JP2010086472A - Query candidate providing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the efficiency of input of a user's retrieval keyword by presenting suitable query candidates based on an image size, a reading deviation of retrieval target data, or the like. <P>SOLUTION: A query candidate extracting part 106 extracts a character string of a query candidate from a reading try structure index generated on the basis of the retrieval target data 102. A score calculating part 107 calculates a score of a query candidate by taking into consideration the number of keying times which can be saved by presenting a query candidate, a determined score showing a reduction in the number of keying times, which is predicted during input of the remaining character string after the user's selecting a query candidate, the number of items displayable in one screen, or the like. A query candidate selecting part 109 registers a query candidate based on the score and generates a query candidate dictionary 103. When the user inputs a reading character string of a retrieval keyword by one character at a time through an input part 112, a query candidate displaying part 113 presents a query candidate based on a reading character sting with reference to the query candidate dictionary 103. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a query candidate presentation device that presents search keyword candidates when a few characters are read.

In recent years, various electronic devices have become highly functional, and there is an increasing need to browse and search electronic manuals and the like on the devices. Also, from the viewpoint of environmental problems, there is a great need to digitize conventional paper instructions.
However, devices that do not have a keyboard, such as a car navigation device and an FA display device, have a problem in that it is difficult to input search keywords and it takes time to operate, so that digitized documents on the device cannot be fully utilized. It was.

  Therefore, in order to solve this problem, Patent Document 1 proposes a technique for reducing the labor of input operation. The character string predicting device disclosed in Patent Literature 1 predicts a desired character string to be input by a user based on characters input by the user. The character string predicting apparatus is configured to select a first operation cost necessary for the user to select a candidate character string from the candidate list and a candidate character string to be narrowed down by the user inputting the next character. The input efficiency is improved by comparing the required second operation cost and excluding candidate character strings whose first operation cost is larger than the second operation cost from the candidate list to reduce the presentation of unnecessary candidate character strings. It was.

JP 2008-46775 A

  Since the conventional character string predicting apparatus is configured as described above, since only the candidate character string whose first operation cost is larger than the second operation cost is deleted from the candidate list, there are many candidate character strings that are not deleted. If it exists, there was a problem that it would take time to input.

  In addition, when there are a large number of candidate character strings in the candidate list, the candidate character strings cannot be grouped and input in a hierarchical manner, resulting in an increase in the number of candidates presented to the user. .

  In addition, for devices that have a limited number of candidates that can be displayed on one screen, cost calculation that considers the number of candidates that can be displayed has not been performed. There was a problem that the column could not be presented.

  The present invention has been made to solve the above-described problems, and can provide suitable query candidates according to the screen size, reading bias of data to be searched, and the like. The purpose is to improve the input efficiency.

  A query candidate presentation device according to the present invention includes a query candidate display unit that obtains query candidates starting from a reading character of a search keyword input character by character from the query candidate dictionary and presents the query candidates as candidates for the search keyword. A candidate presentation device, a query candidate extraction unit for extracting a character string as a query candidate from data obtained by hierarchically structuring a word extracted from search target data based on a reading character of the word, a search keyword For each reading character of the word extracted from the search target data, a score calculation unit that calculates a score indicating the effect of reducing input labor of the search keyword obtained by setting a character string starting with the reading character as a query candidate, Query candidate selection that registers character strings as query candidates in the query candidate dictionary based on the score calculated by the score calculator It is obtained so as to include a part.

  According to the present invention, it is possible to reduce the labor for inputting the search keyword obtained by extracting a character string as a query candidate from the hierarchically structured data and using the character string starting with the reading character of the search keyword as a query candidate. Since a score indicating the effect is calculated and a character string is registered in the query candidate dictionary as a query candidate based on the score for each reading character of the word extracted from the search target data, the screen size, the search target data Appropriate query candidates according to reading bias and the like can be presented, and the user's search keyword input efficiency can be improved.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of a query candidate presentation device according to Embodiment 1 of the present invention. The query candidate presentation device according to the present embodiment includes a query candidate dictionary generation unit 101 that generates a query candidate dictionary 103 based on search target data 102, and a search that starts with a reading when a reading is input character by character by the user. The query candidate presenting unit 110 selects and presents a candidate for a keyword, that is, a query candidate from the query candidate dictionary 103.

First, the query candidate dictionary generation unit 101 will be described. The query candidate dictionary generation unit 101 analyzes the search target data 102 and extracts an important phrase that should be a target of the query candidate, and generates a reading trie structure index that is structured by reading the extracted important phrase. A trie structure index generation unit 105, a query candidate extraction unit 106 that extracts a character string to be a query candidate from the reading trie structure index, a score calculation unit 107 that calculates a score of the extracted query candidate, and an optimal query candidate based on the score A query candidate selection unit 109 that selects a combination and generates query candidate dictionary data is provided.
The query candidate presentation device uses data such as facility names and instructions as the search target data 102.

The important word / phrase extraction unit 104 analyzes the search target data 102 and extracts important words / phrases to be used as query candidates by using screen layout information such as font size and document structure information such as chapter structure. This process can be realized by using a known technique such as “A summary sentence creation method and apparatus and a storage medium storing a summary sentence creation program” in Japanese Patent Application Laid-Open No. 2001-52032, and detailed description thereof is omitted. To do.
FIG. 2 is an explanatory diagram illustrating an example of important phrases extracted as query candidates by the important phrase extraction unit 104. The important word / phrase extraction unit 104 extracts important words / phrases as query candidates from the search target data 102 as headlines, information on word breaks obtained by dividing the headings into words, and reading breaks obtained by dividing the reading of the headings into words. Is generated. Then, the important phrase extraction unit 104 includes at least these three types of information in the important phrase data and outputs them to the trie structure index generation unit 105. The important phrase data may include information indicating the importance of the phrase in addition to these three types of information.

The trie structure index generation unit 105 receives as input important word / phrase data having heading, word breaker and reading breaker information, and generates a read trie structure index in which the readings of each important word / phrase are arranged as a tree structure.
FIG. 3 is an explanatory diagram showing the reading trie structure index generated by the trie structure index generation unit 105. FIG. 3 shows a reading trie structure index generated using the 14 key words shown in FIG. TRIE structure readings, and the node N _x readings for each keyword, a data hierarchically structures put together as a tree structure the intersection of the top match of the read in the base root node N _r, natural language processing This is a well-known data structure frequently used in dictionary search processing and the like.
As a high-speed implementation method of the same data structure, a double array ("Key deletion efficiency improvement method in a double array" Information Processing Society of Japan, Natural Language Processing Study Group, Vol. 2003 No. 23 2002-NL-154) is well known. The trie structure index generation unit 105 may generate the read trie structure index data by a double array.

The query candidate extraction unit 106 receives the reading trie structure index data as input, and extracts query candidates to be registered in the query candidate dictionary 103 from the reading trie structure index for each node according to the query candidate selection condition 108. Specifically, for each node N _x of the reading trie structure index shown in FIG. 3, all the nodes that are the end of the word break among the descendant nodes N _y of the node N _x are extracted from the root node N _r. the phrase connecting the readings for each node to reach the descendant nodes N _y, and query candidates. However, since the reading trie structure index of FIG. 3 is simplified for the sake of simplicity of explanation, the node that is the end of the word break is not distinguished.

When the search target data 102 is large-scale data, the query candidate selection condition 108 is “descendant node N _y is limited to W words from node N _x ”, “descendant node N _y is L characters from node N _x By specifying conditions such as “limited within” and “descendant node N _y limit the number of branches on the route from node N _x to B”, query candidate extraction unit 106 can extract query candidates. The search range can be limited.

Here, a shortcut between nodes will be described. In query candidate dictionary generator 101, read for each node in the trie structure index, node N _x from the shortcut node N _i is a shortcut destination node descendants in the node N _y of the node N _x, a shortcut between nodes Generate a link to
FIG. 4 is an explanatory diagram showing a shortcut for the reading trie structure index. For example, when the search keyword reading “Tokyo Kyoko” is input, that is, when the node of interest has transitioned from the root node N _r to the node N 6 (hereinafter referred to as readings N _{r to} N 6), details will be described later. The query candidate presentation device presents “Tokyo International Airport First”, “Tokyo International Airport Second”, and “Tokyo International Airport Third Parking” to the user as query candidates. For example, it can be said that the query candidate “Tokyo International Airport First” is a shortcut between nodes from the node N6 corresponding to “ko” of the reading “Tokyo Kyoko” input by the user.
When the user selects “Tokyo International Airport No. 1” from these query candidates, the query candidate presentation device changes the node of interest from N6 to the shortcut destination node N17 (hereinafter referred to as shortcut N6 → N17). ), The readings corresponding to the nodes N7, N8,..., N16 in the shortcut path are automatically displayed to reduce the user's input effort.

The score calculation unit 107 receives the query candidate data as an input, and uses the query candidate selection condition 108 as a score that formulates the effect of reducing input labor obtained by presenting the query candidate corresponding to the reading during the search keyword input. calculate. Then, the score calculation unit 107 outputs score data related to the query candidate to the query candidate selection unit 109.
As the query candidate selection condition 108 input to the score calculation unit 107, there is a maximum display candidate number D _max indicating the number of query candidates that can be displayed per screen.

FIG. 5 is a block diagram illustrating a detailed configuration of the score calculation unit 107. The score calculation unit 107 includes a keystroke number shortening score calculation unit 11, a post-confirmation score calculation unit 12, and a score totaling unit 13. FIG. 6 is an explanatory diagram showing the definition of the score used by the score calculation unit 107.
Keying the number of speed score calculation unit 11 shown in FIG. 5, phrases connecting the readings corresponding to the query candidate corresponding to the reading N _r to N _x ( "reading N _r to N _x" + "shortcut N _x → N _i" ) To calculate the keystroke number shortening score C ₁ (N _x , N _i ). The keystroke shortening score is input by inputting the same query candidate with the number of keystrokes that is less than the number of keystrokes compared to entering the query candidate reading by selecting the query candidate presented in the middle of reading the keyword for search. Show if you can.

Figure 7 is an explanatory diagram showing a score calculation example of the score calculation unit 107, indicating the score when "Tokyo tori" (read N _r ~N66) is input as reading. In FIG. 7, the maximum display candidate number D _max of the query candidate selection condition 108 is defined as 3.
Here, the keystroke number shortening score C ₁ (N66, N _i ) shown in FIG. 7 will be described with reference to FIGS. In the case of the query candidate “Tokyo Metropolitan Daiichi” for the shortcut N66 → N71 corresponding to the readings N _r to N66, among the nodes N _{j in} the middle of the route N66 → N71, the number of leaves (“Chugaku” and “Kokou” ”) Is equal to or less than D _max and the node N _z having the minimum Depth (N _j ) is N70. Therefore, the key stroke reduction score C ₁ (N66, N71) of “Tokyo Metropolitan Daiichi” is “6” from the calculation formula when N _z exists.
On the other hand, in the case of the query candidate “Tokyo Metropolitan Second” of the shortcut N66 → N81 corresponding to the readings N _r to N66, there is no node N _z whose number of leaves is equal to or less than D _{max in the} middle of the route N66 → N81. Therefore, the keystroke reduction score C ₁ (N66, N81) of “Tokyo Metropolitan Second” is “6” from the calculation formula when N _z does not exist.
The keystroke number shortening score calculator 11 calculates the keystroke number shortening score for the shortcut query candidates “Tokyo Metropolitan Third” and “Tokyo Metropolitan Sangyo” corresponding to the node N66 in the same manner as described above.

The after-confirmation score calculation unit 12 shown in FIG. 5 selects a query candidate of shortcut N _x → N _i corresponding to the readings N _{r to} N _x by the user and determines that it is a part of the search keyword. Then, the key hit reduction effect of the query candidate corresponding to the next shortcut N _i → N _ii following this query candidate is predicted and set as a post-determined score C ₂ (N _i ). Here, N _ii is the shortcut destination node when the base shortcut node N _i.

FIG. 8 is an explanatory diagram illustrating an example of a score calculation after determination by the score calculation unit 12 after determination, and illustrates a case where “Tokyo International Airport” is selected as a query candidate. Assuming that the shortcut node N13 of the query candidate “Tokyo International Airport” is a new base node, the next shortcut nodes N _ii include N17, N31, N54, and N64, for example. In this case, the combination of the following shortcut _{N i → N ii, N13 →} N17, N13 → N41, N13 → N54, N13 → is the N64. The post-confirmation score calculation unit 12 calculates a score S _c (N _i , {N _y }) of these shortcut combinations. Here, {N _y } represents the next set of shortcut nodes N _ii , and the number of elements is equal to or less than the maximum display candidate number D _max per screen defined by the query candidate selection condition 108. The score _Sc will be described later.
When following shortcut node N _ii for N _i are present or D _max, also there are many combinations of shortcut N _i → N _ii. In such a case, the post-confirmation score calculation unit 12 calculates a score S _c (N _i , {N _y }) obtained by combining the query candidates of the large number of shortcuts N _i → N _ii by D _max or less, Among them, the maximum one (Max (S _c (N _i , {N _y })) is set as the post-established score C ₂ (N _i ) of the shortcut node N _i .

In the example of score calculation after determination in FIG. 8, two types of patterns 8A and 8B are shown as representative examples of how to select {N _y }. Pattern 8A shows a combination of shortcuts N13 → N17, N13 → N41, N13 → N54, and pattern 8B shows a case where the combination is N13 → N17, N13 → N54, and N13 → N64. The post-confirmation score calculation unit 12 compares the score S _c (N13, {N _y }) of each pattern, and corresponds the score of the pattern 8A that is the maximum to the query candidate “Tokyo International Airport” (node N13). The score after determination is C ₂ (N13).

In the score calculation example of FIG. 7, when the next shortcut node N _ii is considered for a query candidate such as “Tokyo Metropolitan Daiichi”, the number of query candidates is less than the maximum display candidate number D _max in any shortcut node N _ii. It becomes. According to the definition shown in FIG. 6, the post-confirmation score calculation unit 12 is given the post-confirmation score C ₂ (N71) as a score S _c (N71, {N _y }). In this case, the node N _z is N71. Therefore, the keystroke number shortening score C ₁ (N71, N _ii ) is all 0. Therefore, C ₁ (N71, N _ii ) for the next shortcut node N _ii of D _max or less is added, and the score C ₂ (N71) after confirmation of the query candidate “Tokyo Metropolitan Daiichi” is also 0. Become.

The score totaling unit 13 shown in FIG. 5 uses the keystroke number shortening score calculated by the keystroke number shortening score calculation unit 11 and the post-confirmation score calculated by the post-confirmation score calculation unit 12 to use the shortcut score S _e (N _x , N _y ) is calculated, and a score S _c (N _x , {N _y }) corresponding to each query candidate combination is calculated using the shortcut score. {N _y } represents a set of shortcut nodes N _i , and the number of elements is equal to or less than the maximum display candidate number D _max .

The score calculation example of FIG. 7, showing two kinds of patterns 7A and 7B as a typical example of selection of {N _y}. Pattern 7A shows the case where the combinations of query candidates displayed per screen are “Tokyo Metropolitan First”, “Tokyo Metropolitan Second” and “Tokyo Metropolitan Third” for readings N _{r to} N66. Pattern 7B shows the case where the combination is “Tokyo Metropolitan First”, “Tokyo Metropolitan Second” and “Tokyo Metropolitan Sangyo”. The score totaling unit 13 can reduce the length of reading (Depth (N _z ) −Depth (N _x ) and Depth (N _i ) that can be shortened by the shortcut if the number of leaves L _z that can be input efficiently by the shortcut is the same. A high score is given to the pattern 7A having a higher Depth (N _x )).
In the prior art, the priority order between the query candidates could not be determined according to the length of reading that can be shortened by presenting the query candidates. Therefore, these patterns 7A and 7B cannot be superior or inferior, and the number of display candidates is appropriately set. It was not possible to narrow down to three kinds. In contrast, in the present embodiment, query candidates can be prioritized according to the maximum display candidate number _Dmax .
The superiority or inferiority between query candidates in the same pattern may be determined according to the shortcut score, and a query candidate with a high shortcut score is given priority.

FIG. 9 is an explanatory diagram showing a score calculation example of the score calculation unit 107, and shows a score when “Tokyo Kyoko” (reading N _{r to} N6) is input as a reading. Also in FIG. 9, the maximum display candidate number D _max of the query candidate selection condition 108 is defined as 3. Pattern 9A shows the case where the combination of query candidates is “Tokyo International Airport First”, “Tokyo International Airport Second”, and “Tokyo International Airport Third Parking” for readings N _{r to} N6, Pattern 9B shows a case where the combination is “Tokyo International Airport First”, “Tokyo International Airport Third Parking” and “Tokyo International Airport Fourth Parking”. In this example, if the reading lengths (Depth (N _z ) −Depth (N _x ) and Depth (N _i ) −Depth (N _x )) that can be shortened by the shortcut are the same, the number L _z of giving a high score to the higher pattern 9A.
Since the prior art did not calculate the score in consideration of the number of leaves, the patterns 9A and 9B could not be superior or inferior, and the number of display candidates could not be narrowed down to three appropriate types. In contrast, in the present embodiment, it is possible to prioritize query candidates including more leaves.

FIG. 10 shows an example in which the post-confirmation score contributes to the score. FIG. 10 is an explanatory diagram showing a score calculation example of the score calculation unit 107, and shows a score when “Tokyo” (reading N _{r to} N5) is inputted as a reading. In the pattern 10A, although the number of query candidates is less than the maximum display candidate number _Dmax , the scores of the pattern 10A and the pattern 10B have the same value. This is because the score totaling unit 13 calculates the score in consideration of the score after determination, thereby reflecting the score reduction effect predicted after the query candidate is selected by the user in the score.
Here, the post-confirmation score C ₂ (N13) of the query candidate “Tokyo International Airport” in the patterns 10A and 10B is the post-confirmation score calculated by the post-confirmation score calculation unit 12 in FIG. This post-determination score corresponds to the score S _c (N13, {N17, N31, N54}) of the pattern 8A.

Further, according to the score calculation of the score calculation unit 107 as described above, the portion of the reading trie structure index in which the tree structure is biased also corresponds to the size of the screen defined by the search target data 102 and the maximum display candidate number Dma. Appropriate query candidate selection becomes possible. For example, FIG. 11 shows a score calculation example of the score calculation unit 107 when the maximum display candidate number D _max is defined as five. Here, the respective scores of the patterns 11A and 11B are shown as typical examples of combinations of {N _y } when “Tokyo” (reading N _{r to} N66) is input as the reading. In this example, the score totaling unit 13 gives a high score in preference to the pattern 11B that can cover a wider range of leaves.

The query candidate selection unit 109 in FIG. 1 receives the maximum display candidate number D _max of the query candidate selection condition 108 and the score data, and a query having the maximum score based on the score S _c (N _x , {N _y }). Shortcut information is generated by selecting a combination of candidates. The query candidate selection unit 109 adds the generated shortcut information to the query candidate dictionary data by combining the important phrase data extracted by the important phrase extraction unit 104 and the reading trie structure index data generated by the trie structure index generation unit 105. Is registered in the query candidate dictionary 103.

  FIG. 12 is a block diagram showing a configuration of the query candidate dictionary 103 of the query candidate presentation device. The query candidate dictionary 103 includes a heading list 21, a reading trie structure index 22, and shortcut information 23. FIG. 13 shows an example of the heading list 21 registered in the query candidate dictionary 103, and FIG. 14 shows an example of the shortcut information 23 registered in the query candidate dictionary 103. An example of the reading trie structure index 22 is shown in FIGS.

13 includes a record number for reference from other data such as the reading trie structure index 22 and the shortcut information 23 and a headline presented to the user as a query candidate.
Shortcut information 23a, 23b, 23c in FIG. 14 represent the shortcut information for the node N _x in the list information. The elements of the list are composed of (A) to (D), and “(A) Pointer to the next element of the list”, “(A) Shortcut node N _i ”, “(C) Record number referring to the heading list”, respectively. “, (D) Shortcut score S _e (N _x , N _y )”.

The query candidate selection unit 109 selects a combination of query candidates that gives the maximum score from all combinations of query candidates for the node N _x calculated by the score calculation unit 107, and displays information on shortcuts of these query candidates as shown in FIG. It is linked to the node N _x as shown in. FIG. 14 shows shortcut information linked to the node N6. This shortcut information corresponds to a combination of the pattern 9A in FIG.

FIG. 15 is a flowchart showing the operation of the query candidate dictionary generation unit 101 of the query candidate presentation device according to Embodiment 1. Hereinafter, the operation of the query candidate dictionary generation unit 101 will be described with reference to FIGS.
Step ST1 is an important word / phrase extraction process, in which the important word / phrase extraction unit 104 analyzes the search target data 102, extracts important words / phrases using screen layout information and document structure information, and performs word division and reading. Execute. Then, the important phrase extraction unit 104 outputs the important phrase data shown in FIG. 2 to the trie structure index generation unit 105.

Step ST2 is a trie structure index generation process, and the trie structure index generation unit 105 generates the reading trie structure index shown in FIG. 3 using the information of the key words extracted in step ST1. Then, the trie structure index generation unit 105 outputs the read trie structure index data to the query candidate extraction unit 106.
Step ST3 is a query candidate extraction process, and the query candidate extraction unit 106 extracts all query candidates to be registered in the query candidate dictionary 103 using the query candidate selection condition 108 and the reading trie structure index generated in step ST2. To do. Then, the query candidate extraction unit 106 outputs the generated query candidate data to the score calculation unit 107.

  Step ST4 is a score calculation process, and the score calculation unit 107 calculates a score for each query candidate extracted in step ST3. For each query candidate, the keystroke number shortening score calculation unit 11 calculates the keystroke number reduction score, and the post-confirmation score calculation unit 12 calculates the post-confirmation score. Then, the score totaling unit 13 calculates a score for each combination of query candidates using the keystroke number shortening score and the finalized score, and outputs the score data to the query candidate selecting unit 109.

Step ST5 is a query candidate selection process. Using the score data calculated in step ST4, the query candidate selection unit 109 selects a combination of query candidates having the maximum score, and the headings shown in FIGS. Query candidate dictionary data including the list 21, reading trie structure index 22 and shortcut information 23 is generated and registered in the query candidate dictionary 103.
The same score calculation is performed for all the nodes, and when a combination of query candidates having the maximum score is registered for each node (step ST6 “Yes”), the query candidate dictionary generation process ends. While there is a node for which a combination of query candidates is not determined, the process returns to the query candidate extraction process in step ST3 (step ST6 “No”), and the query candidate extraction unit 106 selects a query corresponding to the next node. Extract candidates.

Next, the query candidate presentation unit 110 will be described. The query candidate presentation unit 110 illustrated in FIG. 1 includes a query candidate dictionary 103 generated by the query candidate dictionary generation unit 101, an input unit 112 that receives a search keyword reading character string 111 input by a user, and a search keyword reading character string. A query candidate display unit 113 that displays the next query candidate with reference to the query candidate dictionary 103 based on 111 is provided.
The search keyword reading character string 111 is a search keyword reading character string input by the user or a user selection result for the query candidate displayed on the query candidate display unit 113. The input is performed by a software keyboard, a numeric keypad, etc. displayed on the touch panel, and the input unit 112 and the query candidate display unit 113 are realized by this touch panel. The maximum display candidate number D _max of the query candidate selection condition 108 described above is defined according to the size of the display screen of the touch panel that realizes the query candidate display unit 113.

The input unit 112 accepts input of a search keyword reading character string 111 and a query candidate selection result 111 a from the user, and outputs the information to the query candidate display unit 113.
When the query candidate display unit 113 receives the information of the reading character string 111 of the search keyword and the query candidate selection result 111a, the query candidate display unit 113 refers to the query candidate dictionary 103 and changes the target node of the reading trie structure index. Then, shortcut information for the transition destination node is acquired from the query candidate dictionary 103, and a list of query candidates is created and presented.

Specifically, in the initial state in which no reading character string 111 of the search keyword is input, the query candidate display unit 113 sets the root node _Nr as the target node in the reading trie structure index 22 of the query candidate dictionary 103. To do. When a reading is input, the query candidate display unit 113 reads the input reading one character at a time, and changes the attention node to a node that matches the input reading among the descendant nodes of the attention node. Let it be a new attention node.
On the other hand, when the query candidate selection result 111a indicating the selection result selected by the user from the plurality of presented query candidates is input, the query candidate display unit 113 displays “(A)” of the shortcut information illustrated in FIG. With reference to the shortcut node N _i , the transition is made to the corresponding node number.
Subsequently, the query candidate display unit 113 refers to “(c) a record number referring to a heading list” of each shortcut information based on all shortcut information linked to the node number of the transition destination. The heading of the record number is acquired from the heading list shown in FIG. 13, and this heading is presented to the user as a query candidate. At this time, the query candidate display unit 113 controls the display order so that the display order of query candidates having a high “(D) shortcut score S _e (N _x , N _y )” in the shortcut information is high.

Next, the operation of the query candidate presentation unit 110 will be described. FIG. 16 is a flowchart showing the operation of the query candidate presentation unit 110 of the query candidate presentation device according to Embodiment 1 of the present invention.
Step ST11 is a character input process or a query candidate selection process. Immediately after the start of the query candidate presentation process, that is, in step ST11 in the initial state, the input unit 112 receives the reading character string 111 of the search keyword input from the user.

Step ST12 is a state transition process in the reading tri structure index, and the query candidate display unit 113 updates the attention node of the reading tri structure index shown in FIG. 3 according to the information of the reading character string 111 of the search keyword received in step ST11. To do. For example, when “Tokyo Kyoko” of readings N _{r to} N 6 is input, the query candidate display unit 113 changes the target node from the root node N _r to N 6.

Step ST13 is an end determination process. If there is no transition destination node that matches the input reading among the descendant nodes of the target node in step ST12, there is no query candidate corresponding to the input input by the user. Therefore, the query candidate display unit 113 ends the query candidate presentation process (step ST13 “No”).
If there is a transition destination node, the query candidate display unit 113 advances the processing to the next (step ST13 “Yes”). When the reading “Tokyo Kyoko” is input, since the transition destination node N6 exists, the process proceeds to the next process.

Step ST14 is a query candidate acquisition display process, in which the query candidate display unit 113 refers to shortcut information corresponding to the transition destination node and acquires a list of query candidates.
When the reading “Tokyo Kyoko” is input, the shortcut information corresponding to the transition target node N6 is as shown in FIG. The query candidate display unit 113 first refers to the shortcut information 23a (shortcut N6 → N17), shortcut information 23b (shortcut N6 → N31) and shortcut information 23c (shortcut N6 → N54) corresponding to N6, and the record number of the heading list. Obtain R3, R6 and R9. Next, the query candidate display unit 113 acquires query candidates “Tokyo International Airport First”, “Tokyo International Airport Second”, and “Tokyo International Airport Third Parking Lot” with the same record number from the heading list 21 of FIG. Displayed on the screen.

FIG. 17 is an explanatory diagram illustrating an example of a query candidate presentation screen that the query candidate display unit 113 of the query candidate presentation device displays in response to a user input. As shown in FIG. 17A, the query candidates acquired by the query candidate display unit 113 are displayed as query candidate character strings on the screen.
Then, the process returns to step ST11 again.

When “to”, “u”, “ki”, “yo”, “u”, “ko” are input as the search keyword reading character string 111, the query candidate presentation unit 110 performs steps ST11 to ST13. The process is repeated and the node of interest transits from the root node _Nr to N1, N2, N3, N4, N5, and N6. At this time, the display screen of the query candidate display unit 113 is in the state shown in FIG. Here, when the process returns to step ST11 and the user inputs to the input unit 112 the query candidate “Tokyo International Airport 2” as shown in FIG. 17B, the input unit 112 inputs the input. The query candidate selection process is received that accepts as the query candidate selection result 111a.

In step ST12, the query candidate display unit 113 changes the node of interest to the shortcut node N31 according to the shortcut information for the query candidate “Tokyo International Airport Second”. At the same time, the query candidate display unit 113 displays the screen supplementing the reading character string “Kusakuu Kodani” in the shortcut path (N6 → N31) of the selected query candidate as a reading input (FIG. 17C).
Since there is a transition destination node, the process proceeds (step ST13 “Yes”), and in subsequent step ST14, the query candidate display unit 113 displays the next query candidate corresponding to the shortcut node N31 of the query candidate “Tokyo International Airport Second”. Acquire “Tokyo International Airport Terminal 2” and “Tokyo International Airport Second Parking” and display them on the screen.

  When the process returns to step ST11 again and either one of the query candidates “Tokyo International Airport Second Terminal” or “Tokyo International Airport Second Parking” is selected by the user, in the end determination of step ST13, the transition destination Since it is confirmed that there is no node, the query candidate presentation process ends here.

FIG. 18 is a screen example of query candidate presentation when the maximum display candidate number D _max is set to 5, and shows a case where “Toyotori” of readings N _{r to} N 66 is input. The query candidate presentation unit 110 performs the query candidate presentation process using the query candidate dictionary 103 generated based on the score calculation example of the pattern 11B shown in FIG. Query candidates can be displayed by controlling the display order.

  As described above, according to the first embodiment, the number of keystrokes that can be saved by presenting query candidates and allowing the user to select them, the number of keystrokes that are predicted when the remaining character strings are input after the user selects a query candidate The query candidate dictionary 103 is generated using the score calculation unit 107 that performs score calculation in consideration of the score after confirmation indicating reduction, the maximum number of display candidates that are the number of items that can be displayed on one screen, and the query candidate presentation unit 110 A query candidate is presented with reference to the query candidate dictionary 103. Therefore, it is possible to present appropriate query candidates according to the screen size, the reading bias of the search target data, and the like, and the effect of improving the user's search keyword input efficiency can be obtained.

Embodiment 2. FIG.
FIG. 19 is a block diagram showing a detailed configuration of the score calculation unit of the query candidate presentation device according to Embodiment 2 of the present invention. As shown in FIG. 19, the score calculation unit 107 of the present embodiment includes a keystroke number shortening score calculation unit 11, a post-confirmation score calculation unit 12, a score totaling unit 13, and a candidate selection score calculation unit 14.
Since the configuration of the query candidate presentation device according to the present embodiment is the same as the configuration of the query candidate presentation device according to the first embodiment shown in FIG. 1 except for the internal configuration of the score calculation unit 107, the score calculation unit 107 Other illustrations and detailed descriptions are omitted.

FIG. 20 is an explanatory diagram showing the definition of the score used by the score calculation unit 107. In the score calculation unit 107 of the present embodiment, the candidate selection score calculation unit 14 calculates the candidate selection score C _sel (N _i ), and the keystroke number shortening score calculation unit 11 uses the candidate selection score C _sel (N _i ). The key calculation number shortening score C ₁ (N _x , N _i ) is different from the score calculation unit 107 of the first embodiment.

In the first embodiment, the maximum display candidate number D _max that is the number of query candidates that can be displayed on one screen is set as the query candidate selection condition 108. However, the number of query candidates that is greater than the number of items that can be displayed on one screen is set. It is also possible to set it as D _max . In such a case, the query candidate presentation unit 110 displays the query candidates over several pages, and the user selects a query candidate by sending a page with a scroll bar or a page feed button.
When the candidate selection score calculation unit 14 presents query candidates over a plurality of pages using a scroll bar or a page feed button or the like in this way, the candidate selection score C _sel (N representing the user's input effort that increases with the number of pages. _i ) Calculate.

  The keystroke number shortening score calculation unit 11 calculates a keystroke number reduction score reflecting the candidate selection score of the candidate selection score calculation unit 14. By extending the score definition of the score calculation unit 107 as shown in FIG. 20, even if the query candidate presentation unit 110 presents query candidates over a plurality of pages by a scroll bar or a page feed button, the score calculation unit The score calculation 107 can be performed in consideration of the degree of contribution of the query candidates displayed on the second page and the third page to reducing the number of keystrokes.

  As described above, according to the second embodiment, the number of keystrokes that can be saved by presenting query candidates and allowing the user to select them, the number of keystrokes that are predicted when the remaining character strings are input after the user selects a query candidate The candidate selection score calculation unit 14 is provided in the score calculation unit 107 that performs score calculation in consideration of the post-confirmation score indicating reduction, the maximum number of display candidates that can be displayed on one screen, and the like. Therefore, the score calculation unit 107 can generate the query candidate dictionary 103 by performing an appropriate score calculation considering the input effort when the query candidate presentation unit 110 presents query candidates over a plurality of pages. Since the query candidate presenting unit 110 is configured to present the query candidates with reference to the query candidate dictionary 103, the query candidate presenting unit 110 presents appropriate query candidates according to the screen size, the reading bias of the search target data, and the like. Thus, the effect of improving the user input efficiency can be obtained.

Embodiment 3 FIG.
FIG. 21 is a block diagram showing a detailed configuration of the score calculation unit of the query candidate presentation device according to Embodiment 3 of the present invention. The score calculation unit 107 of the present embodiment is obtained by adding a candidate importance score calculation unit 15 to the configuration of the score calculation unit 107 of the second embodiment shown in FIG.
Since the configuration of the query candidate presentation device according to the present embodiment is the same as the configuration of the query candidate presentation device according to the first embodiment shown in FIG. 1 except for the internal configuration of the score calculation unit 107, the score calculation unit 107 Other drawings and detailed description are omitted.

FIG. 22 is an explanatory diagram showing the definition of the score used by the score calculation unit 107. In the score calculation unit 107 of the present embodiment, the candidate importance score calculation unit 15 calculates the candidate importance score C _L (N _i ), and the keystroke number shortening score calculation unit 11 calculates the candidate importance score C _L (N _i ) Is used to calculate the keystroke number shortening score C ₁ (N _x , N _i ), which is different from the score calculation unit 107 of the first and second embodiments.

For example, when the query candidate dictionary 103 is generated using a place name or facility name as shown in FIG. 2 as a query candidate, the important phrase extraction unit 104 gives high importance to the famous place name or facility name in the important phrase extraction process ( In the case of a description, it is only necessary to give importance based on screen layout information or the like), and in the trie structure index generation process, the trie structure index generation unit 105 reflects the importance of the phrase on each node (FIG. 22). Of Wgt (N _j )).
Candidate importance score calculation unit 15, a shortcut node N _i with the importance of the phrase corresponding to the leaf node N _j is a descendant Wgt (N _j) of the candidate importance indicating the importance of the query candidates score C _L (N _i ) is calculated.

  The keystroke number reduction score calculation unit 11 calculates a keystroke number reduction score reflecting the candidate importance score of the candidate importance score calculation unit 15. By extending the score definition of the score calculation unit 107 as shown in FIG. 22, it becomes possible to preferentially display particularly important phrases as query candidates regardless of the number of reading characters and the number of leaf nodes.

  As described above, according to the third embodiment, the number of keystrokes that can be saved by presenting query candidates and allowing the user to select them, and the number of keystrokes that are predicted when the remaining character strings are input after the user selects a query candidate The score calculation unit 107 that performs score calculation in consideration of the confirmed score indicating reduction, the maximum number of display candidates that can be displayed on one screen, and the like is configured to include the candidate importance score calculation unit 15. Therefore, the score calculation unit 107 can generate the query candidate dictionary 103 by performing score calculation that gives a high priority to particularly important words regardless of the number of reading characters or the number of leaf nodes. Since the query candidate presenting unit 110 is configured to present the query candidates with reference to the query candidate dictionary 103, the query candidate presenting unit 110 can appropriately display the screen size, the reading bias of the search target data, the importance of the phrase, and the like. Query candidates can be presented, and the effect of improving user input efficiency can be obtained.

It is a block diagram which shows the structure of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the example of the important phrase extracted by the important phrase extraction part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the reading trie structure index which the trie structure index production | generation part of the query candidate presentation apparatus concerning Embodiment 1 of this invention produced | generated. It is explanatory drawing which shows the shortcut of the reading try structure index of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the definition of the score which the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention uses. It is explanatory drawing which shows the score calculation example of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the score calculation example after confirmation of the score calculation part after confirmation of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the score calculation example of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the score calculation example of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the example of score calculation of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention, and shows the time when the _maximum number _{Dmax of} display candidates is five. It is a block diagram which shows the structure of the query candidate dictionary of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the example of the heading list registered into the query candidate dictionary of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the example of the shortcut information registered into the query candidate dictionary of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the query candidate dictionary production | generation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the query candidate presentation part of the query candidate presentation apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows the example of a screen of query candidate presentation which the query candidate display part of the query candidate presentation apparatus concerning Embodiment 1 of this invention displays with respect to a user's input. It is explanatory drawing which shows the example of a screen of query candidate presentation when the maximum display candidate number _Dmax which the query candidate display part of the query candidate presentation apparatus concerning Embodiment 1 of this invention displays is five. It is a block diagram which shows the detailed structure of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows the definition of the score which the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 2 of this invention uses. It is a block diagram which shows the detailed structure of the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 3 of this invention. It is explanatory drawing which shows the definition of the score which the score calculation part of the query candidate presentation apparatus which concerns on Embodiment 3 of this invention uses.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 11 Keystroke shortening score calculation part, 12 Post-confirmation score calculation part, 13 Score totaling part, 14 Candidate selection score calculation part, 15 Candidate importance score calculation part, 21 Heading list, 22 Reading tri structure index, 23, 23a, 23b , 23c Shortcut information, 101 Query candidate dictionary generation unit, 102 Search target data, 103 Query candidate dictionary, 104 Important phrase extraction unit, 105 Tri structure index generation unit, 106 Query candidate extraction unit, 107 Score calculation unit, 108 Query candidate selection Condition, 109 Query candidate selection part, 110 Query candidate presentation part, 111 Reading character string of search keyword, 111a Query candidate selection result, 112 input part, 113 Query candidate display part.

Claims (5)

In a query candidate presentation device including a query candidate display unit that obtains query candidates starting from a reading keyword of a search keyword input character by character from the query candidate dictionary and presents the query candidates as candidates for the search keyword.
A query candidate extraction unit that extracts a character string serving as the query candidate from data obtained by hierarchically structuring the word extracted from the search target data based on the reading characters of the word;
A score calculation unit for calculating a score indicating the effect of reducing input labor of the search keyword obtained by setting the character string starting with the reading character of the search keyword as a query candidate;
A query candidate selection unit that registers the character string as a query candidate in the query candidate dictionary based on the score calculated by the score calculation unit for each reading character of the phrase extracted from the search target data. A query candidate presentation device as a feature. The score calculator
A keystroke number shortening score calculation unit for calculating a keystroke number shortening score indicating the number of keystrokes of the search keyword to be reduced by setting a character string starting with a reading character of the search keyword as a query candidate;
After confirming the effect of reducing the number of keystrokes of the search keyword predicted by presenting a new query candidate starting with the character string when it is determined that the character string is a part of the search keyword A post-confirmation score calculator for calculating the score;
A score totaling unit that calculates a score indicating an effect of reducing input labor of the search keyword obtained by using the character string as a query candidate by using the keystroke number shortening score and the post-confirmation score. The query candidate presentation device according to claim 1, wherein: The score calculator
A keystroke number shortening score calculation unit for calculating a keystroke number shortening score indicating the number of keystrokes of the search keyword to be reduced by setting a character string starting with a reading character of the search keyword as a query candidate;
After confirming the effect of reducing the number of keystrokes of the search keyword predicted by presenting a new query candidate starting with the character string when it is determined that the character string is a part of the search keyword A post-confirmation score calculator for calculating the score;
A candidate selection score calculation unit that calculates a candidate selection score indicating the input effort of the search keyword that increases by presenting query candidates over a plurality of pages based on the number of query candidates that can be presented simultaneously by the query candidate display unit. ,
The score totaling unit is a score indicating an effect of reducing input labor of the search keyword obtained by presenting the character string as a query candidate using the keystroke number shortening score, the post-confirmation score, and the candidate selection score. The query candidate presentation device according to claim 1, wherein: The score calculator
A keystroke number shortening score calculation unit for calculating a keystroke number shortening score indicating the number of keystrokes of the search keyword to be reduced by setting a character string starting with a reading character of the search keyword as a query candidate;
After confirming the effect of reducing the number of keystrokes of the search keyword predicted by presenting a new query candidate starting with the character string when it is determined that the character string is a part of the search keyword A post-confirmation score calculator for calculating the score;
A candidate selection score calculation unit that calculates a candidate selection score indicating the input effort of the search keyword that is increased by presenting query candidates over a plurality of pages based on the number of query candidates that can be simultaneously presented by the query candidate display unit;
A candidate importance score calculation unit that calculates a candidate importance score indicating the importance of the character string based on the importance of the phrase extracted from the search target data,
The score totaling unit reduces the input effort of the search keyword obtained by using the character string as a query candidate by using the keystroke reduction score, the post-confirmation score, the candidate selection score, and the candidate importance score. The query candidate presentation device according to claim 1, wherein a score indicating the effect of the query is calculated.   The query candidate selection unit registers, for each reading character of the phrase extracted from the search target data, a number of character strings corresponding to the number of query candidates that can be presented simultaneously by the query candidate display unit as query candidates in the query candidate dictionary. 5. The query candidate presentation device according to claim 1, wherein: the query candidate presentation device according to claim 1.

Images