A search system based on knowledge graph

Abstract

The present invention discloses a search system based on a knowledge graph, comprising: a knowledge graph part, which includes a tree structure and a molecular structure, wherein the tree structure is set as an upper unit-lower unit; the molecular structure is set as a macromolecule keyword-N small molecule keywords, and the small molecule keyword represents a certain aspect of the macromolecule keyword; a keyword operation part, which includes conventional operations and additional operations, wherein the conventional operations include adding or deleting, and the additional operations include generalization, decomposition and synthesis. The generalization operation means that when a keyword is deleted, a prompt of the upper unit of the deleted keyword will appear at the same time according to the tree structure. The decomposition operation means that a macromolecule keyword is decomposed into multiple small molecule keywords according to the molecular structure. The synthesis operation means that multiple small molecule keywords are randomly generated with reference to the molecular structure to generate suitable new macromolecule keywords.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular to a search system based on a knowledge graph. Background Art

[0002] In today's world of explosive information growth, content search systems play a central role in any information, database, or network system. The most common content search systems are keyword-based. Based on the keywords entered by the searcher, the system returns a series of search results, including the most relevant, important, or recent results.

[0003] Common search aids include: Before a search, the machine automatically generates supplementary tags in the search bar based on machine statistics and recommends them to the searcher, usually within eight. After a search, the machine automatically generates supplementary search links based on machine statistics and recommends them to the searcher, usually within eight. Search aids can be thought of as specialized content recommendation systems. Search systems generally assume that users are clear about their purpose and can enter specific keywords. However, users may not always have a clear intention. Content recommendation systems were developed to address this need.

[0004] Whether it's a keyword search system or a content recommendation system, the system's internal logical structure is disconnected from the user's own cognitive structure. For example, a user's like or dislike of a piece of content may have multiple reasons. Current systems, lacking direct user interaction with the system's internal logical structure, only see the results of likes or dislikes, not the user's cognitive structure. This makes it difficult for the system to adapt to the rapidly shifting, irregular, and cyclical interests of humans in different situations. In this case, even if the content recommendation system itself has several different columns, most of the recommendations are based on statistical results and have a fixed tone. Over time, users will still easily become bored.

[0005] Relatively speaking, keyword search systems can meet the needs of users' rapidly changing interests, but the biggest problem with keyword search systems is that users sometimes cannot say the keywords of the content they are interested in, and sometimes the results must be displayed first before users can discover their interests. The current search system uses search assistants, that is, displaying no more than eight entries at a time, to help users discover their interests. The problem here is that when a user enters a keyword search system, it is likely that he has already thought of the "keywords". In this case, the so-called search assistant is likely to have nothing to do with the "keywords" that the user has thought of. The "keywords" generated by the search assistant are not generated based on the user. Therefore, for most users, it may be a noisy, useless, or even negative user experience. Therefore, users will not interact with the recommendation assistant. Without interaction, the search assistant is difficult to "cold start." Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a search system based on knowledge graphs and propose a set of improvement methods, which can not only improve the keyword search system and content recommendation system separately, but also use a unified framework to combine the two to provide users with a unified and smooth user experience.

[0007] The above solution is a search system based on knowledge graph that integrates keyword search system and content recommendation system.

[0008] If we understand our proposed new system as a keyword search system, we can say that we replace keywords with keywords in the knowledge graph; if we understand our proposed new system as a recommendation system, we can say that the system no longer directly generates content, but generates a set of intermediaries between computers and humans, namely knowledge graph keywords.

[0009] The above technical objectives of the present invention are achieved through the following technical solutions:

[0010] A search system based on a knowledge graph, comprising:

[0011] The knowledge graph part includes two structures: tree structure and molecular structure. The tree structure is set as upper unit-lower unit; the molecular structure is set as macromolecule keyword-N small molecule keywords, and the small molecule keyword represents a certain aspect of the macromolecule keyword.

[0012] The keyword operation section includes regular operations and additional operations. Regular operations include adding or deleting, and additional operations include generalization, decomposition, and synthesis. Generalization operations refer to prompts based on tree structures, while decomposition and synthesis operations refer to prompts based on molecular structures.

[0013] More preferably, the generalized operation means that when a keyword is deleted, a prompt of the upper unit of the deleted keyword will appear at the same time according to the tree structure.

[0014] More preferably, the decomposition operation refers to decomposing a macromolecule keyword into multiple small molecule keywords based on the molecular structure, and the synthesis operation refers to randomly generating appropriate new macromolecule keywords from multiple small molecule keywords with reference to the molecular structure.

[0015] More preferably, the search system further includes a storage list representing favorite keyword combinations, the keyword combination storage list includes a new column, an option to add a keyword tag, and a stored keyword combination list, and the stored keyword combination list is provided with a keyword tag.

[0016] More preferably, the keyword tag is set as text or graphics.

[0017] In summary, the present invention has the following beneficial effects compared to the prior art:

[0018] First of all, these operations are different from those of the input method, both in purpose and result. The purpose of the input method is to input characters, while these operations are performed based on the meaning of the characters. Therefore, these operations are not mutually exclusive, but complementary.

[0019] Similarly, we must compare these methods with the search assistance provided by the system in the keyword search system. The difference here is that the focus of search assistance is to guess what the user wants, while using our method, the user can generate enough options he wants through his own operations, so the need for the system to provide search assistance is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which:

[0021] Figure 1 The mobile phone application operation interface of the embodiment;

[0022] Figure 2 A list of keyword combinations for an embodiment is stored;

[0023] Figure 3 This is the keyword display area in the embodiment. DETAILED DESCRIPTION

[0024] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make this disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art. The "embodiment" or "implementation method" in the specification may represent one embodiment or one implementation method, or may represent some embodiments or some implementation methods.

[0025] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present disclosure may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0026] According to an embodiment of the present invention, a search system based on a knowledge graph is proposed.

[0027] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0028] A search system based on a knowledge graph, comprising:

[0029] The knowledge graph part includes two structures: tree structure and molecular structure. The tree structure is set as upper unit-lower unit; the molecular structure is set as macromolecule keyword-N small molecule keywords, and the small molecule keyword represents a certain aspect of the macromolecule keyword.

[0030] The keyword operation part includes regular operations and additional operations. Regular operations include adding or deleting, and additional operations include generalization, decomposition and synthesis. Generalization operation means that when deleting a keyword, a prompt of the upper unit of the deleted keyword will appear at the same time according to the tree structure. Decomposition operation means decomposing a large molecule keyword into multiple small molecule keywords according to the molecular structure. Synthesis operation means randomly generating suitable new large molecule keywords from multiple small molecule keywords with reference to the molecular structure.

[0031] A knowledge graph is a set of keywords that is pre-processed by the system, can be continuously improved, and has a structure. We list two representative structures: tree structure and molecular structure.

[0032] First, let's use an example to illustrate a tree structure. This tree structure is similar to a library classification system. Suppose we have keywords such as "Japan," "China," "USA," "Shanghai," "Asia," "Americas," and "World." A possible structure is that the parent unit of "Shanghai" is "China," the parent unit of "China" is "Asia," the parent unit of "Japan" is "Asia," the parent unit of "USA" is "Americas," the parent unit of "Asia" is "World," and so on. It's worth noting that in our knowledge graph design, a keyword will only have one primary parent unit, but can have multiple child units.

[0033] The knowledge graph means that when we search for content related to "Asia", we may see content related to "Shanghai", even if the keyword "Asia" does not appear in this content. We will introduce related operations based on the tree structure later.

[0034] Molecular structure means that a keyword, like a large molecule, can be repeatedly broken down into smaller molecules, all the way down to atoms. For example, the concept "Kiyomizu-dera Temple" can be broken down into keywords such as "Japan," "Architecture," and "Famous Attraction." Unlike a tree structure, a single keyword can be broken down into multiple keywords.

[0035] The keyword search system does not have many options when entering keywords. Users can enter / modify the string themselves, or use the system-recommended string.

[0036] The search system based on the knowledge graph has more operation methods for entering keywords. Similar to the keyword search system, we have "add" or "delete" keywords. In addition, based on the structure of the knowledge graph, we have additional operations such as "generalization", "decomposition", and "synthesis".

[0037] Keyword search systems all have the ability to add keywords to the search bar, whether adding bytes or strings. Knowledge graph-based search systems also offer new ways to add keywords through "generalization," "decomposition," and "synthesis."

[0038] "Generalization" means that when we delete a keyword, a prompt of the upper-level unit will appear at the same time, and the user can select the upper-level unit as the new keyword. Since the upper-level unit will be broader than the deleted keyword, we call this operation "generalization."

[0039] Decomposition means we can break down a keyword into multiple keywords based on its "molecular structure." For example, we can break down "Kiyomizu-dera Temple" into "Japan," "architecture," and "famous attractions." This type of decomposition is relatively stable.

[0040] "Synthesis" means we can combine several keywords and randomly generate new keywords based on their "molecular structure." For example, we can randomly combine "Japan," "architecture," and "famous attractions" to create "Kinkaku-ji Temple" or other suitable keywords. Each synthesis will generally generate different results, with the number of possibilities ranging from dozens to hundreds.

[0041] Here, we must compare the above methods with users entering keywords by themselves. First of all, our operations are different from the input method, from the purpose to the result. Because the purpose of the input method is to input words, and these operations are performed according to the meaning of the words, so these operations are not mutually exclusive, but complement each other.

[0042] Similarly, we must compare these methods with the search assistance provided by the system in the keyword search system. The difference here is that the focus of search assistance is to guess what the user wants, while using our method, the user can generate enough options he wants through his own operations, so the need for the system to provide search assistance is reduced.

[0043] Next, we'll explore the structure of the knowledge graph and how it can transform content recommendation systems. As we mentioned earlier, content recommendation systems have a large number of hidden parameters. Our goal is to transform content recommendation systems into more like keyword search systems. For example, there might be a keyword called "interest" that represents user interests, and a keyword called "hotspot" that represents statistically significant hot topics.

[0044] When the system detects a user's particular interest in certain topics based on user behavior, such as browsing time and details viewed, it can proactively adjust the keywords on the current page. For example, if a topic changes from "interest" to "interest plus travel," once the user has perceived the system's intent, they can interact with the content recommendation system using the aforementioned methods of "addition," "deletion," "generalization," "decomposition," and "synthesis," quickly and accurately reflecting the user's short-term interests.

[0045] It is worth noting here that when using the knowledge graph structure to transform the content recommendation system, all pages are generated by operational keywords, which also means that the internal logic of all pages is consistent. This is different from the common hybrid keyword search system and content recommendation system methods. In the common method, each page can be classified as belonging to the keyword search system or the content recommendation system, while in the method proposed in this patent, the operation of each page is consistent. We can regard the content recommendation system as a search system with special keywords such as "interest" and "hot spots". The difference between the two may be that the keyword search system tends to present the "most relevant" content, while the recommendation system tends to provide "new" content.

[0046] Below we use a mobile phone application implementation to specifically illustrate the search system proposed in this invention. Figure 1 Represents the main operation interface.

[0047] 11 is the information flow display area, which is composed of multiple 111s (a piece of content, which may include a title, picture, date, and other related information). Users can slide up and down, left and right to obtain more information.

[0048] 12 is the keyword display area, which contains 121, 122, and 123;

[0049] 121 is the keyword label, which can be text or graphics and can be switched by clicking;

[0050] 122 is the “Favorite” button;

[0051] 123 is the "Go" button;

[0052] 124 is a button for restoring the previous step.

[0053] Figure 2 Represents a list of favorite keyword combinations

[0054] 21 represents a new column;

[0055] 211 stands for adding keyword tags;

[0056] 22 represents the stored keyword combination;

[0057] 221 represents a keyword tag, which can be text or a graphic.

[0058] Figure 3 Represents a list of favorite keyword combinations

[0059] 31 represents a new column;

[0060] 311 represents a keyword tag, which can be text or graphics and can be switched by clicking;

[0061] 32 represents searching for keyword tags using text;

[0062] 321 represents the search box;

[0063] 322 represents the recommended keyword, which can be text or graphics and can be switched by clicking;

[0064] 33 represents the main keyword tree structure display area;

[0065] 331 represents the keyword at the top level of the tree structure, which can be text or graphics. You can click to expand the lower level.

[0066] The specific operations are as follows:

[0067] The first page users see is the information flow page (11), where users can perform common operations. These operations are slightly different depending on whether we are implementing a search system or a content recommendation system. Generally speaking, search systems tend to place jump links at (111), while content recommendation systems usually allow users to view details, like, comment, etc.

[0068] The keyword display area (12) below lists several keywords (121) that define this information flow. Users can add favorites (122). The go button (123) and the restore button (124) are usually hidden.

[0069] The user has several ways to operate in the display area (12). Clicking on a keyword (121) can switch the image / text display; long pressing on a keyword (121) can drag it. Dragging it outside (12) is considered "deleting" and "generalizing" occurs at the same time, with its temporary upper structure keyword appearing at the original keyword position in (12). If the user clicks on the temporary keyword, it will become a normal keyword. If the user clicks elsewhere, the temporary keyword disappears. The user can also drag the brick keyword to other keywords, and the system will automatically determine whether it can be "synthesized". If it can be synthesized, the original keyword disappears and a new keyword is generated. Long pressing on a keyword (121) and not moving it for a certain period of time will trigger the "decomposition" action, the original keyword disappears, and several component temporary keywords appear. Similarly, if the user clicks on the temporary keyword, it will become a normal keyword. If the user clicks elsewhere, the temporary keyword disappears.

[0070] After directly operating the keyword display area (12), the favorite button (122) will disappear, and the go button (123) and the restore button (124) will appear. Before executing the go button (123), you can use (124) to restore the previous step. After executing the go button (123), the system will generate a new information flow (11) based on the new keyword combination. After that, the favorite button (122) will appear, and the go button (123) and the restore button (124) will disappear.

[0071] Long press the blank space in the keyword display area (12) to enter the Figure 3 , and the keyword display area (31) will be a copy of the keyword display area (12).

[0072] Long press the favorite (122) to enter directly Figure 2 .

[0073] The user's activities on the information flow page (11), such as clicking on details, swiping left or right, may cause changes to the keywords (121) in the keyword display area (12) depending on different implementation plans. In this case, the information flow page (11) has been redefined.

[0074] Figure 2 The operation is as follows: The user can choose to add a new column (21), click Add keyword (211), and enter Figure 3 Users can also select a keyword combination (22) that they have already “favorited” and click to enter Figure 1 , and keyword combination (121) is a copy of keyword combination (221).

[0075] Figure 3The operation is as follows: The keyword display area (31) is similar to the keyword display area (12). Users can perform operations such as "delete", "generalize", "decompose", and "synthesize". Users can enter a character string through the character search bar (321) in the new keyword area (32). The system will display the appropriate keyword (322). After long pressing, it can be added to the display keyword area (31). The keyword tree structure display area (33) provides another method for adding keywords. Clicking the top tree structure keyword (331) can expand the tree structure. Clicking again can close the tree structure. Long pressing can add the keyword (331) to the display area (31).

[0076] We propose to establish a user-specific progress bar above each keyword. The system can increase the progress bar of each keyword based on the user's usage time and interaction to unlock the daily usage of special functions such as "generalization", "decomposition", and "synthesis".

[0077] As users themselves become experts in the "keyword" over time, they are also better able to use the system to operate this keyword, discover other related keywords, and further expand their territory.

[0078] The above description is merely an exemplary embodiment of the present invention and is not intended to limit the scope of protection of the present invention. The scope of protection of the present invention is determined by the appended claims.

Claims

1. A search system based on knowledge graph, characterized in that: include: The knowledge graph part includes two structures: tree structure and molecular structure. The tree structure is set as upper unit-lower unit; The molecular structure is set as macromolecule keyword-N small molecule keywords, where the small molecule keyword represents a certain aspect of the macromolecule keyword; Keyword operation section, including regular operations and additional operations. Regular operations include adding or deleting, and additional operations include generalization, decomposition, and synthesis operations. Generalization operations refer to prompts based on tree structures, while decomposition and synthesis operations refer to prompts based on molecular structures. An operation interface, the operation interface including an information display area and a keyword display area, the keyword display area lists a number of keywords that define the information flow in the information display area; The keyword can be deleted by long pressing the keyword and dragging it outside the keyword display area. At the same time, a general operation occurs. The general operation means that when a keyword is deleted, its temporary upper-level unit keyword appears in the original keyword position; By dragging a keyword onto other keywords to trigger the synthesis operation, the system will automatically determine whether the synthesis can be performed. If it can be synthesized, the original keyword will disappear. The synthesis operation refers to randomly generating a new macromolecule keyword by referring to the molecular structure of multiple small molecule keywords, and each synthesis will generate a different result; By long pressing a keyword without moving it for a certain period of time, the decomposition operation will be triggered and the original keyword will disappear. The decomposition operation is relatively fixed, which means that a large molecule keyword is decomposed into multiple small molecule keywords according to the molecular structure, and several tentative keywords of the components will appear; By creating a user-specific progress bar on each keyword and increasing the progress bar of each keyword based on the user's usage time and interaction, the daily usage of generalization, decomposition and synthesis operations is unlocked; and By clicking on details or sliding left and right in the information display area, the keywords in the keyword display area will be changed, so that the information display area will be redefined.

2. A search system based on knowledge graph according to claim 1, characterized in that: The search system also includes a storage list for favorite keyword combinations, the keyword combination storage list including a new column, an option to add a keyword tag, and a stored keyword combination list, wherein the stored keyword combination list is provided with a keyword tag.

3. A search system based on knowledge graph according to claim 2, characterized in that: Keyword tags are set as text or graphics.

Images