CN118140219A - Interactive next query recommendation based on knowledge attributes and paragraph information - Google Patents

Abstract

The computing system receives an indication that a user has selected a paragraph presented on a display that is shown on a Search Engine Results Page (SERP). Upon receiving the indication, the computing system identifies a plurality of suggested queries related to the paragraph, wherein the plurality of suggested queries are generated based on the paragraph and an entry for the entity in the knowledge-graph. After identifying the plurality of suggested queries, the computing system presents the plurality of suggested queries in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP. When a query of the plurality of suggested queries is selected, a second SERP is presented on the display, wherein the second SERP is based on the query.

Description

Interactive next query recommendation based on knowledge attributes and paragraph information

Background

Traditionally, search engines receive queries from a user's computing device. The search engine obtains search results based on the query and transmits a Search Engine Results Page (SERP) to the computing device for presentation to the user, wherein the SERP includes a Uniform Resource Locator (URL) of a web page related to the query. The SERP may also include suggested queries based on the query and historical queries submitted by other users. When a SERP is presented on a display, suggested queries tend to be presented at the top of the SERP or at the bottom of the SERP.

Disclosure of Invention

The following is a brief overview of the subject matter described in more detail herein. This summary is not intended to limit the scope of the claims.

Described herein are various techniques related to generating suggested queries based on (1) paragraphs displayed in a Search Engine Results Page (SERP) (or paragraphs displayed on another web page) and (2) data from a knowledge-graph. The techniques described herein learn about the user's points of interest and generate suggested queries that are contextually related to paragraphs shown on the SERP. The techniques described herein help a user explore entities referenced in paragraphs in an intuitive and non-invasive manner. In essence, the techniques described herein predict content of interest to a user and present suggested queries based on such information. Additionally, the techniques described herein are applicable to paragraphs of different lengths, from simple one-line sentences to multi-sentence paragraphs.

In example operations, the computing system obtains a paragraph from the electron source. The paragraph references an entity (e.g., person, place, event, etc.). The computing system also obtains an entry for the entity in the knowledge-graph. The knowledge graph includes nodes and edges connecting the nodes, wherein the nodes represent entities or attributes of the entities, and the edges represent relationships between the entities or between the entities and the attributes. In this way, the entries for the entities in the knowledge-graph include part of the nodes and edges of the knowledge-graph. The computing system generates a plurality of candidate suggested queries based on the entries for the entity and the paragraph text in the knowledge-graph. According to an embodiment, a computing system generates a plurality of candidate suggested queries using a generative model, such as a transformer model. The computing system removes a subset of the queries from the plurality of candidate suggested queries based on the relevance criteria, thereby generating a plurality of pruned candidate suggested queries. In an example, a computing system utilizes a machine learning model to classify a plurality of candidate suggested queries as being related or unrelated to a paragraph based on entries for an entity in a knowledge graph. The computing system ranks the plurality of pruned candidate suggested queries based on a ranking criteria, thereby generating a ranked plurality of pruned candidate suggested queries. In an example, the computing system ranks the plurality of pruned candidate suggested queries based on previous queries submitted by the plurality of users. The computing system selects a number (e.g., three) of top-ranked queries from the ranked plurality of pruned candidate suggested queries, thereby generating a plurality of suggested queries related to the paragraph.

The computing system receives a query from a user, wherein the query is related to an entity. In an example, the query is the name of a football coach: "football coach X". The computing system obtains the SERP based on the query and presents the query on the display. The SERP includes a Uniform Resource Locator (URL) of a web page related to a query and paragraph. The paragraph may be from an online encyclopedia, a knowledge graph, or another website. In an example, the paragraph is "coach X is considered the most historically biggest coach for university football. He won 7 national champions, the most historically college football. In an example, the paragraph is included in a knowledge card shown on the SERP. The knowledge card includes information about the entity, wherein at least a portion of the information included in the knowledge card is from a knowledge graph.

The computing system receives an indication that the user has selected a paragraph shown on the SERP, wherein the indication includes an identifier for the paragraph. In an example, the computing system receives the indication when a mouse cursor of a mouse operated by the user hovers over a paragraph shown on the SERP. Upon receiving the indication, the computing system identifies a plurality of suggested queries related to the paragraph. After identifying the plurality of suggested queries, the computing system presents the plurality of suggested queries in a pop-up graphical element located near the paragraph in the SERP, wherein the pop-up graphical element overlays a portion of the SERP. In an example, the plurality of suggested queries includes "most historically biggest coaches", "top university football coaches" and "university football histories".

The computing system may receive a second indication that the user has selected a query of the plurality of suggested queries shown on the SERP. In an example, the computing system receives the second indication when the query is clicked via a cursor of a user-operated mouse. The computing system obtains a second SERP based on a query of the plurality of suggested queries. The computing system presents the second SERP on a display. According to the example given above, the user selects the "top university football coach" query, and as such, the second SERP presents information about the top university football coach. The second SERP may comprise a knowledge card previously presented in the SERP or a modified version of the knowledge card, such as a knowledge card comprising different paragraphs than those in the (original) knowledge card. Alternatively, the second SERP may not include a knowledge card.

The above-described techniques present various advantages over conventional query suggestion techniques. First, unlike conventional query suggestion techniques, the above techniques combine knowledge-graph mining methods with generative machine learning techniques to present relevant suggested queries to users. Second, while conventional query suggestion techniques present suggested queries to users based on queries that the user has submitted, the techniques described above present suggested queries based on paragraphs shown within the SERP. Thus, the above-described techniques present more relevant queries to the user than conventional techniques. Third, unlike conventional query suggestion techniques that present suggested queries at the top or bottom of a SERP when presenting a SERP, the above techniques may present suggested queries to a user when a mouse cursor of a user-operated mouse hovers over a paragraph shown on the SERP, where the hovering serves as an indicator of user interest in the paragraph. Thus, the above-described techniques present suggested queries in a non-invasive, intuitive, and user-friendly manner. Fourth, by using a generative model that generates suggested queries based on paragraphs and data in the knowledge-graph, the techniques described above may generate related suggested queries without requiring other users to previously submit suggested queries to a search engine. Fifth, because the knowledge-graph includes comprehensive information about many different entities in a single location, the above-described techniques do require extensive searches to be performed on different sources in order to find information to supplement paragraphs (which are used to generate suggested queries). Thus, the above described techniques save network and storage resources.

The foregoing summary presents a simplified summary in order to provide a basic understanding of some aspects of the systems and/or methods discussed herein. This summary is not an extensive overview of the systems and/or methods discussed herein. It is not intended to identify key/critical elements or to delineate the scope of such systems and/or methods. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is presented later.

Drawings

FIG. 1 is a functional block diagram of an example computing environment that facilitates providing contextual query-intent to an interactive user.

FIG. 2 illustrates an example pipeline that facilitates providing contextual query-intent to an interactive user.

3A-3C illustrate different states of an example Search Engine Results Page (SERP) that facilitates providing contextual query-intent to an interactive user.

Fig. 4-8 are example screen shots illustrating providing contextual query-intent to an interactive user.

Fig. 9 is an illustration of an example knowledge graph.

FIG. 10 is a flowchart illustrating an example method performed by a computing system.

FIG. 11 illustrates an example computing device.

Various techniques related to providing contextual query-intent (e.g., suggested queries) to an interactive user are now described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. It may be evident, however, that such aspect(s) may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to facilitate describing one or more aspects. Furthermore, it should be understood that the functionality described as being performed by certain system components may be performed by multiple components. Similarly, for example, a component may be configured to perform functionality described as being performed by multiple components.

Detailed Description

As noted above, conventional search engines recommend queries to users based on queries submitted by users and queries previously submitted by other users. Furthermore, as noted above, conventional search engines tend to present suggested queries near the top or bottom of the query's Search Engine Results Page (SERP), which is often prone to be ignored by users.

To address these issues, a computing system is described herein that provides contextual query-intent (e.g., suggested query) to a user. The computing system generates suggested queries based on paragraphs that are shown on the SERP (or paragraphs that are shown on another web page) for the query (rather than the query itself), and thus the computing system tends to suggest more relevant queries to the user. Further, when a mouse cursor and/or a touchpad of a user's mouse hovers over the paragraph (indicating that the user is interested in the paragraph), the computing system may present the suggested query to the user, and thus the computing system may be better able to present the suggested query to the user when the user is actually interested in viewing the suggested query. Additionally, the computing system utilizes a generative model (e.g., a transformer model) to generate suggested queries (rather than simply identifying popular related queries submitted by other users), and thus may generate related suggested queries even if such suggested queries have not been previously submitted by other users. Furthermore, the computing system utilizes both the paragraph and data from the knowledge-graph to generate suggested queries, which further helps identify related suggested queries.

In example operations, the computing system obtains a paragraph from the electron source. The paragraph references an entity (e.g., person, place, event, etc.). The computing system also obtains an entry for the entity in the knowledge-graph. The knowledge graph includes nodes and edges connecting the nodes, wherein the nodes represent entities or attributes of the entities, and the edges represent relationships between the entities or between the entities and the attributes. In this way, the entries for the entities in the knowledge-graph include part of the nodes and edges of the knowledge-graph. The computing system generates a plurality of candidate suggested queries based on the paragraph text in the knowledge-graph and the entries for the entity. According to an embodiment, a computing system generates a plurality of candidate suggested queries using a generative model, such as a transformer model. The computing system removes a subset of the queries from the plurality of candidate suggested queries based on the relevance criteria, thereby generating a plurality of pruned candidate suggested queries. In an example, a computing system utilizes a machine learning model to classify a plurality of candidate suggested queries as being related or unrelated to a paragraph based on entries for an entity in a knowledge graph. The computing system ranks the plurality of pruned candidate suggested queries based on a ranking criteria, thereby generating a ranked plurality of pruned candidate suggested queries. In an example, the computing system ranks the plurality of pruned candidate suggested queries based on previous queries submitted by the plurality of users. The computing system selects a number (e.g., three) of top-ranked queries from the ranked plurality of pruned candidate suggested queries, thereby generating a plurality of suggested queries related to the paragraph.

The computing system receives a query from a user, wherein the query is related to an entity. In an example, the query is the name of the scientist: "scientist X". The computing system obtains the SERP based on the query and presents the query on the display. The SERP includes a Uniform Resource Locator (URL) of a web page related to a query and paragraph. The paragraph may be from an online encyclopedia, a knowledge graph, or another website. In an example, the paragraph is "scientist X's most important work is considered a study of black holes". In an example, the paragraph is included in a knowledge card shown on the SERP. The knowledge card includes information about the entity, wherein at least a portion of the information included in the knowledge card is from a knowledge graph.

The computing system receives an indication that the user has selected a paragraph shown on the SERP, wherein the indication includes an identifier for the paragraph. In an example, the computing system receives the indication when a mouse cursor of a mouse operated by the user hovers over a paragraph shown on the SERP. Upon receiving the indication, the computing system identifies a plurality of suggested queries related to the paragraph. After identifying the plurality of suggested queries, the computing system presents the plurality of suggested queries in a pop-up graphical element located near the paragraph in the SERP, wherein the pop-up graphical element overlays a portion of the SERP. In an example, the plurality of suggested queries includes "cosmic theory for scientist X", "black hole interpretation", and "black hole in the world".

The computing system may receive a second indication that the user has selected a query of the plurality of suggested queries shown on the SERP. In an example, the computing system receives the second indication when the query is clicked via a cursor of a user-operated mouse. The computing system obtains a second SERP based on a query of the plurality of suggested queries. The computing system presents the second SERP on a display. According to the example given above, the user selects the "black hole interpretation" query, and the second SERP presents an interpretation of the black hole. The second SERP may comprise a knowledge card previously presented in the SERP or a modified version of the knowledge card, such as a knowledge card comprising different paragraphs than those in the (original) knowledge card. Alternatively, the second SERP may not include a knowledge card.

The above-described techniques present various advantages over conventional query suggestion techniques. First, unlike conventional query suggestion techniques, the above techniques combine knowledge-graph mining methods with generative machine learning techniques to present relevant suggested queries to users. Second, while conventional query suggestion techniques present suggested queries to users based on queries that the user has submitted, the techniques described above present suggested queries based on paragraphs shown within the SERP. Thus, the above-described techniques present more relevant queries to the user than conventional techniques. Third, unlike conventional query suggestion techniques that present suggested queries at the top or bottom of a SERP when presenting a SERP, the above techniques may present suggested queries to a user when a mouse cursor of a user-operated mouse hovers over a paragraph shown on the SERP, where the hovering serves as an indicator of user interest in the paragraph. Thus, the above-described techniques present suggested queries in a non-invasive, intuitive, and user-friendly manner. Fourth, by using a generative model that generates suggested queries based on paragraphs and data in the knowledge-graph, the techniques described above may generate related suggested queries without requiring other users to previously submit suggested queries to a search engine. Fifth, because the knowledge-graph includes comprehensive information about many different entities in a single location, the above-described techniques do require extensive searches to be performed on different sources in order to find information to supplement paragraphs (which are used to generate suggested queries). Thus, the above described techniques save network and storage resources.

Referring to FIG. 1, an example computing environment 100 that facilitates providing contextual query-intent (e.g., suggested queries) to an interactive user is illustrated. The computing environment 100 includes a computing system 102. According to an embodiment, computing system 102 is a cloud-based computing platform. According to an embodiment, computing system 102 is a server computing device. The computing system 102 includes a processor 104, a memory 106, and a data store 108.

According to an embodiment, the memory 106 has a server search engine 110 loaded therein. According to other embodiments, the server search engine 110 is included in memory of a second computing system (not shown in FIG. 1). The server search engine 110, when executed by the processor 104 (or another processor), is generally configured to (1) receive a query from a computing device of a user; (2) searching the index based on the query; (3) Generating a Search Engine Results Page (SERP) based on the search of the index; and (4) communicating the SERP to the user's computing device, wherein the SERP is presented on a display of the computing device.

The memory 106 includes a query generator 112. As will be described in greater detail below, the query generator 112, when executed by the processor 104, is configured to identify contextual query-intent and provide it to the interactive user. The query generator 112 includes a candidate generator Machine Learning (ML) model 114 (computer-implemented). As will be described in more detail below, the candidate generator ML model 114 is configured to generate a plurality of candidate suggested queries that are relevant to the paragraph. In an example, the computing system 102 (or another computing system) trains the candidate generator ML model 114 based on the plurality of paragraphs 116 and data from the knowledge-graph 118. The plurality of paragraphs 116 and knowledge-graph 118 may be stored in the data store 108 (or another data store).

According to an embodiment, candidate generator ML model 114 is a generative ML model. According to an embodiment, candidate generator ML model 114 is or includes a transformer model. The decoder of the transformer model takes as input paragraphs and entries for entities in the knowledge-graph 118. The decoder outputs a plurality of candidate suggested queries based on the input.

A plurality of paragraphs 116 are obtained from a plurality of electron sources. According to an embodiment, a plurality of paragraphs 116 are obtained from a web page included in a SERP. According to an embodiment, the plurality of paragraphs 116 are obtained from an online encyclopedia. According to an embodiment, a plurality of paragraphs 116 are obtained from a news article. The entities referenced in the paragraphs may include people, places, things, ideas, and/or events. Paragraphs of the plurality of paragraphs 116 may include a portion of a sentence, a phrase, or several sentences.

Knowledge graph 118 includes nodes and edges connecting the nodes. The nodes included in the knowledge graph 118 represent entities or attributes of entities. Edges included in the knowledge-graph 118 represent relationships between entities or relationships between entities and attributes. The entities represented by the nodes in the knowledge graph 118 may include people, places, things, ideas, and/or events. A first node representing an entity is connected by an edge to a second node representing a second entity or an attribute of the entity, which may be considered an entry for the entity in the knowledge-graph 118. In an example, a first node represents a scientist, a second node represents a book, and edges represent an "author" relationship, that is, the scientist has written the book. In another example, the first node represents a scientist, the second node is a date attribute, and the edges represent a "birth" relationship, that is, the scientist is born on that date. The expected knowledge-graph 118 includes entries for entities referenced in the plurality of paragraphs 116.

The query generator 112 includes a pruning component 120 configured to prune (e.g., remove) queries from the plurality of candidate suggested queries generated by the candidate generator ML model 114 based on relevance criteria. In an example, pruning component 120 is or includes a machine learning model that classifies a plurality of candidate suggested queries as relevant or irrelevant to a paragraph (that references an entity) based on entries for the entities in knowledge-graph 118. In this way, pruning component 120 removes candidate suggested queries generated by candidate generator ML model 114 that are not relevant to a given paragraph.

The query generator 112 includes a ranker component 122 configured to rank the plurality of pruned candidate suggested queries based on ranking criteria. In an example, the ranking criteria is user interaction data 124 stored in the data store 108. The user interaction data 124 includes a query submitted by a user and an indication of a Uniform Resource Locator (URL) presented on a SERP selected by the user, wherein the SERP is generated based on the query.

For a paragraph of the plurality of paragraphs 116, the query generator 112 uses the candidate generator ML model 114 to generate a plurality of candidate suggested queries for the paragraph. The query generator 112 uses the pruning component 120 to remove a subset of the queries from the plurality of candidate suggested queries, thereby generating a plurality of pruned candidate suggested queries. The query generator 112 ranks the plurality of pruned candidate suggested queries using the ranker component 122, thereby generating a ranked plurality of pruned candidate suggested queries. The query generator 112 selects a number of top-ranked queries from the ranked plurality of pruned candidate suggested queries, thereby generating a plurality of suggested queries for the paragraph. In an example, the query generator 112 selects the top 3 queries from the ranked plurality of pruned candidate suggested queries. The query generator 112 stores an identifier for the paragraph and a plurality of suggested queries for the paragraph as part of a paragraph-query pair 126 stored in the data store 108. Although not depicted in FIG. 1, it should be appreciated that the data store 108 can store paragraph-query pairs for many different paragraphs.

The query generator 112 also includes a service component 128. The service component 128 is generally configured to communicate a plurality of suggested queries for a given paragraph to a user's computing device. The service component 128 can cause a plurality of suggested queries to be included in the SERP before the SERP is presented, or the service component 128 can transmit the plurality of suggested queries to a computing device for presentation within the SERP after a paragraph in the SERP is selected.

The computing environment 100 also includes a computing device 130 operated by a user 132. In examples, computing device 130 is a desktop computing device, a laptop computing device, a tablet computing device, a smart phone, or a video game console. Computing device 130 communicates with computing system 102 over network 134.

Computing device 130 includes a processor 136 and a memory 138. Memory 138 includes a client search engine 140. The client search engine 140 may execute within a web browser (not shown in fig. 1) on the computing device 130. When executed by the processor 136, the client search engine 140 is generally configured to (1) transmit a query to the server search engine 110; (2) receiving search results from server search engine 110; and (3) present the search results to user 132. The search results may be in the form of a SERP. The server search engine 110 and the client search engine 140 may be collectively referred to as "search engines". According to an embodiment, the search engine isBing。

The computing device 130 includes an input component 142 that enables the computing device 130 to receive input from the user 132. Input component 142 may include a mouse, keyboard, touch pad, scroll wheel, microphone, camera, video camera, controller, and/or touch screen. Computing device 130 also includes an output component 144 that enables computing device 130 to output information to user 132. The output assembly 144 includes a display 146. The display 146 may be a touch screen display. Client search engine 140 presents SERP 148 on display 146. Output assembly 144 may also include speakers and/or haptic feedback devices (not shown in fig. 1).

Referring now to FIG. 2, an example conduit 200 that facilitates providing contextual query-intent (e.g., suggested query) to an interactive user is illustrated. The steps illustrated in pipeline 200 are performed by query generator 112. At step 202, the query generator 112 obtains paragraphs referencing entities from the plurality of paragraphs 116. At step 204, the query generator 112 uses the paragraph to obtain data about the entity from the knowledge-graph 118. At step 206, the query generator 112 generates a query intent (e.g., a plurality of candidate suggested queries) based on the text of the paragraph and data about the entity from the knowledge-graph 118.

At step 208, the query generator 112 prunes the plurality of candidate suggested queries based on the relevance criteria, thereby generating a plurality of pruned candidate suggested queries. In an example, the query generator 112 removes, by the pruning component 120, a subset of queries from the plurality of candidate suggested queries based on the relevance criteria. For example, pruning component 120 classifies each candidate suggested query of the plurality of candidate suggested queries as either related or unrelated to the passage based on data from knowledge graph 118. At step 210, the query generator 112 ranks the plurality of pruned candidate suggested queries based on ranking criteria by the ranker component 122 to generate a ranked plurality of pruned candidate suggested queries. In an example, the ranking criteria is user interaction data 124. The query generator 112 selects a number of top-ranked queries (e.g., top three) from the ranked plurality of pruned candidate suggested queries to generate a plurality of suggested queries for the paragraph. The query generator 112 saves the identifier for the paragraph and the plurality of suggested queries for the paragraph as part of the paragraph-query pair 126. At step 212, the query generator 112 causes, via the service component 128, a plurality of suggested queries to be presented to the user 132 when the user 132 selects the paragraph.

Referring back to FIG. 1, an example operation of the computing environment 100 will now be described. The client search engine 140 transmits the query to the server search engine 110. The query is related to entities having entries in the knowledge-graph 118. In an example, the entity is a politician. In an example, the client search engine 140 receives a query from the user 132 as a manual input, and the client search engine 140 transmits the query to the server search engine 110. The server search engine 110 performs a search on an index (not shown in FIG. 1) based on the query to obtain search results. Server search engine 110 generates SERP 148 based on the search results. In an example, SERP 148 includes, among other things, a query, a URL of a web page, and a knowledge card. The knowledge card includes data from entries for entities in the knowledge-graph 118. The knowledge card also includes at least one paragraph associated with the entity. Server search engine 110 transmits SERP 148 to client search engine 140, and client search engine 140 then presents SERP 148 on display 146.

Turning now to fig. 3A, an example of a SERP 148 is illustrated. Client search engine 140 presents SERP 148 on display 146 of computing device 130. SERP 148 includes query 302 (received as input from user 132). In an example, the query is the name of the politician having an entry in the knowledge-graph 118. SERP 148 also includes a top-most result 304, where top-most result 304 includes the URL that server search engine 110 has determined to be most relevant to query 302 and a description of that URL. SERP 148 may include news 306 about the entities referenced in query 302. In an example, news 306 includes news stories about legislation that politicians are sponsoring. In addition to the top most result 304, SERP 148 also includes a first result 308 and an Nth result 310, where N is a positive integer greater than 1 (collectively, "multiple results 308-310"). Each result of the plurality of results 308-310 may include a URL of the web page, a title of the web page, and a description of the web page.

SERP 148 includes knowledge card 312. Knowledge cards 312 include information about the entities referenced in query 302, where at least some of the information in knowledge cards 312 is from knowledge-graph 118. The knowledge card 312 may include an entity image 314, wherein the entity image 314 is an image of or related to an entity. In the example where the entity is a politician, the entity image 314 is an image of the politician. The knowledge card 312 includes an entity description 316 that describes the entity. In the example where the entity is a politician, the entity description 316 describes the politician. The entity description 316 may be from the knowledge-graph 118, an online encyclopedia, or another website. Knowledge card 312 may include entity facts 318 about the entity. Entity facts 318 may come from knowledge graph 118, an online encyclopedia, or another website. In examples where the entity is a politician, facts include political parties of the politician, date of birth of the politician, job position of the politician's role, and the like.

The knowledge card 312 includes a first paragraph 320 that is related to the entity referenced in the query 302, wherein the first paragraph 320 is included in the plurality of paragraphs 116. The first paragraph 320 may include a phrase, a single sentence, or several sentences. In an example, the first paragraph 320 is "politician X began his election politics career as a member of the Y party in 1986, which originated from the battle sports. Knowledge card 312 may also include a P-th paragraph 322, wherein P-th paragraph 322 is included in the plurality of paragraphs 116, and wherein P is a positive integer greater than one.

Turning now to FIG. 3B, a cursor 324 selects the first paragraph 320. According to an embodiment, the user 132 moves the mouse to cause the cursor 324 to hover over the first paragraph 320. According to other embodiments, the user 132 clicks the mouse while the cursor 324 is positioned over the first paragraph 320. According to other embodiments in which the display 146 is a touch screen display, the user 132 touches the first paragraph 320 presented on the touch screen display.

When the first paragraph 320 is selected, the client computing device 130 presents the first plurality of suggested queries 326 in pop-up graphical elements located near the first paragraph 320 in the SERP 148, wherein the first plurality of suggested queries 326 are generated by the query generator 112 based on the data in the first paragraph 320 and the knowledge-graph 118 (as described above). The pop-up graphical element overlays a portion of the SERP 148. In an example, the first plurality of suggested queries 326 includes three queries related to politicians referenced in the first paragraph 320. In an example, the three queries include "politician X biography", "politician X political party" and "politician X political view".

According to some embodiments, server search engine 110 provides SERP 148 to query generator 112 prior to transmitting the SERP to client search engine 140. Query generator 112 identifies a first paragraph 320 in SERP 148 via service component 128. Query generator 112 performs a search for paragraph-query pairs 126 stored in data store 108 based on the identifier for first paragraph 320. The search results in search results, wherein the search results include paragraph-query pairs corresponding to the first paragraph 320. The paragraph-query pair includes an identifier for the first paragraph 320 and a first plurality of suggested queries 326. The query generator 112 causes the first plurality of suggested queries 326 to be included in the SERP 148, wherein the SERP 148 includes code that causes the first plurality of suggested queries 326 to be initially hidden from display when the SERP 148 is initially presented on the display 146. Query generator 112 provides SERP 148 (with first plurality of suggested queries 326 and code) to server search engine 110. Server search engine 110 transmits SERP 148 (with first plurality of suggested queries 326 and code) to client search engine 140. When the cursor 324 hovers over the first paragraph 320 shown on the SERP 148, the code causes a first plurality of suggested queries 326 to be displayed. When the cursor 324 is not hovering over the first paragraph 148, the code causes the first plurality of suggested queries 326 to be hidden from display.

According to other embodiments, when the client search engine 140 presents the SERP 148 on the display 146, the first plurality of suggested queries 326 are not initially included in the SERP 148. When the cursor 324 hovers over the first paragraph 320, the client search engine 140 transmits an indication to the query generator 112, wherein the indication includes an identifier for the first paragraph 320. Upon receiving the indication, query generator 112 performs a search for paragraph-query pairs 126 stored in data store 108 based on the identifier for first paragraph 320. The search results in search results, wherein the search results include paragraph-query pairs corresponding to the first paragraph 320. The paragraph-query pair includes an identifier for the first paragraph 320 and a first plurality of suggested queries 326. The query generator 112 transmits the first plurality of suggested queries 326 to the client search engine 140, and then the client search engine 140 presents the first plurality of suggested queries 326 on the display 146 while the cursor 324 hovers over the first paragraph 320.

The client search engine 140 may receive a selection of a query of the first plurality of suggested queries 326 as input from the user 132. In an example, the client search engine 140 receives an indication that the user 132 has clicked on a query in the first plurality of suggested queries 326. Upon receiving the indication, the client search engine 140 transmits the query (among the first plurality of suggested queries 326) to the server search engine 110. The server search engine 110 performs a search (as described above) on the index based on the query (among the first plurality of suggested queries 326). The server search engine 110 generates a second SERP based on the results of the query (in the first plurality of suggested queries 326). The server search engine 110 transmits the second SERP to the client search engine 140, and the client search engine 140 then presents the second SERP on the display 146. When the second SERP is presented on the display 146, the client search engine 140 may remove the SERP 148 from the display 146. Alternatively, the client search engine 140 may present the second SERP in a window presented on the display 146 that is separate from the window presenting the SERP 148. Alternatively, the client search engine 140 may present the second SERP in a separate tab of the web browser that is separate from the tab displaying the SERP 148. The second SERP includes a knowledge card 312 (or another knowledge card).

Turning now to FIG. 3C, cursor 324 selects paragraph P322. When the P-th paragraph 322 is selected, the client computing device 130 presents a second plurality of suggested queries 328 in a second pop-up graphical element located near the second paragraph 322 in the SERP 148, wherein the second plurality of suggested queries 328 are generated by the query generator 112 based on the data in the P-th paragraph 322 and the knowledge-graph 118 (in a manner similar to that described above with respect to the first paragraph 320). The second pop-up graphical element overlays a portion of the SERP 148. As illustrated in fig. 3C, when the cursor 324 no longer selects the first paragraph 320, the client search engine 140 may remove the pop-up graphical element comprising the first plurality of suggested queries 326 from the display 146.

Fig. 4-8 include example screen shots illustrating providing contextual query intent to an interactive user. Fig. 4 illustrates a screenshot 400 of a portion of a SERP for "stefin-hopkin". The screenshot 400 includes a knowledge card and paragraph associated with "still-hall". Fig. 5 illustrates writing "stefin hall" on a mouse cursor hovering over a SERP to study black hole physics. He also written a free-selling book, the most notable of which is the "time brief history: a screen shot 500 of the portion of the SERP in fig. 4 after the paragraph from the large explosion to the black hole (1988) ". Screenshot 500 includes a plurality of suggested queries (generated and identified by query generator 112) related to a paragraph: "Stefan Hold book abstract", "Stefan Hold book List" and "Stefan Hold theory". Notably, none of the plurality of suggested queries is directly answered in the paragraph, but such queries are logical extensions of the paragraph and the user 132 may participate in one or more of them.

Fig. 6 illustrates a screenshot 600 of a portion of a SERP of "bernil mordes". Screenshot 600 includes a knowledge card and paragraph associated with "Bery Mordes". FIG. 7 illustrates a screen shot 700 of a portion of a SERP in FIG. 6 after a mouse cursor hovers over a first paragraph on a portion of a SERP that was written "Mordses in 1971 as a member of the liberty alliance party that originated in the anti-war sport and the people party" and after a mouse cursor hovers over a second paragraph on a portion of a SERP that was written "Mordses described itself as a civilian societies, supports workplace demographics, and praise some elements of the Nordic mode". Screenshot 700 includes a first plurality of suggested queries (generated and identified by query generator 112) related to a first paragraph: "bernile morus biography", "bernile morus political trend", and "bernile morus political view". Screenshot 700 includes a second plurality of suggested queries (generated and identified by query generator 112) related to a second paragraph: "Bery-Morus political views", "Bery-Morus website" and "Nordic societies".

Fig. 8 illustrates a screenshot 800 of a portion of a SERP including "nikkasaban" comprising: the paragraphs related to nikka sarban from the knowledge card, a first plurality of suggested queries (generated and identified by the query generator 112) related to a first paragraph on the portion of the SERP, and a second plurality of suggested queries (generated and identified by the query generator 112) related to a second paragraph on the portion of the SERP.

Although the above techniques have been described in the context of SERPs, other possibilities are contemplated. According to an embodiment, query generator 112 generates a suggested query for a paragraph shown on a web page that is not a SERP, where the paragraph references an entity having a corresponding entry in knowledge-graph 118.

Referring now to fig. 9, an example knowledge-graph 900 is illustrated. Knowledge-graph 900 may be or include knowledge-graph 118, or knowledge-graph 118 may be or include knowledge-graph 900. Knowledge graph 900 includes a first node 902 representing "football coach X". The first node 902 is connected to a second node 904 via a first edge 906, wherein the second node 904 represents "team Y". The first side 906 includes criteria indicating a "coaching" relationship, i.e., a "football coach X" teaching "team Y". The knowledge-graph 900 includes a third node 908 connected to the first node 902 via a second edge 910, wherein the third node 908 represents a date ("09-09-1960"). The second side 910 includes criteria indicating a "birth" relationship, i.e., "football coach X" occurs at "09-09-1960". Knowledge graph 900 includes a fourth node 912 connected to first node 902 via a third side 914, wherein fourth node 912 represents "politician Z". Third side 914 includes criteria indicating a "wedding" relationship, i.e., "politician Z" is wedding with "football coach X".

The query generator 112 may utilize the data in the knowledge-graph 900 along with the paragraphs in order to generate suggested queries for the paragraphs. In an example, query generator 112 extracts data indicating that "football coach X" is married with "politician Z" and provides such data and paragraphs referencing "football coach X" to the transformer model, which then generates (candidate) suggested queries based on the data and paragraphs.

Although the knowledge-graph 900 is described as including four nodes and three edges, it should be understood that the knowledge-graph may include more than four nodes and three edges. Furthermore, it should be understood that each node may be connected to many different nodes. It should be appreciated that the nodes and edges in the knowledge-graph 900 may include metadata that enables access to information about the entities represented in the knowledge-graph 900. Further, it should be appreciated that the query generator 112 may utilize a graph search algorithm to find and extract information from the nodes and edges of the knowledge-graph 900.

FIG. 10 illustrates an example methodology related to providing contextual query-intent to an interactive user. While the method is illustrated and described as a series of acts that are performed in a sequence, it should be understood and appreciated that the method is not limited by the order of the sequences. For example, some acts may occur in a different order than described herein. Further, one action may occur simultaneously with another action. Moreover, in some cases, not all acts may be required to implement a methodology described herein.

Moreover, the acts described herein may be computer-executable instructions that may be implemented by one or more processors and/or stored on one or more computer-readable media or mediums. Computer-executable instructions may include routines, subroutines, programs, threads of execution, and the like. Still further, the results of the actions of the method may be stored in a computer readable medium, displayed on a display device, or the like.

Referring now to FIG. 10, a method 1000 performed by a computing system that facilitates providing contextual query-intent to an interactive user is illustrated. The method 1000 begins at 1002, and at 1004, a computing system receives a first indication that a user has selected a paragraph presented on a display that is shown on a Search Engine Results Page (SERP). At 1006, upon receiving the indication, the computing system identifies a plurality of suggested queries related to the paragraph, wherein the plurality of suggested queries are generated based on the paragraph and an entry for the entity in the knowledge-graph. At 1008, after identifying the plurality of suggested queries, the computing system presents the plurality of suggested queries in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP. At 1010, the computing system receives a second indication that the user has selected a query of the plurality of suggested queries. At 1012, upon receiving the second indication, the computing system presents a second SERP on the display, wherein the second SERP is based on the query. The method 1000 ends at 1014.

Referring now to FIG. 11, there is illustrated a high-level diagram of an example computing device 1100 that can be used in accordance with the systems and methods disclosed herein. For example, the computing device 1100 may be used in a system that facilitates providing contextual query-intent to an interactive user. As another example, computing device 1100 may be used in a system that displays SERPs on a display. The computing device 1100 includes at least one processor 1102 that executes instructions stored in a memory 1104. The instructions may be, for example, instructions for implementing functionality described as being performed by one or more components discussed above or instructions for implementing one or more of the methods described above. The processor 1102 may access a memory 1104 through a system bus 1106. In addition to storing executable instructions, the memory 1104 may also store knowledge maps, paragraphs, user interaction data, paragraph-query pairs, SERPs, and the like.

Computing device 1100 additionally includes a data store 1108 that is accessible by processor 1102 through system bus 1106. The data store 1108 may include executable instructions, knowledge maps, paragraphs, user interaction data, paragraph-query pairs, SERPs, and the like. Computing device 1100 also includes an input interface 1110 that allows external devices to communicate with computing device 1100. For example, input interface 1110 may be used to receive instructions from an external computer device, from a user, and so on. Computing device 1100 also includes an output interface 1112 that connects computing device 1100 with one or more external devices. For example, computing device 1100 may display text, images, etc. through output interface 1112.

It is contemplated that external devices that communicate with computing device 1100 via input interface 1110 and output interface 1112 may be included in an environment that provides essentially any type of user interface with which a user may interact. Examples of user interface types include graphical user interfaces, natural user interfaces, and the like. For example, a graphical user interface may accept input from a user employing input device(s) such as a keyboard, mouse, remote control, etc., and provide output on an output device such as a display. Further, the natural user interface may enable a user to interact with the computing device 1100 in a manner that is unconstrained by input devices such as keyboards, mice, remote controls, and the like. Instead, natural user interfaces may rely on speech recognition, touch and stylus recognition, gesture recognition on and near the screen, air gestures, head and eye tracking, sound and speech, vision, touch, gestures, machine intelligence, and the like.

Additionally, although illustrated as a single system, it should be appreciated that the computing device 1100 may be a distributed system. Thus, for example, several devices may communicate over a network connection and may collectively perform tasks described as being performed by the computing device 1100.

The present disclosure relates to providing contextual query-intent to an interactive user according to at least the following examples:

(A1) In one aspect, some embodiments include a method (e.g., 1000) performed by a processor (e.g., 104) of a computing system (e.g., 102). The method includes receiving (e.g., 1004) a first indication that a user (e.g., 132) has selected a paragraph (e.g., 320) shown on a Search Engine Results Page (SERP) (e.g., 148) presented on a display (e.g., 146). The method further includes, upon receiving the indication, identifying (e.g., 1006) a plurality of suggested queries (e.g., 326) related to the paragraph, wherein the plurality of suggested queries are generated based on the paragraph and the entry for the entity in the knowledge-graph (e.g., 118, 900). The method additionally includes, after identifying the plurality of suggested queries, presenting (e.g., 1008) the plurality of suggested queries in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP. The method also includes receiving (e.g., 1010) a second indication that the user has selected a query of the plurality of suggested queries. The method further includes, upon receiving the second indication, presenting (e.g., 1012) a second SERP on the display, wherein the second SERP is based on the query.

(A2) In some embodiments of the method of A1, the paragraph references the entity.

(A3) In some embodiments of any of the methods of A1-A2, the SERP is removed from the display when the second SERP is presented on the display.

(A4) In some embodiments of any of methods A1-A3, the paragraph is located in a knowledge card (e.g., 312) shown on the SERP, wherein a portion of the information displayed in the knowledge card is from a knowledge graph.

(A5) In some embodiments of any of methods A1-A4, the entity is one of a person, a place, or an event.

(A6) In some embodiments of any of methods A1-A5, the plurality of suggested queries are ranked based on ranking criteria, wherein the ranking criteria include queries submitted by a plurality of users and selection of Uniform Resource Locators (URLs) on the SERP by the plurality of users (e.g., 124), wherein the SERP is based on the queries.

(B1) In another aspect, some embodiments include a computing system (e.g., 102) including a processor (e.g., 104) and a memory (e.g., 106). The memory stores instructions that, when executed by the processor, cause the processor to perform any of the methods described herein (e.g., any of methods A1-A6).

(C1) In yet another aspect, some embodiments include a non-transitory computer-readable storage medium comprising instructions that, when executed by a processor (e.g., 104) of a computing system (e.g., 102), cause the processor to perform any of the methods described herein (e.g., any of A1-A6).

(D1) In another aspect, some embodiments include a method performed by a computing system (e.g., 102) including a processor (e.g., 104) and a memory (e.g., 106). The method includes receiving an indication that a user (e.g., 132) has selected a paragraph (e.g., 320) shown on a Search Engine Results Page (SERP) (e.g., 148) presented on a display (e.g., 146). The paragraph references an entity. The method also includes, upon receiving the indication, identifying a plurality of suggested queries (e.g., 326) related to the paragraph, wherein the plurality of suggested queries are based on the paragraph and an entry (e.g., 118, 900) for the entity in the knowledge-graph. The method additionally includes, after identifying the plurality of suggested queries, presenting the plurality of suggested queries in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP.

(D2) In some embodiments of the method of D1, a paragraph is selected when a mouse cursor (e.g., 324) of a mouse hovers over the paragraph shown on the SERP.

(D3) In some embodiments of any of the methods of D1-D2, the indication includes an identifier for the paragraph and the plurality of suggested queries are identified based on the identifier for the paragraph.

(D4) In some embodiments of any of the methods of D1-D3, the paragraph is located in a knowledge card shown on the SERP. The method also includes receiving a second indication that the user has selected a query of the plurality of suggested queries. The method additionally includes obtaining a second SERP based on the query. The method further includes presenting a second SERP on the display. The second SERP comprises a knowledge card.

(D5) In some embodiments of any of the methods of D1-D4, the method further comprises receiving a query before receiving an indication that the user has selected the paragraph. The method additionally includes obtaining a SERP based on the query. The method further includes presenting the SERP on a display.

(D6) In some embodiments of any of the methods of D1-D5, the method further comprises receiving a second indication that the user has selected a query of the plurality of suggested queries. The method additionally includes obtaining a second SERP based on the query. The method further includes presenting a second SERP on the display.

(D7) In some embodiments of any of the methods of D1-D6, the method further comprises, prior to receiving the indication, generating a plurality of candidate suggested queries based on the paragraph and an entry for the entity in the knowledge-graph. The method additionally includes removing a subset of the queries from the plurality of candidate suggested queries based on the relevance criteria, thereby generating a plurality of pruned candidate suggested queries. The method also includes ranking the plurality of pruned candidate suggested queries based on ranking criteria, thereby generating a ranked plurality of pruned candidate suggested queries. The method also includes selecting a number of top-ranked queries from the ranked plurality of pruned candidate suggested queries, thereby generating the plurality of suggested queries.

(D8) In some embodiments of any of the methods of D1-D7, the method further comprises receiving a second indication that the user has selected a second paragraph (e.g., 322) shown on the SERP after presenting the plurality of suggested queries in the pop-up graphical element. The method additionally includes, upon receiving the second indication, identifying a second plurality of suggested queries (e.g., 328) related to the second paragraph, wherein the second plurality of suggested queries are generated based on the second paragraph and the entry for the entity in the knowledge-graph. The method further includes, after identifying the second plurality of suggested queries, presenting the second plurality of suggested queries in a second pop-up graphical element overlaying a second portion of the SERP, wherein the second pop-up graphical element is located near a second paragraph shown on the SERP.

(D9) In some embodiments of the method of D8, the pop-up graphical element is removed from the SERP when the paragraph is no longer selected.

(D10) In some embodiments of any of the methods of D1-D9, the plurality of suggested queries are included in the SERP, the plurality of suggested queries are initially hidden from display in the SERP, and the plurality of suggested queries are presented in a pop-up graphical element when a cursor (e.g., 324) hovers over a paragraph in the SERP.

(D11) In some embodiments of any of the methods of D1-D10, the plurality of suggested queries are generated by a computer-implemented transformer model having as input the paragraph and an entry for the entity in the knowledge-graph, wherein the plurality of suggested queries are identified based on an output of the computer-implemented transformer model.

(E1) In another aspect, some embodiments include a computing system (e.g., 102) including a processor (e.g., 104) and a memory (e.g., 106). The memory stores instructions that, when executed by the processor, cause the processor to perform any of the methods described herein (e.g., any of D1-D11).

(F1) In yet another aspect, some embodiments include a non-transitory computer-readable storage medium comprising instructions that, when executed by a processor (e.g., 104) of a computing system (e.g., 102), cause the processor to perform any of the methods described herein (e.g., any of D1-D11).

(G1) In another aspect, some embodiments include a method performed by a computing system (e.g., 102) including a processor (e.g., 104). The method includes receiving, from a computing device (e.g., 130) of a user (e.g., 132) in network (e.g., 134) communication with the computing system, an indication that the user has selected a paragraph (e.g., 320) shown on a Search Engine Results Page (SERP) (e.g., 148) presented on a display (e.g., 146) of the computing device. The method also includes, upon receiving the indication, identifying a plurality of suggested queries (e.g., 326) related to the paragraph, wherein the plurality of suggested queries are generated based on the paragraph and an entry (e.g., 118, 900) for the entity in the knowledge-graph. The method additionally includes, upon identifying the plurality of suggested queries, transmitting the plurality of suggested queries to the computing device, wherein the plurality of suggested queries are presented in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP.

(G2) In some embodiments of the method of G1, the method further comprises receiving, from the computing device, a second indication that the user has selected a query of the plurality of suggested queries, wherein the second indication comprises the query. The method additionally includes obtaining a second SERP based on the query. The method also includes transmitting the second SERP to the computing device, wherein the second SERP is presented on the display.

(G3) In some embodiments of any of the methods of G1-G2, the knowledge graph includes nodes (e.g., 902, 904, 908, 912) and edges (e.g., 906, 910, 914) connecting the nodes, wherein the nodes represent entities or attributes of the entities, wherein the edges represent relationships between the entities or relationships between the entities and the attributes.

(H1) In another aspect, some embodiments include a computing system (e.g., 102) including a processor (e.g., 104) and a memory (e.g., 106). The memory stores instructions that, when executed by the processor, cause the processor to perform any of the methods described herein (e.g., any of G1-G3).

(I1) In yet another aspect, some embodiments include a non-transitory computer-readable storage medium comprising instructions that, when executed by a processor (e.g., 104) of a computing system (e.g., 102), cause the processor to perform any of the methods described herein (e.g., any of G1-G3).

The various functions described herein may be implemented in hardware, software, or any combination thereof. If implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. The computer-readable medium includes a computer-readable storage medium. Computer readable storage media can be any available storage media that can be accessed by a computer. Such computer-readable storage media may include Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), compact disk read only memory (CD-ROM) or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. Disk and disc, as used herein, includes Compact Disc (CD), laser disc, optical disc, digital Versatile Disc (DVD), floppy disk and blu-ray disc (BD), where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Furthermore, the propagated signal is not included within the scope of computer-readable storage media. Computer-readable media also includes communication media including any medium that facilitates transfer of a computer program from one place to another. The connection may be a communication medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital Subscriber Line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of communication medium. Combinations of the above should also be included within the scope of computer-readable media.

Alternatively or additionally, the functionality described herein may be performed, at least in part, by one or more hardware logic components. By way of example, and not limitation, illustrative types of hardware logic components that may be used include Field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), system-on-Chip Systems (SOCs), complex Programmable Logic Devices (CPLDs), and the like.

What has been described above includes examples of one or more embodiments. It is, of course, not possible to describe every conceivable modification or variation of the aforementioned devices or methodologies for purposes of describing the aforementioned aspects, but one of ordinary skill in the art may recognize that many further modifications and arrangements of the various aspects are possible. Accordingly, the described aspects are intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims. Furthermore, where the term "comprising" is used in the detailed description or claims, the term is intended to be inclusive in a manner similar to the term "comprising" as "comprising" is interpreted when employed as a transitional word in a claim.

As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless indicated otherwise or clear from the context, the phrase "X employs a or B" is intended to mean any of the natural inclusive permutations. That is, the phrase "X employs a or B" is satisfied by either: x is A; x is B; x employs both A and B. Furthermore, the articles "a" and "an" as used in this disclosure and the appended claims should generally be construed to mean "one or more" unless specified otherwise or clear from context to be directed to a singular form.

Furthermore, as used herein, the terms "component" and "system" are intended to encompass a computer-readable data store configured with computer-executable instructions that, when executed by a processor, cause certain functionality to be performed. Computer-executable instructions may include routines, functions, and the like. It should also be understood that a component or system may be located on a single device or distributed across several devices. Furthermore, as used herein, the term "exemplary" is intended to mean serving as an illustration or example of something, and is not intended to indicate a preference.

Claims (15)

1. A computing system, comprising:

a processor; and

A memory storing instructions that, when executed by the processor, cause the processor to perform actions comprising:

Receiving an indication that a user has selected a paragraph presented on a display that is shown on a Search Engine Results Page (SERP), wherein the paragraph references an entity;

upon receiving the indication, identifying a plurality of suggested queries related to the paragraph, wherein the plurality of suggested queries are generated based on the paragraph and an entry for the entity in a knowledge graph; and

After identifying the plurality of suggested queries, the plurality of suggested queries is presented in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP.

2. The computing system of claim 1, wherein the paragraph is selected when a mouse cursor of a mouse hovers over the paragraph shown on the SERP.

3. The computing system of at least one of claims 1-2, wherein the indication includes an identifier for the paragraph, wherein the plurality of suggested queries are identified based on the identifier for the paragraph.

4. The computing system of at least one of claims 1-3, wherein the paragraph is located in a knowledge card shown on the SERP, the acts further comprising:

receiving a second indication that the user has selected a query of the plurality of suggested queries;

Obtaining a second SERP based on the query; and

A second SERP is presented on the display, wherein the second SERP includes the knowledge card.

5. The computing system of at least one of claims 1-4, the acts further comprising:

Receiving a query prior to receiving the indication that the user has selected the paragraph;

Obtaining the SERP based on the query; and

The SERP is presented on the display.

6. The computing system of at least one of claims 1-5, the acts further comprising:

receiving a second indication that the user has selected a query of the plurality of suggested queries;

Obtaining a second SERP based on the query; and

The second SERP is presented on the display.

7. The computing system of at least one of claims 1-6, the acts further comprising:

Generating a plurality of candidate suggested queries based on the paragraphs and the entries for the entity in the knowledge-graph prior to receiving the indication;

Removing a subset of queries from the plurality of candidate suggested queries based on relevance criteria, thereby generating a plurality of pruned candidate suggested queries;

Ranking the plurality of pruned candidate suggested queries based on ranking criteria, thereby generating a ranked plurality of pruned candidate suggested queries; and

Selecting a number of top-ranked queries from the ranked plurality of pruned candidate suggested queries, thereby generating the plurality of suggested queries.

8. The computing system of at least one of claims 1-7, the acts further comprising:

after presenting the plurality of suggested queries in the pop-up graphical element, receiving a second indication that the user has selected a second paragraph shown on the SERP;

upon receiving the second indication, identifying a second plurality of suggested queries related to the second paragraph, wherein the second plurality of suggested queries are generated based on the second paragraph and the entry for the entity in the knowledge-graph; and

After identifying the second plurality of suggested queries, the second plurality of suggested queries is presented in a second pop-up graphical element overlaying a second portion of the SERP, wherein the second pop-up graphical element is located proximate to the second paragraph shown on the SERP.

9. The computing system of claim 8, wherein the pop-up graphical element is removed from the SERP when the paragraph is no longer selected.

10. The computing system of at least one of claims 1-9, wherein the plurality of suggested queries are included in the SERP, wherein the plurality of suggested queries are initially hidden from display in the SERP, wherein the plurality of suggested queries are presented in the pop-up graphical element when a cursor hovers over the paragraph in the SERP.

11. The computing system of at least one of claims 1-10, wherein the plurality of suggested queries are generated by a computer-implemented transformer model having the paragraphs and the entries for the entity in the knowledge-graph as inputs, wherein the plurality of suggested queries are identified based on an output of the computer-implemented transformer model.

12. A method performed by a processor of a computing system, the method comprising:

Receiving a first indication that a user has selected a paragraph presented on a display that is shown on a Search Engine Results Page (SERP);

Upon receiving the indication, identifying a plurality of suggested queries related to the paragraph, wherein the plurality of suggested queries are generated based on the paragraph and an entry for an entity in a knowledge graph;

after identifying the plurality of suggested queries, presenting the plurality of suggested queries in a pop-up graphical element overlaying a portion of the SERP, wherein the pop-up graphical element is located near the paragraph shown on the SERP;

receiving a second indication that the user has selected a query of the plurality of suggested queries; and

Upon receiving the second indication, a second SERP is presented on the display, wherein the second SERP is based on the query.

13. The method of claim 12, wherein the paragraph references the entity.

14. The method of at least one of claims 12-13, wherein the SERP is removed from the display when the second SERP is presented on the display.

15. The method of at least one of claims 12-14, wherein the paragraph is located in a knowledge card shown on the SERP, wherein a portion of the information displayed in the knowledge card is from the knowledge-graph.