Information processing device and program

Abstract

To enhance the accuracy of timing when a user needs assistance compared with a case where timing for assistance is determined while focusing on words used for searching.SOLUTION: When displaying candidates selected by a user from among a plurality of candidates obtained as search results in sequence, a processor of an information processing device provides the user with information for assisting the search when predetermined conditions are met on a trend related to a plurality of operations by the user performed on each search result.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to an information processing apparatus and a program.

Background Art

[0002] Some application programs (hereinafter also referred to as "software" or "applications") have an assistant function for assisting users' work. The assistant function is a type of user interface, and estimates and provides information for assisting users' operations through display of characters and messages.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] In recent years, from the perspective of technology inheritance, digitization and accumulation of know-how, experience, and other personal information of skilled workers have been promoted by various operators. In addition, various documents handled in daily business are also accumulated in users' terminals or storage on the network. On the other hand, the utilization of accumulated documents in business currently does not meet the expectations of operators and users. For example, users cannot find the required documents, or a lot of trial and error is required to find them. Therefore, it is conceivable to use an assistant function that supports document search by users. However, with current technology, assistance is not provided at the timing required by users.

[0005] The present invention aims to improve the accuracy of determining when a user needs assistance, compared to determining the timing of assistance by focusing on the words used in the search. [Means for solving the problem]

[0006] The invention described in claim 1 has a processor, wherein the processor selects from a plurality of candidates which are the results of a search, the user of The choice hair When displaying the selected candidates in order, Regarding viewing the displayed candidates in order tendency This meets the criteria for providing information that assists in the search. At that time, The aforementioned This is an information processing device that provides users with information to assist in their searches. The invention described in claim 2 is that the processor is the result of a first search Regarding the length of time spent viewing multiple first candidates displayed in order: tendency and, The multiple First Candidate Content of Regarding similarity Trends and the results of a second search performed after the first search. Regarding the length of time spent viewing multiple second candidates displayed in order: Trends and 、 The multiple Second Candidate Content of Regarding similarity Trends and 、 During to The aforementioned offer When the condition is met, The aforementioned Help with your search ru The information processing device according to claim 1, which provides information to a user. The invention described in claim 3 is the same as offer One of the conditions is display done candidate Length of viewing time The information processing apparatus according to claim 2, wherein the tendency for the length to become shorter is detected in both the first search and the second search. The invention described in claim 4 is the same as offer One of the conditions is, The aforementioned First search Trends regarding similarity in and the aforementioned second search The trend regarding similarity in the data is continuous, or the difference in the trend is within a threshold. The information processing device is as described in claim 2 or 3. The invention described in claim 5 is that the processor displays the candidate The words included are cursor Length of time spent Based on this, we estimate the words that are most likely to attract attention. The aforementioned Assist in search ru The information processing apparatus according to claim 1, which provides the word as information. Claim 6 In the invention according to, for each word where the cursor stayed, the The time when the conditions for providing the above-mentioned conditions were met was traced back to the time when the conditions for providing the service were detected. time Then, the value obtained by dividing by the time the cursor was located. Based on, estimate the word with high attention degree, claim 5 The information processing apparatus according to. Claim 7 In the invention according to, among a plurality of candidates which are search results, the user of Receives selection hair A computer that displays the selected candidates in order, Regarding viewing the displayed candidates in order Trend The information provided must meet the conditions for assisting the search. A function of detecting, and the offer Condition Satisfying the conditions When is detected, The aforementioned A program for realizing a function of providing information for assisting search to the user.

Advantages of the Invention

[0007] According to the invention described in claim 1, compared with the case of determining the timing of assistance by paying attention to the words used in the search, the accuracy of the timing when the user needs assistance can be improved. According to the invention described in claim 2, by combining two types of trends, the accuracy of the timing when the user needs assistance can be improved. According to the invention described in claim 3, a situation where assistance is needed can be detected from the user's operation situation. According to the invention described in claim 4, a situation where browsing of documents with similar contents continues can be detected. According to the invention described in claim 5, a search adding the words that the user pays attention to can be realized. Claim 6 According to the invention described in, the words that the user pays attention to at the time of assistance can be provided. Claim 7 According to the invention described in, compared with the case of determining the timing of assistance by paying attention to the words used in the search, the accuracy of the timing when the user needs assistance can be improved. [Brief explanation of the drawing]

[0008] [Figure 1] This diagram illustrates an example configuration of an information processing system assumed in the embodiment. [Figure 2] This diagram illustrates an example of the device's functional configuration. [Figure 3] This flowchart illustrates an example of the process performed by a terminal operated by a user searching for electronic documents. [Figure 4] This diagram illustrates an example of information obtained about viewed pages. (A) shows the information obtained when the first page has been viewed, and (B) shows the information obtained when the second page has been viewed. [Figure 5] This diagram illustrates an example of information obtained about pages viewed during a series of searches. [Figure 6] This diagram illustrates an example of calculating the average page transition and similarity score after viewing up to the third page. [Figure 7] This diagram illustrates an example of calculating the average page transition and similarity score after viewing up to the fourth page. [Figure 8] This diagram illustrates an example of calculating the average page transition and similarity score after viewing up to the eighth page. [Figure 9] This is a diagram illustrating an example of the information saved after viewing up to the eighth page. [Figure 10] This diagram illustrates the conditions under which information to assist in a search is provided. [Figure 11] This diagram illustrates an example of a process for estimating words that a user is paying attention to. (A) shows an example of the information used to estimate highly-regarded words, and (B) shows a list of candidate highly-regarded words. [Figure 12]This diagram illustrates an example of the process for narrowing down new search keywords. (A) shows a list of highly relevant word candidates, (B) shows a list of word candidates sorted based on the calculated score, and (C) shows examples of recommended search keywords. [Figure 13] This diagram illustrates how information to assist the user can be displayed on the screen showing search results. (A) shows an example of the screen used to display search results when it is determined that assistance is not needed, and (B) shows an example of the screen used to display search results when it is determined that assistance is needed. [Figure 14] This diagram illustrates another example of how user-assistant information may be displayed on the screen where search results are shown. [Figure 15] This diagram illustrates another example of how user-assistant information may be displayed on the screen where search results are shown. [Modes for carrying out the invention]

[0009] Embodiments of the present invention will be described below with reference to the drawings. <System Configuration> Figure 1 is a diagram illustrating an example of the configuration of the information processing system 1 assumed in the embodiment. The information processing system 1 shown in Figure 1 consists of a terminal 100, a database (hereinafter also referred to as "DB") 200, and a web server 300. The terminal 100, DB 200, and web server 300 are connected to network N. Network N can be a LAN (Local Area Network) or the Internet. Network N can also be a 4G, 5G, or other mobile communication system. In Figure 1, terminal 100, DB 200, and web server 300 are on the same network N, but network N may consist of multiple networks.

[0010] Terminal 100 is a device used by the user to search for information. Terminal 100 can be, for example, a desktop computer, a laptop computer, a tablet computer, a smartphone, or smart glasses. Smart glasses are a device that displays a virtual image in front of the user's line of sight. Terminal 100 is an example of an information processing device within the scope of the claims. Figure 1 shows one terminal 100, but multiple terminals 100 may be used. Furthermore, while the example in Figure 1 assumes that the electronic documents to be searched are stored in DB200 and web server 300, the electronic documents to be searched may also be stored in terminal 100. In that case, a system configuration consisting only of terminal 100 is also possible.

[0011] The electronic documents handled in this embodiment include file data created by various applications, as well as file data digitized from paper documents and file data output from various electronic devices. Applications here include, for example, word processing software, spreadsheet software, presentation software, drawing software, database software, email software, groupware software, accounting software, CAD (Computer-Aided Design) software, DTP (Desktop Publishing) software, project management software, web design software, image editing software, and audio editing software.

[0012] Furthermore, digitized paper documents can be converted into file data, such as scanned data and faxed data. Furthermore, various types of electronic devices include, for example, cameras, microphones, scanners, fax machines, and medical equipment. The electronic documents handled in this embodiment include, for example, text, still images, videos, audio, and data processed by programs. Electronic documents include not only data generated in daily operations, but also data that electronically records the know-how, experience, and other personal information of skilled workers. Skilled workers are defined as individuals with extensive experience in various occupations or those with high levels of professional skills.

[0013] Terminal 100 includes a processor 111 that controls the operation of the entire device, a ROM (Read Only Memory) 112 that stores the BIOS (Basic Input Output System), a RAM (Random Access Memory) 113 used as the work area of ​​the processor 111, an auxiliary storage device 114, a display device 115, an input receiving device 116 that accepts information input using a mouse or keyboard, and a communication device 117 used for communication with network N. The processor 111 and each device are connected via signal lines such as buses.

[0014] The processor 111, ROM 112, and RAM 113 function as a so-called computer. The processor 111 implements various functions through program execution. For example, the processor 111 performs processes such as providing information to assist the user in their searches. The auxiliary storage device 114 is, for example, a hard disk drive or semiconductor storage. The auxiliary storage device 114 is used to store the OS (Operating System), applications and other programs, retrieved data, etc. The auxiliary storage device 114 may also store the documents that are the target of the search. The communication device 117 is a device that enables communication with other devices connected to the network N. The communication device 117 uses a module compliant with Ethernet®, WiFi®, or any other communication standard.

[0015] Database 200 is, for example, a hard disk drive or semiconductor storage. Database 200 stores the electronic documents to be searched. Alternatively, storage may be used in place of, or in conjunction with, database 200. Figure 1 shows a single database 200, but multiple databases 200 may be used.

[0016] The web server 300 is a server equipped with, for example, a hard disk drive or semiconductor storage. The web server 300 provides various services through a web browser running on the terminal 100. One of these services is a search service. Alternatively, a file server may be used instead of, or in conjunction with, the web server 300. Figure 1 shows a single web server 300, but multiple web servers 300 may be used.

[0017] <Device Functional Configuration> Figure 2 is a diagram illustrating an example of the functional configuration of terminal 100. The functions shown in Figure 2 correspond to the assistant functions that support information retrieval, among the various functions realized through program execution. As mentioned earlier, the timing and content of user assistance provided by the assistant function are crucial. In this embodiment, by improving the accuracy of the timing, efficient information retrieval by the user is achieved.

[0018] The information acquisition unit 121 is a functional unit that acquires various types of information related to the search. One example of information to be acquired is information entered by the user for searching electronic documents. One type of information entered by the user is, for example, a keyword (hereinafter referred to as "search keyword"). In this embodiment, the search keyword is a string of characters. However, images may also be included in the information entered by the user for searching electronic documents. In addition, the information entered by the user may include search criteria. These search criteria may include, for example, conditions defining an electronic document. Examples of conditions defining an electronic document include type, duration, language used for writing, and author.

[0019] Other examples of information obtained include information about electronic documents selected by the user from the search results (hereinafter referred to as "search results" or "candidates"). In other words, it includes information about electronic documents that the user has selectively viewed. If the viewed electronic document is a web page, information about the accessed web page is retrieved. A web page is an electronic document written in HTML (Hypertext Markup Language). The retrieved web page information includes, for example, the page name, the page's URL (Uniform Resource Locator), the text written on the page, embedded links, and other information that can be retrieved. For other types of documents, information appropriate to the document type will be obtained. For example, if the viewed electronic document is an image, the image's attribute data or features extracted from the image may be obtained.

[0020] Other examples of information to be retrieved include information related to browsing candidates. If the viewed electronic document is a web page, the following information can be obtained: the date and time the page was viewed, the length of time the page was viewed (hereinafter referred to as "viewing time"), the speed at which the page was scrolled (hereinafter referred to as "scrolling speed"), the commands entered by the user, the coordinates where the cursor was positioned, the text at the cursor's position, the time the cursor was positioned over the text (hereinafter referred to as "dwell time"), and other information. This information is just one example of information about the user's behavior while browsing.

[0021] The behavioral trend calculation unit 122 is a functional unit that calculates the trend of operations (hereinafter also referred to as "behavioral trend") from information related to the viewing of candidates. The behavioral trend includes, for example, a moving average of the time required to view the presented candidates one after another (hereinafter referred to as "viewing time"). In this embodiment, the moving average is calculated as the average of the viewing times of the three most recent candidates. In other words, the viewing time per candidate is calculated for every three candidates.

[0022] When the currently viewed candidate is far from the electronic document the user is looking for, the moving average of viewing time tends to be small. This is because the user will quickly move on to other candidates. On the other hand, when the candidate being viewed is close to the electronic document the user is looking for, the moving average of viewing time tends to be larger. This is because the user spends more time reviewing the content. However, the number of candidates used to calculate the moving average is not limited to three. For example, the moving average could be calculated using the four most recent candidates as the unit. In addition, a moving average of behavioral trends, such as scroll speed, could also be calculated.

[0023] The page similarity calculation unit 123 is a functional unit that calculates the similarity between two candidate pages viewed by the user, one before the other. The similarity is calculated based on information about the selected candidate. Known techniques are used to calculate the similarity. For example, electronic documents are vectorized, and the cosine similarity of the two vectors is calculated. In this embodiment, we assume that the candidate is a web page, so the calculated similarity is called "page similarity." In this embodiment, page similarity is calculated based on the feature quantities of the viewed web page and the feature quantities of other web pages viewed immediately before it. The feature quantities are represented, for example, as vectors.

[0024] The calculation result storage unit 124 is a functional unit that stores the information acquired by the information acquisition unit 121, the moving average calculated by the behavior trend calculation unit 122, and the similarity calculated by the page similarity calculation unit 123 in an auxiliary storage device 114 or the like. The state determination unit 125 determines whether or not support is needed based on the slope of the moving average calculated by the behavior trend calculation unit 122 and the slope of the similarity calculated by the page similarity calculation unit 123. A state in which support is needed is when the user is "in trouble".

[0025] In this embodiment, the state determination unit 125 determines that support is needed if the following two conditions are met. These two conditions are examples of the "predetermined conditions" in the claims. Furthermore, the simultaneous fulfillment of both conditions means that the predetermined conditions are met. The first condition is that the slope of the moving average calculated for candidate views is "continuously negative" between the two search keywords. The second condition is that the similarity slope is "continuous with a positive slope" between the two search keywords.

[0026] The recommended keyword estimation unit 126 is a functional unit that estimates words that the user is interested in based on information related to browsing candidates. Information regarding the browsing of candidates is provided by the behavioral trend calculation unit 122. The estimation here also takes into account the time elapsed from the time the page was viewed to the time when a state requiring assistance was detected. In this embodiment, the recommended keyword estimation unit 126 also uses the coordinates where the cursor is placed, the string where the cursor is placed, and the time the cursor was on the string (i.e., the time spent) to estimate candidate words that are of high interest to the user.

[0027] Specifically, the recommended keyword estimation unit 126 multiplies the reciprocal of the time elapsed from the time the page containing each candidate was viewed to the time when a state requiring assistance was detected by the cursor's dwell time, thereby identifying the candidate that the user is paying the most attention to. The recommended keyword estimation unit 126 outputs the identified candidates as recommended keywords (hereinafter also referred to as "recommended keywords"). The recommended keyword display unit 127 is a functional unit that displays recommended keywords on the display device 115 (see Figure 1).

[0028] <Processing operation> The following sections use Figures 3 to 15 to explain the processing actions performed in relation to searching for electronic documents. Figure 3 is a flowchart illustrating an example of the process performed by terminal 100 (see Figure 1), which is operated by a user searching for electronic documents. In the diagram, the symbol S represents a step. The process shown in Figure 3 is implemented through the execution of a program by processor 111 (see Figure 1). This program is resident and monitors user searches.

[0029] First, the processor 111 obtains the search keyword (Step 1). The search itself is performed by a search engine (not shown). In this embodiment, the search engine resides, for example, in the database 200 or the web server 300. The search keyword is obtained each time a target for viewing is specified from the list of search results. Next, the processor 111 records the elapsed time since the start of the search (step 2). The search starts, for example, by entering the first search keyword. Note that the time spent viewing each page is measured separately from the elapsed time.

[0030] Next, the processor 111 obtains information about the pages the user viewed from the presented search results and trends in the user's behavior while browsing (Step 3). Figure 4 is a diagram illustrating an example of information obtained about viewed pages. (A) shows the information obtained when the first page has been viewed, and (B) shows the information obtained when the second page has been viewed. Each row corresponds to the viewed page, and each column corresponds to the retrieved information.

[0031] As shown in Figures 4(A) and (B), each time a new page is viewed, the following information is acquired and recorded: "Elapsed time from the start of the search to the completion of viewing (seconds)", "Related search keywords", "Title of the viewed page", "Page viewing time (seconds)", and "Words the cursor was hovering over and time spent on them (seconds)". A search using "document + important words + extract" as search keywords is one example of the first type of search. Furthermore, the operation of selecting a page to view from the search results using "document + important words + extract" as search keywords, the viewing time of the selected page, and the words the cursor was on and the time spent on the selected page are all examples of the first operation.

[0032] The title of the first page viewed was "The Path to Tech Begins with a Single Step, Part 12: Learning the Fundamental Concept of TF-IDF." This page was found as a result of a search using the keywords "document + important words + extraction." The viewing time for the first page was 360 seconds.

[0033] In Figure 4(A), the elapsed time from the start of the search to the completion of browsing is the same as the browsing time, but in reality, there is a discrepancy. Additionally, the words the user hovered over with their cursor while browsing the first page, along with the duration of their dwell time, are recorded. Figure 4(A) shows the three words in descending order of dwell time. For reference, the dwell time for "natural language" was 4 seconds, for "-Idf" it was 2.4 seconds, and for "accuracy" it was 1.5 seconds. The title of the second page viewed was "Introduction to Natural Language Processing, Part 5: Keyphrase Extraction using pke". This page was also the result of a search using the keywords "document + important words + extraction". The viewing time for the second page was 180 seconds. Furthermore, the words the cursor was hovering over and the time spent on them within the second page were "accuracy" for 3 seconds, "pyton" for 1.4 seconds, and "pke" for 1 second.

[0034] Figure 5 illustrates an example of information obtained about pages viewed during a series of searches. Figure 5 shows information obtained from eight pages, including six newly viewed pages. Incidentally, the viewing time for the third page (i.e., the third page viewed) was 300 seconds, the viewing time for the fourth page (i.e., the fourth page viewed) was 255 seconds, and the viewing time for the fifth page (i.e., the fifth page viewed) was 105 seconds. The search keywords are the same for pages from the first to the fifth line.

[0035] The first column, "Elapsed time from start of search to completion of viewing (seconds)," records the sum of the viewing times for each page. Therefore, the elapsed time until the viewing of the second page (i.e., the second page viewed) was completed is 540 seconds (= 360 seconds + 180 seconds). In Figure 5, the search keywords change starting from the sixth page (i.e., the sixth page viewed). Specifically, the search keywords are changed to "document + important words + extraction + dataset".

[0036] A search using the keywords "document + important words + extraction + dataset" is an example of a second type of search. Furthermore, selecting a page to view from search results using the search keywords "document + important words + extraction + dataset," the "viewing time (seconds) of the selected page," and the "word the cursor was hovering over and the time spent on it (seconds)" within the selected page are all examples of the second operation. However, the first column, "Time elapsed from the start of the search to the completion of the browsing," records the elapsed time calculated regardless of the differences in search keywords.

[0037] Let's return to the explanation of Figure 3. Once information about the newly viewed page is obtained, the processor 111 determines whether or not there was a page viewed immediately before using the same search keyword (step 4). If there is a page that was viewed immediately before, the processor 111 obtains a positive result in step 4. For example, in Figure 5, this would be the pages from lines 2 to 5 and lines 7 and 8. On the other hand, if there is no page that was viewed immediately before, the processor 111 obtains a negative result in step 4. For example, in Figure 5, this applies to the page on line 1 and the page on line 6.

[0038] If a negative result is obtained in step 4, processor 111 returns to step 1 and waits to view another page. If a positive result is obtained in step 4, the processor 111 performs the following processes: calculating the trend of page transitions from information on the user's behavior (step 5), and extracting features from the previously viewed page and the currently viewed page to calculate the similarity between pages (step 6). Figure 6 illustrates an example of calculating the average transition time and similarity between pages after viewing up to the third page. The chart in Figure 6 includes additional items: "Moving average viewing time (seconds)", "Similarity to the previous page", and "Activation of support".

[0039] Figure 6 shows an example of calculating the moving average from the first page (i.e., the first page viewed) to the third page (i.e., the third page viewed). In Figure 6, the moving average is 280 (=(360+180+300) / 3). The similarity between the first page viewed and the second page viewed was 0.7, and the similarity between the second page viewed and the third page viewed was 0.8.

[0040] Figure 7 illustrates an example of calculating the transition average and similarity between pages after viewing up to the fourth page. In Figure 7, an example of calculating the moving average from the second page (i.e., the second page viewed) to the fourth page (i.e., the fourth page viewed) is shown. In Figure 7, the moving average is 245 (=(180+300+255) / 3). The similarity between the third and fourth pages viewed is 0.5.

[0041] Figure 8 illustrates an example of calculating the transition average and similarity between pages after viewing up to the eighth page. In Figure 8, the search keyword changes from the sixth page (i.e., the sixth page viewed). Therefore, after calculating 220 (=300+255+105) / 3) as the moving average between the three pages including the fifth page (i.e., the fifth page viewed), the moving average column is left blank for two rows. The moving average from the sixth to the eighth page is 210 (=200+170+260) / 3).

[0042] Note that the similarity between pages is calculated before and after the search keyword changes. Therefore, the similarity between the fifth page viewed and the sixth page viewed is 0.4. In the example in Figure 8, "dataset" is added as a new search keyword, but the similarity with the fifth page viewed has decreased. Incidentally, the similarity between the sixth page viewed and the seventh page viewed was 0.8, and the similarity between the seventh page viewed and the eighth page viewed was also 0.8.

[0043] Let's return to the explanation of Figure 3. After steps 5 and 6 are completed, the processor 111 saves information about the viewed page and the user's behavior during the browsing (step 7). Figure 9 is a diagram illustrating an example of information saved after viewing up to the eighth page. Next, the processor 111 determines whether or not assistance is needed (step 8). The determination in step 8 corresponds to the processing of the state determination unit 125 (see Figure 2). If it is determined that assistance is needed, the processor 111 obtains a positive result in step 8. If it is determined that assistance is not needed, the processor 111 obtains a negative result in step 8. If a negative result is obtained in step 8, the processor 111 returns to step 1 and prepares for the next browsing.

[0044] The need for support indicates that the user is in a difficult situation. In this embodiment, if a high frequency of user transitions between pages is detected, or if the similarity between viewed pages remains high, this is considered a situation where the user's search is not working well, i.e., a situation where the user is having trouble. A high frequency of page transitions suggests that the user is not finding the information they are looking for. Similarly, a continuous viewing of similar pages suggests that the user is not finding new information.

[0045] In this embodiment, a situation in which the user continues to view pages with high transition frequency and similarity is determined to be a situation in which user assistance is needed. Figure 10 illustrates the conditions under which information to assist in a search is provided. Here, using Figure 10, we explain that the conditions for the need for assistance are met when the viewing of the 8th page (i.e., the 8th page viewed) is completed.

[0046] First, let's explain the situation where page transitions are occurring at a high frequency. The moving average of viewing time for the first search keyword gradually decreases from 280 seconds to 245 seconds to 220 seconds. In other words, a negative slope is observed. The moving average of viewing time for the second search keyword is only recorded at 210 seconds. This moving average (i.e., 210 seconds) is smaller than the last moving average of viewing time calculated for the first search keyword (i.e., 220 seconds), and the negative slope continues. This condition satisfies the requirement that the slope of the moving average of viewing time is "continuously negative" between the two search keywords.

[0047] Next, we will explain the situation where users continue to view pages with high similarity even after changing their search keywords. The similarity score for the first search keyword to the previous page viewed changed from 0.7 to 0.8 to 0.5 to 0.6 to 0.4. The average of these scores is 0.65. The similarity between the page viewed for the second search keyword and the previous page changed from 0.8 to 0.8. Incidentally, the first similarity is the similarity between the page viewed for the first search keyword and the last page viewed. The average of these is 0.8.

[0048] Figure 10 shows the average value calculated when two pages have been viewed. However, if the minimum number of pages required to calculate the average is three or more, the average can be calculated after three or more pages have been viewed. This condition satisfies the requirement that the similarity slope is "continuous with a positive slope" between the two search keywords.

[0049] Let's return to the explanation of Figure 3. If a positive result is obtained in step 8, the processor 111 estimates the word the user is focusing on (step 9). The estimated word is provided to the user as helpful information. The words displayed here are estimated based on user activity, such as cursor movement during browsing. Words included in search keywords are excluded from this estimation. Figure 11 illustrates an example of a process for estimating words that a user is paying attention to. (A) shows an example of the information used to estimate highly-regarded words, and (B) shows a list of candidate highly-regarded words.

[0050] In the chart shown in Figure 11(A), each row corresponds to the viewed page, and each column corresponds to the information used for estimation based on its perceived importance. In Figure 11(A), the first column is "Elapsed time from the start of the search to the completion of the browsing," the second column is "Page browsing time (seconds)," the third column is "Time (seconds) from the time the need for assistance was detected to the completion of each individual browsing," the fourth column is "Words the cursor was touching and the time spent on them (seconds)," and the fifth column is "Activation of assistance."

[0051] The information in the third column of Figure 11(A) is the new information. The time when the need for support was detected in the third column is the time when the browsing that resulted in a positive outcome in Step 8 (see Figure 3) was completed. In other words, it is the time when the browsing of the page in which the need for support was detected was completed. In other words, this information represents the time elapsed from the time each page was left until the time the page requiring assistance was completed. Therefore, the earlier a page was viewed, the larger the time value recorded in the third column.

[0052] In this embodiment, the time at which the viewing of the 8th line page, where the need for assistance was detected, was completed was 1830 seconds after the start of the search. Therefore, the time in the third column for the first page is calculated as 1470 seconds (= 1830 seconds - 360 seconds). Similarly, the time in the third column for the second page is calculated as 1290 seconds (= 1830 seconds - 540 seconds). The time shown here is an example of "the time elapsed from the time the cursor left the document containing the word until a predetermined condition is met."

[0053] In the chart shown in Figure 11(B), each row corresponds to the word the cursor was hovering over on each page, and each column corresponds to the information used to evaluate the level of attention each word received. In Figure 11(B), the first column is "word," the second column is the "time spent" at the cursor, the third column is the "time back from the time when the need for assistance was detected," and the fourth column is "time spent / time back." In the chart shown in Figure 11(B), the words recorded in the fourth column of Figure 11(A) are rearranged in order of duration of stay.

[0054] For example, the "time spent" on the "natural language" extracted from the first page is 4 seconds, the "time back from the time the need for assistance was detected" is 1470 seconds, and the "time spent / time back" is 0.002721 (=4 / 1470). The value in the fourth column increases the closer the page is to the time the need for assistance was detected, assuming the time spent on the page is the same. For words on the same page, the value increases the longer the time spent on the page.

[0055] Figure 12 illustrates an example of the process for narrowing down new search keywords. (A) shows a list of highly relevant word candidates, (B) shows a list of word candidates sorted based on the calculated score, and (C) shows examples of recommended search keywords. The chart shown in Figure 12(A) is the same as the chart shown in Figure 11(B). Figure 12(B) is a chart in which words are sorted in descending order based on the numerical values ​​in the fourth column of the chart shown in Figure 12(A). In Figure 12(B), "BERT," which attracted attention when viewing page 7, is at the top of the list. The numerical value corresponding to "BERT" is 2.25. The second-ranked word is "co-occurrence," which attracted attention when viewing page 8. In Figure 12(C), "BERT," which ranks first, is added as the first search keyword, resulting in "document + important words + extraction + BERT" as the new search keyword.

[0056] Let's return to the explanation of Figure 3. Once the estimation of the word the user is interested in is complete, the processor 111 displays recommended keywords that include the estimated word (step 10). Below, Figures 13 to 15 illustrate an example of a screen used to assist the user. Figure 13 illustrates a case where information to assist the user is displayed on the screen showing the search results. (A) shows an example of screen 400 used to display search results when it is determined that assistance is not needed, and (B) shows an example of screen 410 used to display search results when it is determined that assistance is needed.

[0057] The screen 400 shown in Figure 13(A) displays the search results for the first search keyword. Therefore, the first row of the search keyword input field 401 on screen 400 displays "Document Key Word Extraction," and below it is a list of web page titles and URLs that constitute the search results. The screen 410 shown in Figure 13(B) appears when it is determined that assistance is needed while displaying the search results for the second search keyword. Therefore, the input field 411 for the first line of the search keyword on screen 410 displays "document important words extraction dataset".

[0058] Note that the initial screen when the search results for the second search keyword are displayed is the same as in Figure 13(A). However, when it is determined through the user's browsing behavior that the user needs assistance, the screen 410 displayed includes an assistance field 420 inserted in the space between the search keyword input field 411 and the search results. The support column 420 shown in Figure 13(B) includes the explanatory text 421 and recommended keywords 422-424.

[0059] The explanation in 421 includes phrases such as "How about these keywords?", indicating that the content of the support section 420 is a suggestion of new search keywords for the user. In the support section, entry 420, three recommended keywords are shown. The top-ranked recommended keyword (422) and the second-ranked recommended keyword (423) are displayed in a larger font, while the third-ranked recommended keyword (424) is displayed in a smaller font. Furthermore, recommended keywords 422 through 424 are all displayed with hyperlinks.

[0060] The top-ranked recommended keyword, 422, is "document key word extraction BERT," the second-ranked recommended keyword, 423, is "document key word extraction co-occurrence," and the third-ranked recommended keyword, 424, is "document key word extraction summary." In Figure 13(B), the words recommended by the system are shown in bold. In the chart shown in Figure 12(B), the third highest-ranking word is "BERT," but since it has already been presented as the top-ranked recommended keyword (422), "summary," which was ranked fourth, is included as the third-ranked recommended keyword (424) in the chart.

[0061] Each of the recommended keywords 422-424 has a hyperlink embedded in it, so when a user clicks on any of the search keywords, screen 410 switches to a list of the corresponding search results. This allows the user to obtain new search results with just a click. Alternatively, instead of recommending keywords 422-424, it is also possible to display only the recommended words on the screen. In this case, it is desirable that the hyperlink be linked to the search results screen where the recommended words are combined with the first search keyword.

[0062] In addition, the priority of recommended keywords can also be represented by the color of the font. For example, recommended keyword 422, which has the highest priority, could be represented in gold; recommended keyword 423, which has the second highest priority, in silver; and recommended keyword 424, which has the third highest priority, in bronze. Additionally, the priority of recommended keywords may be expressed using numbers or symbols.

[0063] Figure 14 illustrates another example of when user-helpful information is displayed on the screen where search results are shown. In the upper right corner of screen 500 shown in Figure 14, a new tab 501 has been added that displays search results based on recommended keywords. Incidentally, the tab 501 is labeled with the word "Recommended". The additional display of tab 501 is useful for drawing the user's attention.

[0064] Additionally, to make tab 501 stand out, it may be displayed in a different format than other tabs. For example, the font used for displaying the tab may be changed, or the color of the tab may be changed. Alternatively, when adding tab 501, you can force the content displayed on screen 500 to switch to the content of the new tab. Incidentally, the function of displaying tab 501 on screen 500 shown in Figure 14 may also be implemented as a web browser extension (i.e., an add-on).

[0065] Figure 15 illustrates another example of when user-helpful information is displayed on the screen where search results are shown. As shown in Figure 15, screen 600 displays Assistant 601 and Advice 602 if it is determined that the user needs assistance. In Figure 15, Assistant 601 is an image of a robot, but the image of the assistant is arbitrary. Furthermore, the user can pre-set the image of the assistant. In Figure 15, the advice 602 displayed is "Next, we recommend searching with 'Document Key Word Extraction BERT'." The text of advice 602 is arbitrary and, like the support field 420 (see Figure 13), may include multiple search keywords.

[0066] <Summary> In this embodiment, the processor 111 (see Figure 1) of the terminal 100 (see Figure 1) where the user performs a search will display a search support screen 420 (see Figure 13(B)), a tab 501 (see Figure 14), and advice 602 (see Figure 15) on the screen only if the search results desired by the user have not been obtained. Specifically, even after different search keywords are entered, the processor 111 detects that the following two conditions are met and displays a search support screen 420, etc. (1) The trend of the moving average of viewing time shortening continues while viewing the second search result, similar to the trend while viewing the first search result. (2) The high similarity between pages continues while viewing the second search result, just as it was while viewing the first search result.

[0067] By displaying search support screens such as screen 420, provided these conditions are met, the system does not interfere with user operations and makes it easier for users to accept support from the system. Furthermore, it is expected that user dissatisfaction with the display of search-assistance information will be reduced, as the function that provides search-assistance information will not have to be disabled by the user's settings.

[0068] Furthermore, this embodiment employs a method in which the priority of candidate words increases the longer the cursor dwells on them, and the shorter the time elapsed until the need for assistance is detected. Therefore, the search results using the recommended new search keywords are more likely to include electronic documents of interest to the user. As a result, improved user satisfaction is expected. In addition, features that provide information to assist in searching will not be disabled by the user's settings.

[0069] <Other Embodiments> (1) Although embodiments of the present invention have been described above, the technical scope of the present invention is not limited to the embodiments described above. It is clear from the claims that embodiments with various modifications or improvements made to those described above are also included in the technical scope of the present invention.

[0070] (2) In the above-described embodiment, one of the conditions for outputting information to support searching was that the average similarity between viewed pages increases even after the search keyword is changed, that is, the average value has a positive slope, but other conditions may be adopted. For example, it may be required that both the average similarity score for the first search and the average similarity score for the second search exceed a predetermined threshold. Here, the threshold could be, for example, 0.6. In this case as well, the similarity trend in the sixth column of Figure 10 will satisfy the similarity condition.

[0071] (3) In the embodiments described above, the case is assumed to be solely for searching and viewing web pages, but it can also be used to search for electronic documents that match the search keywords from the database 200 (see Figure 1) or auxiliary storage device 114 (see Figure 1).

[0072] (4) In the above-described embodiment, the search support function was described as a function of the terminal 100 operated by the user, but it may also be provided as a function of the processor that constitutes the database 200 or the web server 300. For example, in the case of a thin client system in which the terminal 100 operated by the user is used as an input / output device and the program is executed on the server.

[0073] (5) In the embodiments described above, it was required that the slope of the similarity of the electronic documents being viewed be "continuous with a positive slope" between the two search keywords, but it may also be required that they be equivalent or better. Equivalence here may include the case where the average similarity value corresponding to the second search is smaller than the average similarity value corresponding to the first search, but the difference between the two average values ​​is within a predetermined threshold. A positive slope is an example of equivalent or better.

[0074] (6) The processor in each of the embodiments described above refers to a processor in a broad sense, and includes not only general-purpose processors (e.g., CPUs) but also specialized processors (e.g., GPUs (=Graphical Processing Units), ASICs (=Application Specific Integrated Circuits), FPGAs (=Field Programmable Gate Arrays), programmable logic devices, etc.). Furthermore, the processor operations in each of the embodiments described above may be performed by a single processor alone, or by multiple processors located in physically separate locations working together. Also, the order in which each operation is executed by the processor is not limited to the order described in each of the embodiments described above, but may be changed individually. [Explanation of symbols]

[0075] 1...Information processing system, 100...Terminal, 111...Processor, 121...Information acquisition unit, 122...Behavior trend calculation unit, 123...Page similarity calculation unit, 124...Calculation result storage unit, 125...Status determination unit, 126...Recommended keyword estimation unit, 127...Recommended keyword display unit, 200...Database, 300...Web server

Claims

1. It has a processor, The aforementioned processor, When displaying the user's selection of multiple search results in order, and the browsing trends of the displayed candidates meet the conditions for providing search-supporting information, the search-supporting information is provided to the user. Information processing device.

2. The aforementioned processor, Trends regarding the length of time spent viewing multiple first candidates displayed sequentially from the results of the first search, and trends regarding the similarity of the content of said multiple first candidates, The trend regarding the length of time spent viewing multiple second candidates displayed sequentially among the results of the second search performed after the first search, and the trend regarding the similarity of the content of the multiple second candidates, When the aforementioned conditions for provision are met, information to support the search is provided to the user. The information processing apparatus according to claim 1.

3. One of the aforementioned terms of service is, The tendency for the length of time spent viewing the displayed candidates to decrease is detected in both the first and second searches. The information processing apparatus according to claim 2.

4. One of the aforementioned terms of service is, The similarity trend in the first search and the similarity trend in the second search are continuous, or the difference in the trends is within a threshold. The information processing apparatus according to claim 2 or 3.

5. The aforementioned processor, Based on the length of time the cursor lingered over the displayed candidates, the system estimates the most important words and provides those words as information to support the search. The information processing apparatus according to claim 1.

6. The aforementioned processor, For each word the cursor has stayed in, the word with the highest level of attention is estimated based on a value obtained by dividing the time the cursor stayed in the word by the time preceding the time when the conditions for providing the word were found to be met. The information processing apparatus according to claim 5.

7. In a computer that displays the search results in order based on the user's selection from among multiple candidates, A feature that detects whether the browsing trends of the sequentially displayed candidates meet the criteria for providing information that aids in the search, When it is detected that the aforementioned conditions for provision are met, a function is provided to the user to assist in the search, A program to achieve this.

Images