CN113656668A

CN113656668A - Retrieval method, management method, device, equipment and medium of multi-modal information base

Info

Publication number: CN113656668A
Application number: CN202110955328.9A
Authority: CN
Inventors: 魏翔; 孙逸鹏; 姚锟; 韩钧宇; 丁二锐; 刘经拓
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2021-08-19
Filing date: 2021-08-19
Publication date: 2021-11-16
Anticipated expiration: 2041-08-19
Also published as: CN113656668B; WO2023019948A1

Abstract

The invention provides a retrieval method and a management method for a multi-modal information base, relates to the technical field of artificial intelligence, in particular to the technical field of computer vision and deep learning, and can be applied to scenes such as image recognition, image search and the like. The implementation scheme is as follows: in response to receiving retrieval information including first modal information, extracting first modal features of the retrieval information from the first modal information of the retrieval information using a first multi-modal feature extraction module; selecting a first set of the plurality of pieces of target information based on a similarity of the first modal feature of the search information to each of the first modal feature and the second modal feature of each of the plurality of pieces of target information; and generating a retrieval result based on the first group of target information.

Description

Retrieval method, management method, device, equipment and medium of multi-modal information base

Technical Field

The present disclosure relates to the field of artificial intelligence technologies, and in particular, to the field of computer vision and deep learning technologies, which can be applied to scenes such as image recognition and image search, and in particular, to a retrieval method, a management method, an apparatus, an electronic device, a computer-readable storage medium, and a computer program product for a multimodal information base.

Background

Artificial intelligence is the subject of research that makes computers simulate some human mental processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), both at the hardware level and at the software level. The artificial intelligence hardware technology generally comprises technologies such as a sensor, a special artificial intelligence chip, cloud computing, distributed storage, big data processing and the like, and the artificial intelligence software technology mainly comprises a computer vision technology, a voice recognition technology, a natural language processing technology, machine learning/deep learning, a big data processing technology, a knowledge graph technology and the like.

The traditional retrieval information base is difficult to solve the problem that the target information in the retrieval information base is not consistent in picture and text.

The approaches described in this section are not necessarily approaches that have been previously conceived or pursued. Unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section. Similarly, unless otherwise indicated, the problems mentioned in this section should not be considered as having been acknowledged in any prior art.

Disclosure of Invention

The present disclosure provides a retrieval method, a management method, an apparatus, an electronic device, a computer-readable storage medium, and a computer program product for a multimodal information repository.

According to an aspect of the present disclosure, there is provided a retrieval method for a multimodal information library, wherein the multimodal information library includes a plurality of pieces of target information, each piece of target information including first modality information and second modality information, the method including: in response to receiving retrieval information including first modal information, extracting first modal features of the retrieval information from the first modal information of the retrieval information using a first multi-modal feature extraction module; selecting a first group of target information in the plurality of pieces of target information based on similarity of first modal features of the retrieval information and each of the first modal features and second modal features of each piece of target information, wherein the first modal features of each piece of target information are extracted from the first modal information of the target information by using a first multi-modal feature extraction module, and the second modal features of each piece of target information are extracted from the second modal information of the target information by using a second multi-modal feature extraction module; and generating a retrieval result based on the first group of target information.

According to another aspect of the present disclosure, there is provided a method for managing a multimodal information repository, the method including: in response to receiving warehousing information comprising first modality information and second modality information, extracting first modality features of the warehousing information from the first modality information of the warehousing information by using a first multi-modality feature extraction module, and extracting second modality features of the warehousing information from the second modality information of the warehousing information by using a second multi-modality feature extraction module; calculating multi-modal characteristics of the warehousing information based on the first modal characteristics and the second modal characteristics of the warehousing information; generating one or more retrieval objects corresponding to the warehousing information in a multi-modal information library based on the first modal characteristic, the second modal characteristic and the multi-modal characteristic of the warehousing information; and in response to receiving the retrieval information, performing a retrieval method according to the present disclosure.

According to another aspect of the present disclosure, there is provided a retrieval apparatus for a multimodal information library, wherein the multimodal information library includes a plurality of target information including first modality information and second modality information, the apparatus including: a retrieval feature extraction module configured to: in response to receiving retrieval information including first modal information, extracting first modal features of the retrieval information from the first modal information of the retrieval information using a first multi-modal feature extraction module; a target matching module configured to: selecting a first group of target information in the plurality of pieces of target information based on similarity of first modal features of the retrieval information and each of the first modal features and second modal features of each piece of target information, wherein the first modal features of each piece of target information are extracted from the first modal information of the target information by using a first multi-modal feature extraction module, and the second modal features of each piece of target information are extracted from the second modal information of the target information by using a second multi-modal feature extraction module; and a retrieval result generation module configured to: and generating a retrieval result based on the first group of target information.

According to another aspect of the present disclosure, there is provided a management apparatus for a multimodal information repository, including: a warehousing information extraction module configured to: in response to receiving warehousing information comprising first modality information and second modality information, extracting first modality features of the warehousing information from the first modality information of the warehousing information by using a first multi-modality feature extraction module, and extracting second modality features of the warehousing information from the second modality information of the warehousing information by using a second multi-modality feature extraction module; a multimodal information generation module configured to: calculating multi-modal characteristics of the warehousing information based on the first modal characteristics and the second modal characteristics of the warehousing information; a search object generation module configured to: generating one or more retrieval objects corresponding to the warehousing information in a multi-modal information library based on the first modal characteristic, the second modal characteristic and the multi-modal characteristic of the warehousing information; and a retrieval apparatus for multimodal information repositories as described in the present disclosure.

According to another aspect of the present disclosure, there is provided an electronic device including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a retrieval method and/or a management method as described in the present disclosure.

According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a retrieval method and/or a management method according to the present disclosure.

According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein the computer program, when executed by a processor, implements a retrieval method and/or a management method as described in the present disclosure.

According to one or more embodiments of the present disclosure, multiple modality information in a multi-modality information base can be retrieved, and the problem of inconsistency between different modality information of the same target information in the multi-modality information base is avoided.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the embodiments and, together with the description, serve to explain the exemplary implementations of the embodiments. The illustrated embodiments are for purposes of illustration only and do not limit the scope of the claims. Throughout the drawings, identical reference numbers designate similar, but not necessarily identical, elements.

FIG. 1 illustrates a schematic diagram of an exemplary system in which various methods described herein may be implemented, according to an embodiment of the present disclosure;

FIG. 2 shows a flow diagram of a retrieval method for a multimodal information repository, in accordance with an embodiment of the present disclosure;

FIG. 3 shows a flow diagram of a retrieval method for a multimodal information repository, in accordance with an embodiment of the present disclosure;

FIG. 4 shows a flow diagram of a retrieval method for a multimodal information repository, in accordance with an embodiment of the present disclosure;

FIG. 5 shows a flow diagram of a management method for a multimodal information repository, in accordance with an embodiment of the present disclosure;

FIG. 6 shows a flow diagram of a management method for a multimodal information repository, in accordance with an embodiment of the present disclosure;

fig. 7 illustrates a flow diagram of an example process of extracting single-modality image features of binned information from one or more pieces of subject information of the binned information in the method of fig. 6, in accordance with an embodiment of the present disclosure;

FIG. 8 shows a block diagram of a retrieval arrangement for a multimodal information repository, in accordance with an embodiment of the present disclosure;

FIG. 9 shows a block diagram of a structure of a management apparatus for a multimodal information repository, according to an embodiment of the present disclosure;

FIG. 10 illustrates a block diagram of an exemplary electronic device that can be used to implement embodiments of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, the timing relationship, or the importance relationship of the elements, and such terms are used only to distinguish one element from another. In some examples, a first element and a second element may refer to the same instance of the element, and in some cases, based on the context, they may also refer to different instances.

The terminology used in the description of the various described examples in this disclosure is for the purpose of describing particular examples only and is not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the elements may be one or more. Furthermore, the term "and/or" as used in this disclosure is intended to encompass any and all possible combinations of the listed items.

Embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

Fig. 1 illustrates a schematic diagram of an exemplary system 100 in which various methods and apparatus described herein may be implemented in accordance with embodiments of the present disclosure. Referring to fig. 1, the system 100 includes one or

more client devices

101, 102, 103, 104, 105, and 106, a server 120, and one or more communication networks 110 coupling the one or more client devices to the server 120.

Client devices

101, 102, 103, 104, 105, and 106 may be configured to execute one or more applications.

In embodiments of the present disclosure, the server 120 may run one or more services or software applications that enable the execution of retrieval methods and/or management methods for multimodal information repositories as described in the present disclosure.

In some embodiments, the server 120 may also provide other services or software applications that may include non-virtual environments and virtual environments. In certain embodiments, these services may be provided as web-based services or cloud services, for example, provided to users of

client devices

101, 102, 103, 104, 105, and/or 106 under a software as a service (SaaS) model.

In the configuration shown in fig. 1, server 120 may include one or more components that implement the functions performed by server 120. These components may include software components, hardware components, or a combination thereof, which may be executed by one or more processors. A user operating a

client device

101, 102, 103, 104, 105, and/or 106 may, in turn, utilize one or more client applications to interact with the server 120 to take advantage of the services provided by these components. It should be understood that a variety of different system configurations are possible, which may differ from system 100. Accordingly, fig. 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

The user can use the

client device

101, 102, 103, 104, 105, and/or 106 to retrieve targeted information in the multimodal information store (e.g., upload retrieved information), or to add targeted information to the multimodal information store (e.g., upload entered information). The client device may provide an interface that enables a user of the client device to interact with the client device. The client device may also output information to the user via the interface. Although fig. 1 depicts only six client devices, those skilled in the art will appreciate that any number of client devices may be supported by the present disclosure.

Client devices

101, 102, 103, 104, 105, and/or 106 may include various types of computer devices, such as portable handheld devices, general purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, gaming systems, thin clients, various messaging devices, sensors or other sensing devices, and so forth. These computer devices may run various types and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux, or Linux-like operating systems (e.g., GOOGLE Chrome OS); or include various Mobile operating systems such as MICROSOFT Windows Mobile OS, iOS, Windows Phone, Android. Portable handheld devices may include cellular telephones, smart phones, tablets, Personal Digital Assistants (PDAs), and the like. Wearable devices may include head mounted displays and other devices. The gaming system may include a variety of handheld gaming devices, internet-enabled gaming devices, and the like. The client device is capable of executing a variety of different applications, such as various Internet-related applications, communication applications (e.g., email applications), Short Message Service (SMS) applications, and may use a variety of communication protocols.

Network 110 may be any type of network known to those skilled in the art that may support data communications using any of a variety of available protocols, including but not limited to TCP/IP, SNA, IPX, etc. By way of example only, one or more networks 110 may be a Local Area Network (LAN), an ethernet-based network, a token ring, a Wide Area Network (WAN), the internet, a virtual network, a Virtual Private Network (VPN), an intranet, an extranet, a Public Switched Telephone Network (PSTN), an infrared network, a wireless network (e.g., bluetooth, WIFI), and/or any combination of these and/or other networks.

The server 120 may include one or more general purpose computers, special purpose server computers (e.g., PC (personal computer) servers, UNIX servers, mid-end servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and/or combination. The server 120 may include one or more virtual machines running a virtual operating system, or other computing architecture involving virtualization (e.g., one or more flexible pools of logical storage that may be virtualized to maintain virtual storage for the server). In various embodiments, the server 120 may run one or more services or software applications that provide the functionality described below.

The computing units in server 120 may run one or more operating systems including any of the operating systems described above, as well as any commercially available server operating systems. The server 120 may also run any of a variety of additional server applications and/or middle tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, and the like.

In some implementations, the server 120 may include one or more applications to analyze and consolidate data feeds and/or event updates received from users of the

client devices

101, 102, 103, 104, 105, and 106. The server 120 may also include one or more applications to display data feeds and/or real-time events via one or more display devices of the

client devices

101, 102, 103, 104, 105, and 106 (in some embodiments, the server 120 may include one or more applications, e.g., applications of services based on object detection and recognition, signal conversion, etc. of data such as image, video, voice, text, digital signals, etc. to process task requests such as voice interaction, text classification, image recognition, or keypoint detection, etc. received from the

client devices

101, 102, 103, 104, 105, and 106, the server may train the neural network model with training samples according to specific deep learning tasks, and may test individual ones of the super-network modules of the neural network model, based on the test results of the individual sub-networks, structures and parameters of a neural network model for performing a deep learning task are determined. Various data can be used as training sample data of the deep learning task, such as image data, audio data, video data or text data. After the training of the neural network model is completed, the server 120 may also automatically search out an optimal model structure through a model search technique to perform a corresponding task).

In some embodiments, the server 120 may be a server of a distributed system, or a server incorporating a blockchain. The server 120 may also be a cloud server, or a smart cloud computing server or a smart cloud host with artificial intelligence technology. The cloud Server is a host product in a cloud computing service system, and is used for solving the defects of high management difficulty and weak service expansibility in the traditional physical host and Virtual Private Server (VPS) service.

The system 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as audio files and video files. The data store 130 may reside in various locations. For example, the data store used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection. The data store 130 may be of different types. In certain embodiments, the data store used by the server 120 may be a database, such as a relational database. One or more of these databases may store, update, and retrieve data to and from the database in response to the command.

In some embodiments, one or more of the databases 130 may also be used by applications to store application data. The databases used by the application may be different types of databases, such as key-value stores, object stores, or regular stores supported by a file system.

The system 100 of fig. 1 may be configured and operated in various ways to enable application of the various methods and apparatus described in accordance with the present disclosure.

As described above, the conventional search information library searches based on text keywords or picture contents of target information, and therefore, it is desirable to provide a method for searching based on multi-modality information (e.g., image information and text information) of target information to avoid a situation where there is no match between multi-modality information of the same target information.

An embodiment of the present disclosure provides a retrieval method for a multimodal information base, wherein the multimodal information base includes a plurality of pieces of target information, each piece of target information includes first modality information and second modality information, the method including: in response to receiving retrieval information including first modal information, extracting first modal features of the retrieval information from the first modal information of the retrieval information using a first multi-modal feature extraction module; selecting a first group of target information in the plurality of pieces of target information based on similarity of first modal features of the retrieval information and each of the first modal features and second modal features of each piece of target information, wherein the first modal features of each piece of target information are extracted from the first modal information of the target information by using a first multi-modal feature extraction module, and the second modal features of each piece of target information are extracted from the second modal information of the target information by using a second multi-modal feature extraction module; and generating a retrieval result based on the first group of target information.

FIG. 2 shows a flow diagram of a retrieval method 200 for a multimodal information repository, in accordance with an embodiment of the present disclosure. According to some embodiments, the multimodal information library includes a plurality of pieces of target information, each piece of target information including first modality information and second modality information.

At step S201, in response to receiving search information including first modality information, using a first multi-modality feature extraction module, first modality features of the search information are extracted from the first modality information of the search information.

According to some embodiments, the retrieval information is a retrieval request sent by the client to the server, for example, when a user wishes to retrieve a skirt seen online on the internet, the user can input "skirt" or take a picture of the seen skirt at the client, and send a corresponding retrieval request to the server through the client.

According to some embodiments, the first modality information of the retrieval information may be text information or image information, wherein when the first modality information of the retrieval information is text information, a multi-modality text feature of the retrieval information is extracted using a multi-modality text extraction module (e.g., Bert Base network), and when the first modality information of the retrieval information is image information, a multi-modality image feature of the retrieval information is extracted using a multi-modality image extraction module (e.g., ViT Base network).

At step S203, a first group of target information of the plurality of pieces of target information is selected based on a similarity of a first modal feature of the retrieval information to each of a first modal feature and a second modal feature of each piece of target information, wherein the first modal feature of each piece of target information is extracted from the first modal information of the target information using a first multimodal feature extraction module, and the second modal feature of each piece of target information is extracted from the second modal information of the target information using a second multimodal feature extraction module.

According to some embodiments, selecting a first set of target information of the plurality of pieces of target information comprises: selecting a second set of target information of the plurality of pieces of target information based on a similarity of the first modal feature of the search information and the first modal feature of each piece of target information of the plurality of pieces of target information; and selecting the first set of target information from the second set of target information based on a similarity of the first modal characteristics of the retrieved information to the second modal characteristics of each of the target information in the second set of targets.

According to further embodiments, selecting a first set of target information of the plurality of pieces of target information includes: a retrieval score of each piece of target information is calculated based on a similarity of the first modal feature of the retrieval information to the first modal feature of each piece of target information and a similarity of the second modal feature of the retrieval information to the second modal feature of each piece of target information, and a first set of target information is selected from the plurality of pieces of target information based on the retrieval score.

According to some embodiments, the plurality of pieces of target information are sorted based on a similarity of the first modality feature of the retrieval information to the first modality feature of each of the plurality of pieces of target information, and a first predetermined number of top pieces of target information having a highest similarity among the plurality of pieces of target information are selected as the second group of target information. According to further embodiments, target information, of which the similarity of the first-modality feature of the retrieval information to the first-modality feature thereof is greater than a first similarity threshold, is selected from the plurality of items of target information as the second set of target information.

According to some embodiments, the second group of target information is ranked based on similarity of the second modal characteristics of the retrieved information to the second modal characteristics of each of the plurality of pieces of target information, and the first predetermined number of pieces of target information with the highest similarity among the second group of target information is selected as the first group of target information. According to further embodiments, target information from the second set of target information is selected from the retrieved information having a second modality feature with a similarity to the second modality feature greater than a second similarity threshold as the first set of target information. According to still further embodiments, all target information in the second set of target information (i.e. the first predetermined number is the same as the second predetermined number) may be selected as the first set of target information, and the target information may be ranked only according to the similarity of the second modality feature of the retrieved information to its second modality feature.

According to some embodiments, the similarity of the first-modality feature of the retrieval information to the first-modality feature of each piece of target information and/or the similarity of the second-modality feature of the retrieval information to the second-modality feature of each piece of target information is a cosine similarity.

According to some embodiments, the first modality information is any one of image information and text information, and the second modality information is the other one of image information and text information.

At step S205, a search result is generated based on the first set of target information.

According to some embodiments, the target information arrangement order in the search result is determined based on the similarity between the first modal characteristics of the search information and the first modal characteristics of each piece of target information and/or the similarity between the second modal characteristics of the search information and the second modal characteristics of each piece of target information. According to some embodiments, a retrieval score for each piece of target information is calculated based on a similarity of the first modal features of the retrieved information to the first modal features of each piece of target information and/or a similarity of the second modal features of the retrieved information to the second modal features of each piece of target information, and a retrieval result corresponding to the first set of target information is generated based on the retrieval score.

In the retrieval method for the multi-modal information library according to the embodiment of the present disclosure, even if the user inputs only single-modal retrieval information, retrieval can be performed based on a plurality of modal information of the target information, and the problem of non-coincidence between different modal information of the same target information in the multi-modal information library (for example, image information and text information of a certain target information do not coincide) is avoided.

According to some embodiments, the first modality information is image information, and the retrieval method for the multimodal information library as described in the present disclosure further comprises, before selecting the first set of target information of the plurality of pieces of target information: extracting one or more pieces of subject information from first modality information of the retrieval information using a subject detection module; for each piece of subject information, extracting a single-mode image feature of the subject information from the subject information by using an image feature extraction module; and selecting a third set of the plurality of pieces of target information based on a similarity of a single-modality image feature of the one or more pieces of subject information to a single-modality image feature of each of the plurality of pieces of target information, wherein the single-modality image feature of each of the plurality of pieces of target information is extracted from the first-modality information of the target information using an image feature extraction module, and wherein selecting the first set of the plurality of pieces of target information includes: for each piece of target information in the third group of target information, calculating a similarity score of the target information based on the similarity of the first modal feature of the retrieval information and the first modal feature of the target information and the similarity of the first modal feature of the retrieval information and the second modal feature of the target information; and selecting the first set of target information from the third set of target information based on the similarity score of each piece of target information in the third set of target information.

FIG. 3 shows a flow diagram of a retrieval method 300 for a multimodal information repository, in accordance with an embodiment of the present disclosure.

At step S301, in response to receiving search information including first modality information, using a first multi-modality feature extraction module, a first modality feature of the search information is extracted from the first modality information of the search information, wherein the first modality information is image information. According to some embodiments, step S301 may be performed similarly to step S201 in fig. 2.

At step S303, one or more pieces of subject information are extracted from the first modality information of the retrieval information using a subject detection module.

According to some embodiments, at the subject detection module, a target detector (e.g., YOLO-v3) is used to perform target detection on the first modality information (i.e., image information) of the search information, wherein for the detected detection boxes, the detection boxes with higher confidence and appropriate size and position are screened out (e.g., the detection boxes with lower confidence, smaller size and closer to the picture boundary are filtered out), and information corresponding to the screened-out detection boxes in the first modality information is extracted as one or more pieces of subject information. According to some embodiments, the detection frames with smaller size and closer to the picture boundary may be filtered out from the detection frames, and then two detection frames with the highest confidence level among the remaining detection frames are selected as the screened detection frames.

At step S305, for each piece of subject information, a single-modality image feature of the subject information is extracted from the subject information using an image feature extraction module.

According to some embodiments, the image feature extraction module is structurally identical to the first multimodal feature extraction module. According to some embodiments, before training the image feature extraction module, training of the first multi-modal feature extraction module is performed, and the trained parameters of the first multi-modal training extraction module are used as initialization parameters of the image feature extraction module to perform training of the image feature extraction module (e.g., fine tuning using metric learning based on ID labeled image data). Compared with the direct training image feature extraction module, the training time of the image feature extraction module is shortened by taking the trained parameters of the first multi-modal training extraction module as the initialization parameters of the image feature extraction module.

At step S307, a third set of the plurality of pieces of target information is selected based on the similarity of the single-modality image features of the one or more pieces of subject information to the single-modality image features of each of the plurality of pieces of target information.

According to some embodiments, the one or more pieces of subject information include a plurality of pieces of subject information, and wherein selecting a third set of target information of the plurality of pieces of target information includes: for each piece of subject information, selecting a plurality of pieces of target information corresponding to the subject information from the plurality of pieces of target information based on the similarity between the single-mode image feature of the subject information and the single-mode image feature of each piece of target information in the target information; and selecting a plurality of pieces of object information corresponding to each piece of body information as a third set of object information. According to some embodiments, a plurality of pieces of target information corresponding to each piece of information are aggregated and deduplicated to obtain a third set of target information.

According to still further embodiments, the one or more pieces of subject information include one piece of subject information, and wherein selecting the third set of target information of the plurality of pieces of target information includes: and selecting a plurality of pieces of target information corresponding to the subject information from the plurality of pieces of target information as a third group of target information based on the similarity between the single-mode image features of the subject information and the single-mode image features of each piece of target information.

According to some embodiments, for each piece of subject information, the plurality of pieces of target information are sorted based on a similarity of a single-modality image feature of the subject information to a single-modality image feature of each of the plurality of pieces of target information, and a top third predetermined number of pieces of target information having a highest similarity among the plurality of pieces of target information are selected. According to further embodiments, for each piece of subject information, target information is selected from the plurality of pieces of subject information, the target information having a similarity of a single-modality image feature of the subject information to a single-modality image feature thereof greater than a first similarity threshold.

At step S309, for each piece of target information in the third set of target information, a similarity score of the target information is calculated based on a similarity of the first modality feature of the retrieval information and the first modality feature of the target information and a similarity of the first modality feature of the retrieval information and the second modality feature of the target information.

According to some embodiments, the similarity score of the target information is selected as the maximum of the similarity of the first modal feature of the search information to the first modal feature of the target information and the similarity of the first modal feature of the search information to the second modal feature of the target information.

At step S311, the first group of target information is selected from the third group of target information based on the similarity score of each piece of target information in the third group of target information.

According to some embodiments, a second predetermined number of pieces of target information having a higher similarity score are selected from the third group of target information based on the similarity score of each piece of target information in the third group of target information. According to further embodiments, target information in the third set of target information whose similarity score is higher than the similarity threshold is selected as the first set of target information. According to still further embodiments, all of the third set of target information is selected as the first set of target information, and the information in the first set of target information is sorted based on the similarity score.

At step S313, a search result is generated based on the first set of target information.

According to some embodiments, the search result is generated based on the similarity score of the first set of target information.

In the retrieval method for the multimodal information repository provided by the present disclosure, according to some embodiments, steps S303-S307 are performed between step S301 and step S309. According to other embodiments, steps S301-S309 may be performed in other orders, such as first performing steps S303-S307, then performing step S301, and then performing step S309.

In the retrieval method for a multimodal information library according to the embodiment of the present disclosure, since the single-modality image feature is extracted from the subject information in the image information and the third group of target information among the plurality of targets is preliminarily screened based on the similarity of the single-modality image feature of the subject information and the single-modality image feature of each of the plurality of pieces of target information, the retrieval accuracy when the user inputs only the image information is improved.

According to some embodiments, the retrieval method for a multimodal information repository as described in the present disclosure further comprises, before generating the retrieval result based on the first set of target information: in response to receiving retrieval information including first modality information and second modality information, extracting first modality features of the retrieval information using a first multi-modality feature extraction module, and extracting second modality features of the retrieval information using a second multi-modality feature extraction module; generating multi-modal features of the retrieval information based on the first modal features and the second modal features of the retrieval information; and selecting a first group of target information in the plurality of pieces of target information based on the similarity between the multi-modal features of the retrieval information and the multi-modal features of each piece of target information, wherein the multi-modal features of each piece of target information are generated based on the first modal features and the second modal features of the target information.

FIG. 4 shows a flow diagram of a retrieval method 400 for a multimodal information repository, in accordance with an embodiment of the present disclosure.

At step S401, it is determined whether the received retrieval information includes first modality information and second modality information, wherein in response to the determination result being no, proceeding to step S403, and in response to the determination result being yes, proceeding to step S407.

At step S403, in response to receiving the retrieval information including the first modality information, using the first multi-modal feature extraction module, extracting first modality features of the retrieval information from the first modality information of the retrieval information. According to some embodiments, step S403 may be performed similarly to step S201.

At step S405, a first set of target information of the plurality of pieces of target information is selected based on a similarity of the first modal feature of the retrieved information to each of the first modal feature and the second modal feature of each of the plurality of pieces of target information. According to some embodiments, step S405 may be performed similarly to step S203.

At step S407, a first modal feature of the retrieval information is extracted using the first multimodal feature extraction module, and a second modal feature of the retrieval information is extracted using the second multimodal feature extraction module.

At step S409, based on the first modal feature and the second modal feature of the retrieved information, a multi-modal feature of the retrieved information is generated.

According to some embodiments, generating multimodal features of retrieved information comprises: for each of the first modal characteristic and the second modal characteristic of the retrieval information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the retrieval information to obtain the multi-modal characteristic of the retrieval information.

At step S411, a first set of target information of the plurality of pieces of target information is selected based on a similarity of the multimodal features of the retrieval information and the multimodal features of each of the plurality of pieces of target information.

At step S413, a search result is generated based on the first set of target information.

According to some embodiments, when the retrieval information includes the first modality information, step S413 may be performed similarly to step S205. According to further embodiments, when the search information includes first modality information and second modality information, the search result may be generated based on a similarity of multi-modal features of the search information to multi-modal features of each piece of target information of the first set of target information.

According to some embodiments, the first and second multi-modal feature extraction modules are trained based on a loss function, wherein the loss function is a function of similarity between features extracted by the first and second multi-modal feature extraction modules, respectively.

By setting the loss function as a function of the similarity between the features extracted by the first and second multimodal feature extraction modules, respectively, the first and second multimodal feature extraction modules can be trained together to shorten the distance between the modal features of the samples whose modal information matches each other and to lengthen the distance between the modal features of the samples whose modal information does not match.

Embodiments of the present disclosure also provide a management method for a multimodal information repository, the method including: in response to receiving warehousing information comprising first modality information and second modality information, extracting first modality features of the warehousing information from the first modality information of the warehousing information by using a first multi-modality feature extraction module, and extracting second modality features of the warehousing information from the second modality information of the warehousing information by using a second multi-modality feature extraction module; calculating multi-modal characteristics of the warehousing information based on the first modal characteristics and the second modal characteristics of the warehousing information; generating one or more retrieval objects corresponding to the warehousing information in a multi-modal information library based on the first modal characteristic, the second modal characteristic and the multi-modal characteristic of the warehousing information; and in response to receiving the retrieval information, performing a retrieval method as described in the present disclosure.

FIG. 5 shows a flow diagram of a management method 500 for a multimodal information repository, in accordance with an embodiment of the present disclosure.

At step S501, in response to receiving warehousing information including first modality information and second modality information, a first modality feature of the warehousing information is extracted from the first modality information of the warehousing information using a first multi-modality feature extraction module, and a second modality feature of the warehousing information is extracted from the second modality information of the warehousing information using a second multi-modality feature extraction module.

At step S503, multi-modal features of the binned information are calculated based on the first modal features and the second modal features of the binned information.

According to some embodiments, computing multi-modal features of the binned information comprises: for each of the first modal characteristic and the second modal characteristic of the warehousing information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the warehousing information to obtain the multi-modal characteristic of the retrieval information.

At step S505, one or more retrieval objects corresponding to the binned information in the multi-modal information bin are generated based on the first modal characteristics, the second modal characteristics, and the multi-modal characteristics of the binned information.

In some embodiments, each feature of the binned information is added to a corresponding index file of the multimodal information repository, e.g., a first modality feature is added to a search file corresponding to the first modality feature and a second modality feature is added to a search file corresponding to the second modality feature, to facilitate independent searching of the various features. In some embodiments, a search object corresponding to the binned information is created in a search file of the multimodal information bin, wherein the search object includes a corresponding characteristic of the binned information, an ID, a related web link, and the like.

At step S507, in response to receiving the retrieval information, target information corresponding to the retrieval information is retrieved in the multimodal information repository. According to some embodiments, a retrieval method as described in the present disclosure is performed to retrieve target information corresponding to the retrieved information in a modal search library.

In some embodiments, the first modality information is any one of image information and text information, and the second modality information is the other one of image information and text information, wherein the management method for the multimodal information base as disclosed in the present disclosure further includes, before generating one or more retrieval objects corresponding to the binned information in the multimodal information base: extracting one or more pieces of main body information of the warehousing information from the image information of the warehousing information by using a main body detection module; and extracting the single-mode image features of the warehousing information from one or more pieces of main information of the warehousing information by using a first multi-mode feature extraction module, a second multi-mode feature extraction module and an image feature extraction module, and wherein the generating of one or more retrieval objects corresponding to the warehousing information in the multi-mode information library comprises: and generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic, the multi-modal characteristic and the single-modal image characteristic of the warehousing information.

FIG. 6 shows a flow diagram of a method 600 for management of a multimodal information repository, in accordance with an embodiment of the present disclosure.

At step S601, in response to receiving the warehousing information including the first modality information and the second modality information, the first modality feature of the warehousing information is extracted from the first modality information of the warehousing information using the first multimodal feature extraction module, and the second modality feature of the warehousing information is extracted from the second modality information of the warehousing information using the second multimodal feature extraction module. According to some embodiments, step S601 may be performed similarly to step S501.

At step S603, multi-modal features of the binned information are calculated based on the first modal features and the second modal features of the binned information. According to some embodiments, step S603 may be performed similarly to step S503.

At step S605, one or more pieces of subject information of the warehousing information are extracted from the image information of the warehousing information using the subject detection module.

According to some embodiments, one or more pieces of subject information of the binned information may be extracted from the image information of the binned information in a similar manner to the way one or more pieces of subject information are extracted from the first modality information of the retrieved information using the subject detection module as described with reference to step S303.

At step S607, single-modality image features of the put-in information are extracted from one or more pieces of subject information of the put-in information using the first multi-modality feature extraction module, the second multi-modality feature extraction module, and the image feature extraction module.

At step S609, one or more retrieval objects corresponding to the binned information in the multi-modal information bin are generated based on the first modal feature, the second modal feature, the multi-modal feature, and the single-modal image feature of the binned information.

According to some embodiments, similar to that described above with reference to step S507, each feature of the binned information is added to the corresponding index file of the multimodal information library, and, in each search file of the multimodal information library, a search object corresponding to the binned information is created.

At step S611, in response to receiving the retrieval information, target information corresponding to the retrieval information is retrieved in the multimodal information repository. According to some embodiments, a retrieval method as described in the present disclosure is performed to retrieve target information corresponding to the retrieved information in a modal search library.

In the retrieval method for the multimodal information repository provided by the present disclosure, according to some embodiments, steps S603-S607 are performed between step S601 and step S609. According to other embodiments, steps S601-S609 may be performed in other sequences, for example, steps S603-S607 are performed first, then step S601 is performed, and then step S609 is performed.

According to some embodiments, the first multi-modal feature extraction module is any one of a multi-modal image extraction module and a multi-modal text extraction module, and the second multi-modal feature extraction module is the other one of the multi-modal image extraction module and the multi-modal text extraction module.

According to some embodiments, extracting the single-modality image feature of the binned information from one or more pieces of subject information of the binned information includes: for each of the image information and the one or more pieces of subject information from the put-in information, extracting a multimodal image feature of the information using a multimodal image extraction module; using a multi-modal text extraction module to extract multi-modal text features of the warehousing information from the text information of the warehousing information; for each of the image information and one or more pieces of main information of the warehousing information, calculating the similarity of the multi-modal image characteristics of the information and the multi-modal character characteristics of the warehousing information as the similarity score of the information; selecting information with the maximum similarity score from the image information and one or more pieces of main body information of the warehousing information; and using an image feature extraction module to extract the single-mode image features of the warehousing information from the information with the maximum similarity score.

Fig. 7 shows a flowchart of an example process of extracting single-modality image features of binned information from one or more pieces of subject information of the binned information (step S607) in the method of fig. 6, according to an embodiment of the present disclosure.

At step S701, for each of the image information and one or more pieces of subject information from the put-in information, a multimodal image feature of the information is extracted using a multimodal image extraction module.

At step S703, multimodal text features of the binned information are extracted from the text information of the binned information using a multimodal text extraction module.

At step S705, for each of the image information and one or more pieces of subject information of the put-in information, the similarity of the multi-modal image feature of the information to the multi-modal character feature of the put-in information is calculated as the similarity score of the information.

At step S707, information having the largest similarity score is selected from among the image information and one or more pieces of subject information of the binning information.

At step S709, using an image feature extraction module, single-modality image features of the warehousing information are extracted from the information having the largest similarity score. According to some embodiments, the information with the greatest similarity score is saved in the multimodal information repository as subject information of the binned information.

In the management method for the multi-modal information base according to the disclosure, since the image information of the warehousing information and the information closest to the text information of the warehousing information in the detected main body information are selected to extract the single-modal image features of the warehousing information, the accuracy and the image-text consistency of the single-modal image features of the extracted warehousing information are ensured.

Fig. 8 shows a block diagram of a retrieval apparatus 800 for a multimodal information repository according to an embodiment of the present disclosure.

According to some embodiments, the retrieving means 800 comprises: a retrieval feature extraction module 801 configured to: in response to receiving retrieval information including the first modal information, extracting, using a first multi-modal feature extraction module, first modal features of the retrieval information from first modal information of the retrieval information; a target matching module 802 configured to: selecting a first group of target information in the plurality of pieces of target information based on similarity between first modal features of the retrieval information and each of the first modal features and second modal features of each piece of target information, wherein the first modal features of each piece of target information are extracted from the first modal information of the target information by using the first multi-modal feature extraction module, and the second modal features of each piece of target information are extracted from the second modal information of the target information by using the second multi-modal feature extraction module; and a retrieval result generation module 803 configured to: and generating a retrieval result based on the first group of target information.

According to some embodiments, the multimodal information repository includes a plurality of target information including first modality information and second modality information.

According to some embodiments, the target matching module 802 comprises: a second target information selection module configured to: selecting a second set of target information of the plurality of pieces of target information based on a similarity of the first modal feature of the search information and the first modal feature of each piece of target information of the plurality of pieces of target information; and a first target information selection module configured to: the first set of target information is selected from the second set of target information based on a similarity of the first modal characteristics of the retrieved information to the second modal characteristics of each of the target information in the second set of targets.

According to some embodiments, the first modality information is image information, wherein the retrieving apparatus 800 further includes a subject feature extraction module including: a subject detection module configured to: extracting one or more pieces of subject information from first modality information of the retrieval information using a subject detection module; a single modality image feature extraction module configured to: for each piece of subject information, extracting a single-mode image feature of the subject information from the subject information by using an image feature extraction module; and a third target information selection module configured to: selecting a third set of the plurality of pieces of target information based on a similarity of a single-modality image feature of one or more pieces of subject information to a single-modality image feature of each piece of target information extracted from the first-modality information of the piece of target information using an image feature extraction module, and wherein the target matching module 802 includes: a similarity calculation module configured to: for each piece of target information in the third group of target information, calculating a similarity score of the target information based on the similarity of the first modal feature of the retrieval information and the first modal feature of the target information and the similarity of the first modal feature of the retrieval information and the second modal feature of the target information; and a first target information selection module configured to: the first set of target information is selected from the third set of target information based on the similarity score of each piece of target information in the third set of target information.

According to some embodiments, the one or more pieces of subject information include a plurality of pieces of subject information, and wherein the first target information selection module includes: a subject information matching module configured to: for each piece of subject information, selecting a plurality of pieces of target information corresponding to the subject information from the plurality of pieces of target information based on the similarity between the single-mode image feature of the subject information and the single-mode image feature of each piece of target information in the target information; and selecting a plurality of pieces of object information corresponding to each piece of body information as a third set of object information.

According to some embodiments, the retrieving means 800 further comprises a multi-modal feature retrieving module configured to: a multi-modal sub-feature extraction module configured to: in response to receiving retrieval information including first modality information and second modality information, extracting first modality features of the retrieval information using a first multi-modality feature extraction module, and extracting second modality features of the retrieval information using a second multi-modality feature extraction module; a multi-modal feature generation module configured to: generating multi-modal features of the retrieval information based on the first modal features and the second modal features of the retrieval information; and a first target information selection module configured to: a first set of target information of the plurality of pieces of target information is selected based on a similarity of multi-modal features of the retrieved information to multi-modal features of each of the plurality of pieces of target information, wherein the multi-modal features of each of the plurality of pieces of target information are generated based on first and second modal features of the target information.

According to some embodiments, the multimodal feature generation module comprises: a product calculation module configured to: for each of the first modal characteristic and the second modal characteristic of the retrieval information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and a normalization module configured to: and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the retrieval information to obtain the multi-modal characteristic of the retrieval information.

According to some embodiments, wherein the first modality information is any one of image information and text information, and the second modality information is the other one of image information and text information.

Fig. 9 shows a block diagram of a management apparatus 900 for a multimodal information repository according to an embodiment of the present disclosure.

According to some embodiments, as shown in fig. 9, the management apparatus 900 includes: a warehousing information extraction module 901 configured to: in response to receiving warehousing information comprising first modality information and second modality information, extracting first modality features of the warehousing information from the first modality information of the warehousing information by using a first multi-modality feature extraction module, and extracting second modality features of the warehousing information from the second modality information of the warehousing information by using a second multi-modality feature extraction module; a multimodal information generation module 902 configured to: calculating multi-modal characteristics of the warehousing information based on the first modal characteristics and the second modal characteristics of the warehousing information; a search object generation module 903 configured to: generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic and the multi-modal characteristic of the warehousing information; and a retrieval apparatus 800 for multimodal information repositories as described in the present disclosure.

According to some embodiments, the first modality information is any one of image information and text information, and the second modality information is the other one of image information and text information, wherein the management apparatus 900 further includes: a warehousing subject detection module configured to: extracting one or more pieces of main body information of the warehousing information from the image information of the warehousing information by using a main body detection module; and a binning feature extraction module configured to: extracting single-modality image features of the put-in information from one or more pieces of subject information of the put-in information using a first multi-modality feature extraction module, a second multi-modality feature extraction module, and an image feature extraction module, and wherein the retrieval object generation module 903 includes: a search object generation submodule configured to: and generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic, the multi-modal characteristic and the single-modal image characteristic of the warehousing information.

According to some embodiments, the first multi-modal feature extraction module is any one of a multi-modal image extraction module and a multi-modal text extraction module, the second multi-modal feature extraction module is the other one of the multi-modal image extraction module and the multi-modal text extraction module, and wherein the binning feature extraction module comprises: a binned image extraction module configured to: for each of the image information and the one or more pieces of subject information from the put-in information, extracting a multimodal image feature of the information using a multimodal image extraction module; a binned text extraction module configured to: using a multi-modal text extraction module to extract multi-modal text features of the warehousing information from the text information of the warehousing information; a warehousing subject selection module configured to: for each of the image information and one or more pieces of main information of the warehousing information, calculating the similarity of the multi-modal image characteristics of the information and the multi-modal character characteristics of the warehousing information as the similarity score of the information; selecting information with the maximum similarity score from the image information of the warehousing information and one or more pieces of main body information; and a warehousing monomodal extraction module configured to: and using an image feature extraction module to extract the single-mode image features of the warehousing information from the information with the maximum similarity score.

According to some embodiments, the multimodal information generation module 902 includes: a binned product calculation module configured to: for each of the first modal characteristic and the second modal characteristic of the warehousing information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and a binning normalization module configured to: and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the warehousing information to obtain the multi-modal characteristic of the retrieval information.

In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, and do not violate the good customs of the public order.

According to an embodiment of the present disclosure, there is also provided an electronic device, a readable storage medium, and a computer program product.

According to some embodiments, the present disclosure provides an electronic device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method as described in the present disclosure.

According to some embodiments, the present disclosure provides a non-transitory computer readable storage medium having stored thereon computer instructions for causing a computer to perform a method as described in the present disclosure.

According to some embodiments, the present disclosure provides a computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements a method as described in the present disclosure.

Referring to fig. 10, a block diagram of a structure of an electronic device 1000, which may be a server or a client of the present disclosure, which is an example of a hardware device that may be applied to aspects of the present disclosure, will now be described. Electronic device is intended to represent various forms of digital electronic computer devices, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 10, the apparatus 1000 includes a computing unit 1001 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM)1002 or a computer program loaded from a storage unit 1008 into a Random Access Memory (RAM) 1003. In the RAM 1003, various programs and data necessary for the operation of the device 1000 can also be stored. The calculation unit 1001, the ROM 1002, and the RAM 1003 are connected to each other by a bus 1004. An input/output (I/O) interface 1005 is also connected to bus 1004.

A number of components in device 1000 are connected to I/O interface 1005, including: input section 1006, output section 1007, storage section 1008, and communication section 1009. Input unit 1006 may be any type of device capable of inputting information to device 1000, and input unit 1006 may receive input numeric or character information and generate key signal inputs related to user settings and/or function controls of the electronic device, and may include, but is not limited to, a mouse, a keyboard, a touch screen, a track pad, a track ball, a joystick, a microphone, and/or a remote control. Output unit 1007 may be any type of device capable of presenting information and may include, but is not limited to, a display, speakers, a video/audio output terminal, a vibrator, and/or a printer. The storage unit 1008 may include, but is not limited to, a magnetic disk, an optical disk. The communications unit 1009 allows the device 1000 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and/or chipsets, such as bluetooth (TM) devices, 1302.11 devices, WiFi devices, WiMax devices, cellular communication devices, and/or the like.

Computing unit 1001 may be a variety of general and/or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various dedicated Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and so forth. The computing unit 1001 performs the various methods and processes described above, such as the

methods

200, 300, 400, 500, and/or 600. For example, in some embodiments,

methods

200, 300, 400, 500, and/or 600 may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 1008. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 1000 via ROM 1002 and/or communications unit 1009. When the computer program is loaded into RAM 1003 and executed by computing unit 1001, one or more steps of

methods

200, 300, 400, 500, and/or 600 described above may be performed. Alternatively, in other embodiments, the computing unit 1001 may be configured by any other suitable means (e.g., by way of firmware) to perform the

methods

200, 300, 400, 500, and/or 600.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server with a combined blockchain.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be performed in parallel, sequentially or in different orders, and are not limited herein as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved.

Although embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it is to be understood that the above-described methods, systems and apparatus are merely exemplary embodiments or examples and that the scope of the present invention is not limited by these embodiments or examples, but only by the claims as issued and their equivalents. Various elements in the embodiments or examples may be omitted or may be replaced with equivalents thereof. Further, the steps may be performed in an order different from that described in the present disclosure. Further, various elements in the embodiments or examples may be combined in various ways. It is important that as technology evolves, many of the elements described herein may be replaced with equivalent elements that appear after the present disclosure.

Claims

1. A retrieval method for a multimodal information library, wherein the multimodal information library includes a plurality of pieces of target information, each piece of target information including first modality information and second modality information, the method comprising:

in response to receiving retrieval information including the first modal information, extracting, using a first multi-modal feature extraction module, first modal features of the retrieval information from first modal information of the retrieval information;

selecting a first group of target information in the plurality of pieces of target information based on similarity between first modal features of the retrieval information and each of the first modal features and second modal features of each piece of target information, wherein the first modal features of each piece of target information are extracted from the first modal information of the target information by using the first multi-modal feature extraction module, and the second modal features of each piece of target information are extracted from the second modal information of the target information by using the second multi-modal feature extraction module; and

and generating a retrieval result based on the first group of target information.

2. The method of claim 1, wherein said selecting a first set of target information of the plurality of pieces of target information comprises:

selecting a second set of target information of the plurality of pieces of target information based on a similarity of a first modality feature of the retrieval information and a first modality feature of each of the plurality of pieces of target information; and

selecting the first set of target information from the second set of target information based on a similarity of first modal features of the retrieved information to second modal features of each of the second set of target information.

3. The method of claim 1, wherein the first modality information is image information,

wherein the method further comprises, prior to said selecting a first set of target information of the plurality of pieces of target information:

extracting one or more pieces of subject information from first modality information of the retrieval information using a subject detection module;

for each piece of subject information, extracting a single-mode image feature of the subject information from the subject information by using an image feature extraction module; and

selecting a third set of the plurality of pieces of target information based on a similarity of a single-modality image feature of the one or more pieces of subject information and a single-modality image feature of each piece of target information extracted from the first-modality information of the piece of target information using the image feature extraction module, and

wherein the selecting a first set of target information of the plurality of pieces of target information comprises:

for each piece of target information in the third group of target information, calculating a similarity score of the target information based on the similarity of the first modal feature of the retrieval information and the first modal feature of the target information and the similarity of the first modal feature of the retrieval information and the second modal feature of the target information; and

selecting the first set of target information from the third set of target information based on a similarity score for each target information in the third set of target information.

4. The method of claim 3, wherein the one or more pieces of subject information include a plurality of pieces of subject information, and wherein the selecting a third set of target information of the plurality of pieces of target information includes:

for each piece of subject information, selecting a plurality of pieces of target information corresponding to the subject information from the plurality of pieces of target information based on the similarity between the single-mode image feature of the subject information and the single-mode image feature of each piece of target information; and

selecting a plurality of pieces of target information corresponding to each piece of body information as the third group of target information.

5. The method of claim 1, further comprising, prior to generating search results based on the first set of target information:

in response to receiving retrieval information including the first modality information and the second modality information, extracting, using the first multi-modality feature extraction module, first modality features of the retrieval information, and extracting, using the second multi-modality feature extraction module, second modality features of the retrieval information;

generating multi-modal features of the search information based on the first modal features and the second modal features of the search information; and

selecting a first group of target information in the plurality of pieces of target information based on similarity between multi-modal features of the retrieval information and multi-modal features of each piece of target information, wherein the multi-modal features of each piece of target information are generated based on the first modal features and the second modal features of the target information.

6. The method of claim 5, wherein the generating multi-modal features of the retrieved information comprises:

for each of the first modal characteristic and the second modal characteristic of the retrieval information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and

and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the retrieval information to obtain the multi-modal characteristic of the retrieval information.

7. The method according to any one of claims 1-6, wherein the first and second multi-modal feature extraction modules are trained based on a loss function, wherein the loss function is a function of similarity between features extracted by the first and second multi-modal feature extraction modules, respectively.

8. The method according to any one of claims 1-6, wherein the first modality information is either one of image information and text information, and the second modality information is the other one of image information and text information.

9. A method for managing a multimodal information repository, the method comprising:

in response to receiving warehousing information comprising first modality information and second modality information, extracting first modality features of the warehousing information from the first modality information of the warehousing information by using a first multi-modality feature extraction module, and extracting second modality features of the warehousing information from the second modality information of the warehousing information by using a second multi-modality feature extraction module;

calculating multi-modal characteristics of the warehousing information based on the first modal characteristics and the second modal characteristics of the warehousing information;

generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic and the multi-modal characteristic of the warehousing information; and

in response to receiving the retrieval information, performing the retrieval method of any one of claims 1-8.

10. The method according to claim 9, wherein the first modality information is any one of image information and text information, the second modality information is the other one of image information and text information,

wherein the method further comprises, prior to said generating one or more search objects in the multimodal information repository corresponding to the binned information:

using a main body detection module to extract one or more pieces of main body information of the warehousing information from the image information of the warehousing information; and

extracting single-mode image features of the warehousing information from one or more pieces of subject information of the warehousing information using the first multi-mode feature extraction module, the second multi-mode feature extraction module, and the image feature extraction module, and,

wherein the generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information repository comprises:

and generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic, the multi-modal characteristic and the single-modal image characteristic of the warehousing information.

11. The method of claim 10, wherein the first multi-modal feature extraction module is any one of a multi-modal image extraction module and a multi-modal text extraction module, the second multi-modal feature extraction module is the other one of the multi-modal image extraction module and the multi-modal text extraction module, and

wherein the extracting of the single-mode image features of the warehousing information from one or more pieces of subject information of the warehousing information includes:

for each of the image information and one or more pieces of subject information from the put-in information, extracting, using the multimodal image extraction module, multimodal image features of the information;

using the multi-modal text extraction module to extract multi-modal text features of the warehousing information from the text information of the warehousing information;

for each of the image information and one or more pieces of main information of the warehousing information, calculating the similarity of the multi-modal image characteristics of the information and the multi-modal character characteristics of the warehousing information as the similarity score of the information;

selecting information with the maximum similarity score from the image information and one or more pieces of main body information of the warehousing information; and

and using the image feature extraction module to extract the single-mode image features of the warehousing information from the information with the maximum similarity score.

12. The method of claim 9, wherein the calculating multi-modal characteristics of the binned information comprises:

for each of the first modal characteristic and the second modal characteristic of the warehousing information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and

and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the warehousing information to obtain the multi-modal characteristic of the retrieval information.

13. A retrieval apparatus for a multimodal information library, wherein the multimodal information library includes a plurality of pieces of target information, each piece of target information including first modality information and second modality information, the apparatus comprising:

a retrieval feature extraction module configured to: in response to receiving retrieval information including the first modal information, extracting, using a first multi-modal feature extraction module, first modal features of the retrieval information from first modal information of the retrieval information;

a target matching module configured to: selecting a first group of target information in the plurality of pieces of target information based on similarity between first modal features of the retrieval information and each of the first modal features and second modal features of each piece of target information, wherein the first modal features of each piece of target information are extracted from the first modal information of the target information by using the first multi-modal feature extraction module, and the second modal features of each piece of target information are extracted from the second modal information of the target information by using the second multi-modal feature extraction module; and

a retrieval result generation module configured to: and generating a retrieval result based on the first group of target information.

14. The apparatus of claim 13, wherein the target matching module comprises:

a second target information selection module configured to: selecting a second set of target information of the plurality of pieces of target information based on a similarity of a first modality feature of the retrieval information and a first modality feature of each of the plurality of pieces of target information; and

a first target information selection module configured to: selecting the first set of target information from the second set of target information based on a similarity of first modal features of the retrieved information to second modal features of each of the second set of target information.

15. The apparatus of claim 13, wherein the first modality information is image information,

wherein, the device still includes the main part characteristic extraction module, includes:

a subject detection module configured to: extracting one or more pieces of subject information from first modality information of the retrieval information using a subject detection module;

a single modality image feature extraction module configured to: for each piece of subject information, extracting a single-mode image feature of the subject information from the subject information by using an image feature extraction module; and

a third target information selection module configured to: selecting a third set of the plurality of pieces of target information based on a similarity of a single-modality image feature of the one or more pieces of subject information and a single-modality image feature of each piece of target information extracted from the first-modality information of the piece of target information using the image feature extraction module, and

wherein the target matching module comprises:

a similarity calculation module configured to: for each piece of target information in the third group of target information, calculating a similarity score of the target information based on the similarity of the first modal feature of the retrieval information and the first modal feature of the target information and the similarity of the first modal feature of the retrieval information and the second modal feature of the target information; and

a first target information selection module configured to: selecting the first set of target information from the third set of target information based on a similarity score for each target information in the third set of target information.

16. The apparatus of claim 15, wherein the one or more pieces of subject information comprise a plurality of pieces of subject information, and wherein the first target information selection module comprises:

a subject information matching module configured to:

17. The apparatus of claim 13, further comprising a multimodal feature retrieval module configured to:

a multi-modal sub-feature extraction module configured to: in response to receiving retrieval information including the first modality information and the second modality information, extracting, using the first multi-modality feature extraction module, first modality features of the retrieval information, and extracting, using the second multi-modality feature extraction module, second modality features of the retrieval information;

a multi-modal feature generation module configured to: generating multi-modal features of the search information based on the first modal features and the second modal features of the search information; and

a first target information selection module configured to: selecting a first group of target information in the plurality of pieces of target information based on similarity between multi-modal features of the retrieval information and multi-modal features of each piece of target information, wherein the multi-modal features of each piece of target information are generated based on the first modal features and the second modal features of the target information.

18. The apparatus of claim 17, wherein the multi-modal feature generation module comprises:

a product calculation module configured to: for each of the first modal characteristic and the second modal characteristic of the retrieval information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and

a normalization module configured to: and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the retrieval information to obtain the multi-modal characteristic of the retrieval information.

19. The apparatus according to any one of claims 13-18, wherein the first and second multi-modal feature extraction modules are trained based on a loss function, wherein the loss function is a function of similarity between features extracted by the first and second multi-modal feature extraction modules, respectively.

20. The apparatus according to any one of claims 13-18, wherein the first modality information is either one of image information and text information, and the second modality information is the other one of image information and text information.

21. A management apparatus for a multimodal information repository, comprising:

a warehousing information extraction module configured to: in response to receiving warehousing information comprising first modality information and second modality information, extracting first modality features of the warehousing information from the first modality information of the warehousing information by using a first multi-modality feature extraction module, and extracting second modality features of the warehousing information from the second modality information of the warehousing information by using a second multi-modality feature extraction module;

a multimodal information generation module configured to: calculating multi-modal characteristics of the warehousing information based on the first modal characteristics and the second modal characteristics of the warehousing information;

a search object generation module configured to: generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic and the multi-modal characteristic of the warehousing information; and

a retrieval arrangement for a multimodal information repository as claimed in any one of claims 13-20.

22. The apparatus according to claim 21, wherein the first modality information is any one of image information and text information, the second modality information is the other one of image information and text information,

wherein the apparatus further comprises:

a warehousing subject detection module configured to: using a main body detection module to extract one or more pieces of main body information of the warehousing information from the image information of the warehousing information; and

a binning feature extraction module configured to: extracting single-mode image features of the warehousing information from one or more pieces of subject information of the warehousing information using the first multi-mode feature extraction module, the second multi-mode feature extraction module, and the image feature extraction module, and,

wherein the retrieval object generation module comprises:

a search object generation submodule configured to: and generating one or more retrieval objects corresponding to the warehousing information in the multi-modal information library based on the first modal characteristic, the second modal characteristic, the multi-modal characteristic and the single-modal image characteristic of the warehousing information.

23. The apparatus of claim 22, wherein the first multi-modal feature extraction module is any one of a multi-modal image extraction module and a multi-modal text extraction module, the second multi-modal feature extraction module is the other one of the multi-modal image extraction module and the multi-modal text extraction module, and

wherein, the storage characteristic extraction module comprises:

a binned image extraction module configured to: for each of the image information and one or more pieces of subject information from the put-in information, extracting, using the multimodal image extraction module, multimodal image features of the information;

a binned text extraction module configured to: using the multi-modal text extraction module to extract multi-modal text features of the warehousing information from the text information of the warehousing information;

a warehousing subject selection module configured to:

for each of the image information and one or more pieces of main information of the warehousing information, calculating the similarity of the multi-modal image characteristics of the information and the multi-modal character characteristics of the warehousing information as the similarity score of the information; and

a warehousing monomodal extraction module configured to: and using the image feature extraction module to extract the single-mode image features of the warehousing information from the information with the maximum similarity score.

24. The apparatus of claim 21, wherein the multimodal information generation module comprises:

a binned product calculation module configured to: for each of the first modal characteristic and the second modal characteristic of the warehousing information, multiplying the modal characteristic by the weight corresponding to the modal characteristic to obtain a product corresponding to the modal characteristic; and a binning normalization module configured to: and normalizing the sum of products corresponding to the first modal characteristic and the second modal characteristic of the warehousing information to obtain the multi-modal characteristic of the retrieval information.

25. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein

The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-12.

26. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-12.

27. A computer program product comprising a computer program, wherein the computer program realizes the method of any one of claims 1-12 when executed by a processor.