determined according to actual conditions, and the embodiment of the application does not limit the number.
And S102, inputting the object detail graph into the attribute detection model to obtain initial attribute information of the target object.
In this embodiment of the application, after the information processing apparatus acquires the object detail graph corresponding to the target object, the information processing apparatus may input the object detail graph into the attribute detection model to obtain initial attribute information of the target object.
In the embodiment of the present application, the attribute detection model may be a model configured in the information processing apparatus; the attribute detection model can also be a model obtained by training the information processing device; the attribute detection model may be a model acquired by the information processing apparatus in another manner; the specific manner in which the information processing apparatus obtains the attribute detection model may be determined according to actual conditions, which is not limited in the embodiment of the present application.
It should be noted that the attribute detection model may be a YOLO _ V3 (young Only Live Once-third edition) neural network; the attribute detection model can be other models; the specific attribute detection model may be determined according to actual conditions, which is not limited in the embodiment of the present application.
In this embodiment of the present application, the initial attribute information may be attribute information of a target object in the object detail diagram.
In this embodiment of the present application, a process of inputting an object detail drawing into an attribute detection model by an information processing apparatus to obtain initial attribute information of a target object includes: determining an attribute information frame in the object detail graph by using an attribute detection model; and acquiring information in a frame-defined range of the attribute information in the object detail drawing to obtain initial attribute information.
In the embodiment of the present application, the information processing apparatus may obtain information within a defined range of an attribute information frame in the object detail drawing to obtain the initial attribute information, and perform Character Recognition on the information within the defined range of the attribute information frame by using an Optical Character Recognition (OCR) technique for the information processing apparatus to obtain the initial attribute information; or using computer vision technology to identify the character of the information in the frame range of the attribute information frame, so as to obtain the initial attribute information; other technologies can also be utilized to perform character recognition on the information within the frame-defined range of the attribute information frame, so as to obtain initial attribute information; the specific information processing apparatus obtains information in the range defined by the attribute information frame in the object detail drawing, and the manner of obtaining the initial attribute information may be determined according to the actual situation, which is not limited in the embodiment of the present application.
In an embodiment of the present application, a process of determining an attribute information frame in an object detail drawing by an information processing apparatus using an attribute detection model includes: determining an initial information frame in the object detail graph by using an attribute detection model; and extending each edge of the initial information frame according to the direction far away from the central point of the initial information frame to obtain the attribute information frame.
In the embodiment of the present application, the information processing apparatus may output the initial information frame by using the attribute detection model by inputting the object detail map into the attribute detection model.
The initial information frame is a picture with an attribute frame.
In this embodiment of the present application, the information processing apparatus may extend each edge of the initial information frame in a direction away from the center point of the initial information frame to obtain an attribute information frame; the initial information frame can also be extended according to other frame extension modes to obtain an attribute information frame; the specific information processing apparatus extends the initial information frame, and the manner of obtaining the attribute information frame may be determined according to actual conditions, which is not limited in the embodiment of the present application.
In this embodiment, the information processing apparatus may extend each edge of the initial information frame in the manner of formula (1) to obtain the attribute information frame. For example, the position coordinate of the initial information frame may be [ X ] min ，Y min ，X max ，Y min ]In order to reduce the interference of noise information in the quotient detailed graph and reduce the frame identification error of the YOLO _ V3 model, the initial information frame is extended, and the finally obtained attribute information frame (bbox) coordinate is shown as formula (1):
X min ＝max(0，X min -x/10)
Y min ＝max(0，T min -y/10)
X max ＝min(x，X max +x/10)
Y max ＝min(y，Y max +y/10) (1)
it should be noted that x and y are the width and length of the detailed drawings, respectively.
It can be understood that, in order to fully display each commodity, a merchant may equip 30 to 50 detailed drawings for the commodity, and most of the 30 to 50 detailed drawings contain descriptive texts, which have strong interference, so that it is necessary to locate an attribute information frame in the detailed drawings by using an attribute detection model, so as to effectively filter useless information (information in the detailed drawings that does not belong to attribute information) and locate attribute information, thereby reducing interference of the useless information in the detailed drawings and improving accuracy in determining initial attribute information (attribute information) of a target object (commodity).
In this embodiment of the present application, the process of acquiring, by an information processing apparatus, information within a range defined by an attribute information frame in an object detail drawing to obtain initial attribute information includes: identifying information in a defined range of an attribute information frame in the object detail graph to obtain initial information; segmenting the initial information by using preset characters to obtain a plurality of segmentation information; and recombining the plurality of segmentation information according to the positions of the plurality of segmentation information in the object detail graph and the intervals among the plurality of segmentation information to obtain initial attribute information.
In the embodiment of the application, the information processing device can identify information in a defined range of an attribute information frame in the object detail graph by using an OCR technology to obtain initial information; or identifying information in the range framed by the attribute information frame in the object detail graph by using a computer vision technology to obtain initial information; the information in the range framed by the attribute information frame in the object detail graph can be identified by other modes to obtain initial information; the specific information processing apparatus identifies information within the range defined by the attribute information frame in the object detail diagram, and the manner of obtaining the initial information may be determined according to the actual situation, which is not limited in the embodiment of the present application.
In the embodiment of the present application, the preset character may be a character configured in the information processing apparatus; the preset characters can also be characters transmitted to the information processing device by other equipment; the preset character can also be a character obtained by the information processing device in other modes; the specific way of obtaining the preset character by the information processing apparatus may be determined according to actual conditions, which is not limited in the embodiment of the present application.
It should be noted that the preset characters can be characters such as semicolon, colon, period, pause, and the like; the preset characters are also other characters; the specific preset characters can be determined according to actual conditions, and the embodiment of the application does not limit the specific preset characters.
For example, the preset character may be a colon, and the initial information includes: "last name" and "first name: li IV "; the information processing apparatus may first segment the initial information (first name: liquad) using preset characters to obtain 2 pieces of segmented information ("first name" and "liquad"), and then recombine "first name", "first name" and "liquad" to obtain two entities of "name" and "liquad" as initial attribute information.
It is understood that, due to the parameter alignment and layout problem, there are often large intervals between words in the quotient detailed diagram (e.g., "name: lie four"), or the attribute name and the attribute value cannot be divided (e.g., "name: lie four"). Therefore, the initial information needs to be segmented by using preset characters to obtain a plurality of segmentation information; and then recombining the plurality of segmentation information according to the positions of the plurality of segmentation information in the object detail graph and the intervals among the plurality of segmentation information, thereby obtaining accurate initial attribute information, namely improving the accuracy of determining the initial attribute information.
In the embodiment of the application, before the information processing device inputs the object detail graph into the attribute detection model and obtains the initial attribute information of the target object, the information processing device also obtains a sample detail graph corresponding to the sample object; and training an initial attribute detection model by using the sample detail drawing and the sample attribute box to obtain an attribute detection model.
It should be noted that, the sample attribute box is labeled in the sample detail drawing.
In the embodiment of the present application, the information processing apparatus trains the initial attribute detection model by using the sample detail drawing and the sample attribute box to obtain the attribute detection model, and may input the sample detail drawing into the initial attribute detection model for the information processing apparatus to obtain an output attribute box; and the information processing device determines the model loss of the initial attribute detection model according to the sample attribute frame and the output attribute frame, and takes the initial attribute detection model as the attribute detection model under the condition that the model loss is less than or equal to a preset loss value. And under the condition that the model loss is greater than the preset loss value, continuously training the initial attribute detection model by using the sample detail graph and the sample attribute box to obtain a training model, and under the condition that the model loss value of the training model is less than or equal to the preset loss value, taking the training model as the attribute detection model.
It should be noted that the preset loss value may be a loss value configured in the information processing apparatus; the preset loss value can also be a loss value transmitted to the information processing device by other equipment; the preset loss value can also be obtained by the information processing device in other modes; the specific way of obtaining the preset loss value by the information processing apparatus may be determined according to actual conditions, which is not limited in the embodiment of the present application.
For example, the number of sample objects may be 100, i.e., 100 items; the number of sample detail maps may be 2000, i.e. 2000 detailed maps; the 2000 quotient detail boxes can be labeled manually with a sample attribute box.
It should be noted that the number of sample objects and the number of sample detail diagrams may be determined according to actual situations, and the embodiment of the present application is not limited to this.
In an embodiment of the present application, the sample object may be a commodity on an e-commerce platform. The sample detail view may be a detailed view of a commercial product.
It should be noted that the sample object and the target object may be the same, or the sample object and the target object may be different; the specific details can be determined according to actual situations, and the embodiment of the present application does not limit the details.
In the embodiment of the present application, the information processing apparatus may acquire a sample detail drawing from a database; the information processing device can also acquire a sample detail map transmitted by other equipment from the information transmission interface; the information processing apparatus may acquire the sample detail map in other ways; the specific manner of acquiring the sample detail diagram by the information processing apparatus may be determined according to actual situations, and is not limited in the embodiment of the present application.
S103, classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values.
In the embodiment of the present application, after the information processing apparatus inputs the object detail map into the attribute detection model to obtain the initial attribute information of the target object, the information processing apparatus may classify the initial attribute information according to the object detail map by using the attribute classification model to obtain a plurality of attribute names and a plurality of attribute values.
In the embodiment of the present application, the attribute classification model may be a model configured in the information processing apparatus; the attribute classification model can also be a model obtained by training the information processing device; the specific manner in which the information processing apparatus obtains the attribute classification model may be determined according to actual conditions, which is not limited in the embodiment of the present application.
It should be noted that the attribute classification model may be a model constructed by a Semantic Entity Recognition (SER) algorithm; the attribute classification model may also be other models; the specific attribute classification model can be determined according to actual conditions, which is not limited in the embodiment of the present application.
In the embodiment of the application, before the information processing device classifies the initial attribute information according to the object detail map by using the attribute classification model to obtain a plurality of attribute names and a plurality of attribute values, the information processing device also removes the sample attribute information in the sample attribute frame from the sample detail map to obtain a removed sample detail map; acquiring target initial attribute information of a target sample object; deploying the target initial attribute information in the removed sample detail drawing to obtain an updated sample detail drawing; and training an initial attribute classification model by using the updated sample detail graph and the target initial attribute information to obtain an attribute classification model.
It should be noted that the target sample object is a different object from the sample object; the target initial attribute information includes a target initial attribute name and a target initial attribute value.
In the embodiment of the application, the target sample object is a commodity on an e-commerce platform.
In the embodiment of the present application, the information processing apparatus may acquire target initial attribute information of a target sample object from a database; the information processing device can also acquire target initial attribute information transmitted by other equipment from the information transmission interface; the information processing apparatus may also acquire the target initial attribute information in another manner; the specific manner in which the information processing apparatus acquires the target initial attribute information may be determined according to actual conditions, which is not limited in this embodiment of the present application.
In the embodiment of the present application, the information processing apparatus trains the initial attribute classification model by using the updated sample detail map and the target initial attribute information to obtain the attribute classification model, and may input the updated sample detail map and the target initial attribute information into the initial attribute classification model for the information processing apparatus to obtain the output attribute name and the output attribute value; determining model loss according to the output attribute name, the output attribute value and the target attribute name and the target attribute value in the target initial attribute information, and determining the initial attribute classification model as an attribute classification model under the condition that the model loss is less than or equal to a first loss value; and under the condition that the model loss is greater than the first loss value, continuously training the initial attribute classification model by using the updated sample detail graph and the target initial attribute information to obtain a training model, and under the condition that the model loss of the training model is less than or equal to the first loss value, determining the training model as the attribute classification model.
It should be noted that the first loss value may be a loss value configured in the information processing apparatus; the first loss value can also be a loss value transmitted to the information processing device by other equipment; the first loss value may also be obtained by the information processing apparatus in other manners; the specific manner of obtaining the first loss value by the information processing apparatus may be determined according to actual conditions, and this is not limited in the embodiment of the present application.
And S104, determining a plurality of corresponding relations between the plurality of attribute names and the plurality of attribute values by using the relation recognition model, and taking the plurality of corresponding relations as target attribute information of the target object.
In the embodiment of the present application, after the information processing apparatus classifies the initial attribute information according to the object detail map by using the attribute classification model to obtain the plurality of attribute names and the plurality of attribute values, the information processing apparatus may determine a plurality of correspondences between the plurality of attribute names and the plurality of attribute values by using the relationship identification model, and use the plurality of correspondences as the target attribute information of the target object.
In the embodiment of the present application, the relationship recognition model may be a model configured in the information processing apparatus; the relationship recognition model can also be a model obtained by training the information processing device; the specific manner in which the information processing apparatus obtains the relationship identification model may be determined according to actual conditions, which is not limited in the embodiment of the present application.
It should be noted that the relationship identification model may be a model constructed by a Relationship Extraction (RE) algorithm; the relationship recognition model may also be other models; the specific relationship identification model can be determined according to actual conditions, which is not limited in the embodiment of the present application.
After obtaining the target attribute information, the information processing apparatus outputs the target attribute information.
In the embodiment of the application, the information processing device determines a plurality of first attribute name positions of a plurality of attribute names in an object detail diagram and a plurality of first attribute value positions of a plurality of attribute values in the object detail diagram; and inputting the plurality of first attribute name positions, the plurality of first attribute value positions and the object detail graph into the relationship identification model, thereby obtaining a plurality of corresponding relationships between the plurality of attribute names and the plurality of attribute values.
In the embodiment of the application, before the information processing device determines the plurality of corresponding relations between the plurality of attribute names and the plurality of attribute values by using the relation recognition model, the information processing device further determines a first position of the target initial attribute name in the updated sample detail drawing and a second position of the target initial attribute value in the updated sample detail drawing; and training an initial relationship recognition model by using the first position, the second position and the updated sample detail graph to obtain a relationship recognition model.
In this embodiment of the present application, in the process of obtaining the relationship recognition model by using the first position, the second position, and the updated sample detail map to train the initial relationship recognition model, the information processing apparatus may train the initial relationship recognition model for the first position, the second position, and the updated sample detail map, and output the first corresponding relationship; determining model loss according to the first corresponding relation and the second corresponding relation, and determining the initial relation recognition model as a relation recognition model under the condition that the model loss is less than or equal to a second loss value; and under the condition that the model loss is greater than the second loss value, continuously training the initial relationship recognition model by using the first position, the second position and the updated sample detail graph to obtain a training model, and under the condition that the model loss of the training model is less than or equal to the second loss value, determining the training model as the relationship recognition model.
In the embodiment of the present application, the first correspondence is a correspondence between a first position and a second position output by the initial relationship identification model; the second correspondence is a correspondence between the target initial attribute name and the target initial attribute value.
It should be noted that the second loss value may be a loss value configured in the information processing apparatus; the second loss value can also be a loss value transmitted to the information processing device by other equipment; the second loss value may also be obtained by the information processing apparatus in other manners; the specific manner of obtaining the second loss value by the information processing apparatus may be determined according to actual conditions, and this is not limited in the embodiment of the present application.
In the embodiment of the application, the K-V key value pair of the attribute value and the attribute name is finally required for commodity attribute standardization. The method is based on the layout XLM method to extract the attribute key information, and mainly depends on Semantic Entity Recognition (SER) and Relation Extraction (RE) tasks. The method comprises the following specific steps:
removing the attribute content of a labeling picture used for a target detection task, and leaving a background picture of the labeling picture to obtain a background picture similar to real data (in the sample detail picture, sample attribute information in a sample attribute frame is removed, and the removed sample detail picture is obtained); reasonably distributing and typesetting the existing commodity attributes (target initial attribute information) and the background graph by utilizing the existing commodity attributes and the background graph, and randomly generating an attribute parameter graph with a label (the target initial attribute information is deployed in the rejected sample detail graph to obtain an updated sample detail graph) as model training data; and (4) SER training: mainly based on image information and text information, using a transform to model, and classifying attribute parameters by query (attribute name) and answer (attribute value) by using a space attention mechanism; RE training: and establishing a relation for the SER identification result, and determining the K-V relation.
It should be noted that, by using the trained models (attribute classification model, relationship identification model), text identification and classification (SER) in the image and Relationship Extraction (RE) of the text content are completed, and finally KV pairs (i.e. key value pairs) of attribute values and attribute names are generated as supplementary parameters and title extraction information check items for commodity standardization.
Illustratively, as shown in FIG. 2: under the condition that a target object (commodity input) is received, acquiring an object detail drawing corresponding to the target object (acquiring all merchant detail drawings under the commodity); determining an initial information frame in the object detail graph by using an attribute detection model (attribute parameter frame detection); and extending each edge of the initial information frame according to the direction far away from the central point of the initial information frame to obtain the attribute information frame. Identifying information (OCR text extraction) in a defined range of an attribute information frame in the object detail graph to obtain initial information; segmenting the initial information by using preset characters to obtain a plurality of segmentation information; the plurality of pieces of division information are recombined (result recombination) in accordance with the positions of the plurality of pieces of division information in the object detail drawing and the intervals between the plurality of pieces of division information, and initial attribute information is obtained. Classifying the initial attribute information according to the object detail graph by using an attribute classification model (DocVQA) to obtain a plurality of attribute names and a plurality of attribute values; a plurality of correspondences (attribute pair results) between the plurality of attribute names and the plurality of attribute values are determined by using the relationship recognition model, and the plurality of correspondences are output as target attribute information of the target object (attribute pair results are output). In the case where the initial information box is not determined in the object detail drawing using the attribute detection model, "null" is output directly.
It should be noted that the Document visual problem solution (ADATASE for VQAON Document Images, docVQA).
In the embodiment of the present application, the flow of the whole system of information processing is shown in fig. 3: inputting a SKU (target object) of the commodity, and acquiring an object detail diagram of the target object (acquiring a full-scale quotient detail diagram); determining an attribute information frame (an object detection and identification parameter frame) in the object detail graph by using an attribute detection model; acquiring information in a defined range of an attribute information frame in the object detail graph to obtain initial attribute information (OCR result processing); classifying (SER) the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values; a plurality of correspondences (REs) between the plurality of attribute names and the plurality of attribute values are determined using the relationship recognition model, and the plurality of correspondences are output as target attribute information of the target object (json result is output).
Note, the Stock Keeping Unit (SKU).
It can be understood that the information processing apparatus acquires initial attribute information of the target object from the object detail map by using the attribute detection model, establishes a plurality of correspondences between a plurality of attribute names and a plurality of attribute values for the plurality of attribute names and the plurality of attribute values determined from the initial attribute information by using the attribute detection model, the attribute classification model and the relationship recognition model, does not need to manually acquire the target attribute information of the target object from the object detail map, and can accurately acquire the target attribute information from the object detail map by using the information processing method even when the number of the object detail maps is large, that is, the accuracy in acquiring the target attribute information is improved.
Based on the same inventive concept as the above-described information processing method, the present embodiment provides an information processing apparatus 1 corresponding to an information processing method; fig. 4 is a schematic diagram illustrating a first composition structure of an information processing apparatus according to an embodiment of the present application, where the information processing apparatus 1 may include:
the acquiring unit 11 is configured to acquire an object detail map corresponding to a target object when the target object is received;
the input unit 12 is configured to input the object detail map into an attribute detection model, so as to obtain initial attribute information of the target object;
a classification unit 13, configured to classify the initial attribute information according to the object detail map by using an attribute classification model to obtain multiple attribute names and multiple attribute values;
a determining unit 14, configured to determine a plurality of correspondences between the plurality of attribute names and the plurality of attribute values by using a relationship recognition model, and use the plurality of correspondences as target attribute information of the target object.
In some embodiments of the present application, the determining unit 14 is configured to determine an attribute information box in the object detail map by using the attribute detection model;
the obtaining unit 11 is configured to obtain information in the defined range of the attribute information frame in the object detail map, so as to obtain the initial attribute information.
In some embodiments of the present application, the apparatus further comprises an extension unit;
the determining unit 14 is configured to determine an initial information frame in the object detail diagram by using the attribute detection model;
and the extending unit is used for extending each edge of the initial information frame according to the direction far away from the central point of the initial information frame to obtain the attribute information frame.
In some embodiments of the present application, the apparatus further comprises an identification unit, a segmentation unit, and a reassembly unit;
the identification unit is used for identifying information in the defined range of the attribute information frame in the object detail graph to obtain initial information;
the segmentation unit is used for segmenting the initial information by using preset characters to obtain a plurality of segmentation information;
and the restructuring unit is configured to restructure the multiple pieces of segmentation information according to the positions of the multiple pieces of segmentation information in the object detail map and the distances between the multiple pieces of segmentation information, so as to obtain the initial attribute information.
In some embodiments of the present application, the apparatus further comprises a training unit;
the acquiring unit 11 is configured to acquire a sample detail drawing corresponding to a sample object; marking a sample attribute box in the sample detail drawing;
and the training unit is used for training an initial attribute detection model by using the sample detail diagram and the sample attribute box to obtain the attribute detection model.
In some embodiments of the present application, the apparatus further comprises a rejection unit and a deployment unit;
the removing unit is used for removing the sample attribute information in the sample attribute frame from the sample detail picture to obtain a removed sample detail picture;
the acquiring unit 11 is configured to acquire target initial attribute information of a target sample object; the target sample object is a different object than the sample object; the target initial attribute information comprises a target initial attribute name and a target initial attribute value;
the deployment unit is used for deploying the target initial attribute information in the removed sample detail drawing to obtain an updated sample detail drawing;
and the training unit is used for training an initial attribute classification model by using the updated sample detail diagram and the target initial attribute information to obtain the attribute classification model.
In some embodiments of the present application, the determining unit 14 is configured to determine a first position of the target initial attribute name in the updated sample detail drawing, and a second position of the target initial attribute value in the updated sample detail drawing;
the training unit is configured to train an initial relationship recognition model by using the first position, the second position, and the updated sample detail map, so as to obtain the relationship recognition model.
In practical applications, the obtaining Unit 11, the input Unit 12, the classifying Unit 13, and the determining Unit 14 may be implemented by a processor 15 on the information Processing apparatus 1, specifically implemented by a CPU (Central Processing Unit), an MPU (micro processor Unit), a DSP (Digital Signal Processing), a Field Programmable Gate Array (FPGA), or the like; the above data storage may be realized by the memory 16 on the information processing apparatus 1.
An embodiment of the present application also provides an information processing apparatus 1, and as shown in fig. 5, the information processing apparatus 1 includes: a processor 15, a memory 16 and a communication bus 17, the memory 16 communicating with the processor 15 through the communication bus 17, the memory 16 storing a program executable by the processor 15, the program, when executed, executing the information processing method as described above through the processor 15.
In practical applications, the Memory 16 may be a volatile Memory (volatile Memory), such as a Random-Access Memory (RAM); or a non-volatile Memory (non-volatile Memory), such as a Read-Only Memory (ROM), a flash Memory (flash Memory), a Hard Disk Drive (HDD) or a Solid-State Drive (SSD); or a combination of the above types of memories and provides instructions and data to the processor 15.
The present embodiment provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by the processor 15 to implement the information processing method as described above.
It can be understood that the information processing apparatus acquires initial attribute information of the target object from the object detail map by using the attribute detection model, establishes a plurality of correspondences between a plurality of attribute names and a plurality of attribute values for the plurality of attribute names and the plurality of attribute values determined from the initial attribute information by using the attribute detection model, the attribute classification model and the relationship recognition model, does not need to manually acquire the target attribute information of the target object from the object detail map, and can accurately acquire the target attribute information from the object detail map by using the information processing method even when the number of the object detail maps is large, that is, the accuracy in acquiring the target attribute information is improved.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present application, and is not intended to limit the scope of the present application.
Claims (10)
1. An information processing method, characterized in that the method comprises:
under the condition that a target object is received, acquiring an object detail graph corresponding to the target object;
inputting the object detail graph into an attribute detection model to obtain initial attribute information of the target object;
classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values;
and determining a plurality of corresponding relations between the plurality of attribute names and the plurality of attribute values by using a relation recognition model, and using the plurality of corresponding relations as target attribute information of the target object.
2. The method of claim 1, wherein the inputting the object detail map into a property detection model to obtain initial property information of the target object comprises:
determining an attribute information frame in the object detail graph by using the attribute detection model;
and acquiring information in the defined range of the attribute information frame in the object detail graph to obtain the initial attribute information.
3. The method according to claim 2, wherein said determining an attribute information box in said object detail view using said attribute detection model comprises:
determining an initial information frame in the object detail graph by using the attribute detection model;
and extending each edge of the initial information frame according to the direction far away from the central point of the initial information frame to obtain the attribute information frame.
4. The method according to claim 2, wherein the obtaining information in the range framed by the attribute information frame in the object detail map to obtain the initial attribute information comprises:
identifying information in the range defined by the attribute information frame in the object detail graph to obtain initial information;
segmenting the initial information by using preset characters to obtain a plurality of segmentation information;
according to the positions of the segmentation information in the object detail map and the distances among the segmentation information, and recombining the plurality of pieces of segmentation information to obtain the initial attribute information.
5. The method of claim 1, wherein before inputting the object detail map into a property detection model to obtain initial property information of the target object, the method further comprises:
acquiring a sample detail drawing corresponding to a sample object; marking a sample attribute box in the sample detail drawing;
and training an initial attribute detection model by using the sample detail drawing and the sample attribute box to obtain the attribute detection model.
6. The method according to claim 1, wherein before the classifying the initial attribute information according to the object detail graph by using the attribute classification model to obtain a plurality of attribute names and a plurality of attribute values, the method further comprises:
in the sample detail drawing, sample attribute information in the sample attribute frame is removed to obtain a removed sample detail drawing;
acquiring target initial attribute information of a target sample object; the target sample object is a different object than the sample object; the target initial attribute information comprises a target initial attribute name and a target initial attribute value;
deploying the target initial attribute information in the removed sample detail drawing to obtain an updated sample detail drawing;
and training an initial attribute classification model by using the updated sample detail graph and the target initial attribute information to obtain the attribute classification model.
7. The method of claim 1, wherein prior to determining the plurality of correspondences between the plurality of attribute names and the plurality of attribute values using the relationship recognition model, the method further comprises:
determining a first position of a target initial attribute name in the updated sample detail drawing and a second position of a target initial attribute value in the updated sample detail drawing;
and training an initial relationship recognition model by using the first position, the second position and the updated sample detail graph to obtain the relationship recognition model.
8. An information processing apparatus characterized in that the apparatus comprises:
the device comprises an acquisition unit, a processing unit and a display unit, wherein the acquisition unit is used for acquiring an object detail graph corresponding to a target object under the condition of receiving the target object;
the input unit is used for inputting the object detail graph into an attribute detection model to obtain initial attribute information of the target object;
the classification unit is used for classifying the initial attribute information according to the object detail graph by using an attribute classification model to obtain a plurality of attribute names and a plurality of attribute values;
a determining unit configured to determine a plurality of correspondence relationships between the plurality of attribute names and the plurality of attribute values using a relationship recognition model, and to use the plurality of correspondence relationships as target attribute information of the target object.
9. An information processing apparatus characterized in that the apparatus comprises:
a memory, a processor, and a communication bus, the memory in communication with the processor through the communication bus, the memory storing an information processing program executable by the processor, the information processing program when executed causing the processor to perform the method of any of claims 1 to 7.
10. A storage medium having stored thereon a computer program for application to an information processing apparatus, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211565452.5A CN115858695A (en) | 2022-12-07 | 2022-12-07 | Information processing method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211565452.5A CN115858695A (en) | 2022-12-07 | 2022-12-07 | Information processing method and device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115858695A true CN115858695A (en) | 2023-03-28 |
Family
ID=85670777
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211565452.5A Pending CN115858695A (en) | 2022-12-07 | 2022-12-07 | Information processing method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115858695A (en) |
-
2022
- 2022-12-07 CN CN202211565452.5A patent/CN115858695A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11416710B2 (en) | Feature representation device, feature representation method, and program | |
| CN108073910B (en) | Method and device for generating human face features | |
| KR102002024B1 (en) | Method for processing labeling of object and object management server | |
| CN111582409A (en) | Training method of image label classification network, image label classification method and device | |
| CN109426831B (en) | Image similarity matching and model training method and device and computer equipment | |
| CN111931859B (en) | Multi-label image recognition method and device | |
| CN113837151B (en) | Table image processing method and device, computer equipment and readable storage medium | |
| CN113705733A (en) | Medical bill image processing method and device, electronic device and storage medium | |
| CN111639970A (en) | Method for determining price of article based on image recognition and related equipment | |
| Do et al. | Automatic license plate recognition using mobile device | |
| CN113378710A (en) | Layout analysis method and device for image file, computer equipment and storage medium | |
| CN113221918A (en) | Target detection method, and training method and device of target detection model | |
| CN112396060A (en) | Identity card identification method based on identity card segmentation model and related equipment thereof | |
| CN114359912B (en) | Software page key information extraction method and system based on graph neural network | |
| JP2020502710A (en) | Web page main image recognition method and apparatus | |
| JP2011238043A (en) | Summarized comic image generation device, program and method for generating summary of comic content | |
| CN110413869B (en) | Method and device for pushing information | |
| JP2018013887A (en) | Feature selection device, tag relevant area extraction device, method, and program | |
| CN115858695A (en) | Information processing method and device and storage medium | |
| CN113221792B (en) | Chapter detection model construction method, cataloguing method and related equipment | |
| CN115063784A (en) | Bill image information extraction method and device, storage medium and electronic equipment | |
| CN115880702A (en) | Data processing method, device, equipment, program product and storage medium | |
| CN115034177A (en) | Presentation file conversion method, device, equipment and storage medium | |
| Rahul et al. | Deep reader: Information extraction from document images via relation extraction and natural language | |
| CN112328812A (en) | Domain knowledge extraction method and system based on self-adjusting parameters and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |