JP6668855B2 - Search device, search method and program - Google Patents

Description

  The present invention relates to a search device, a search method, and a program.

  As the number of electronic documents increases, the importance of technology for efficiently searching for a desired electronic document from a large-scale document database has increased.

  For example, Patent Literature 1 and Patent Literature 2 disclose a search technique using an N-gram. More specifically, Patent Document 1 classifies text of dictionary data to be searched into a headword category and a text category, performs a head-on search for texts belonging to the headword category, and searches for text and text belonging to the headword category. A text search device that performs a partial match search for text belonging to a category is disclosed. Patent Document 2 discloses an ambiguous search technology that can search for a desired document even if a search character string is input by mistake.

JP 2013-161371 A JP 2014-146301 A

  In searching for an electronic document using the N-gram as described above, it is desired to improve the search efficiency and shorten the search time.

  An object of the present invention is to provide a search device, a search method, and a program that can reduce a search time in a search for an electronic document using an N-gram. Aim.

In order to achieve the above object, a search device according to the present invention comprises:
A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. Storage means for storing the second index;
Acquisition means for acquiring a search keyword;
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. First search means;
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. Second search means for performing
Output means for outputting a search result by at least one of the first search means and the second search means;
It is characterized by having.

  According to the present invention, in an electronic document search using an N-gram, the search time can be reduced.

It is a figure showing appearance of a search device concerning an embodiment of the present invention. FIG. 2 is a block diagram illustrating a hardware configuration of a search device. FIG. 2 is a block diagram illustrating a functional configuration of the search device. FIG. 4 is a diagram showing dictionary data and relocation data. FIG. 3 is a first diagram illustrating a configuration of a search index. FIG. 9 is a second diagram illustrating a configuration of a search index. It is a figure showing an example of an incremental search. It is a figure showing an example of a display screen of a result of full-text search. It is a figure showing the example of the display screen of the result of a fuzzy search. It is a flowchart which shows the flow of a search index generation process. It is a flowchart which shows the flow of an incremental search process. 11 is a first flowchart illustrating a flow of a full-text search process. 13 is a second flowchart illustrating the flow of a full-text search process. 13 is a third flowchart illustrating the flow of a full-text search process. 15 is a fourth flowchart illustrating the flow of the full-text search process. 15 is a fifth flowchart illustrating the flow of the full-text search process. It is a flowchart which shows the flow of an ambiguous search process.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the drawings, the same or corresponding portions are denoted by the same reference numerals.

  FIG. 1 shows an appearance of a search device 100 according to an embodiment of the present invention. The search device 100 is, for example, an electronic dictionary. As shown in FIG. 1, the search device 100 includes a keyboard 100 i for receiving an input of a search keyword in accordance with a user operation, an LCD (Liquid Crystal Display) 100 h for displaying a search result obtained by searching a dictionary based on the search keyword. , Is provided.

  As shown in FIG. 2, the search device 100 includes a CPU (Central Processing Unit) 100a, a ROM (Read Only Memory) 100b, a RAM (Random Access Memory) 100c, a hard disk 100d, a media controller 100e, a video card 100g, and a speaker 100j. Is provided inside. These components are connected to the LCD 100h and the keyboard 100i via a bus.

  The CPU 100a is, for example, a microprocessor or the like, and is a central processing unit that executes various processes and calculations. The CPU 100a is also called a central processing unit, a central processing unit, a processor, or the like. The CPU 100a is connected to each unit of the search device 100 via a system bus, which is a transmission path for transferring commands and data, and controls the entire search device 100. The CPU 100a executes various programs including the following search processes by executing programs stored in the ROM 100b or the hard disk 100d. The RAM 100c is used as a work area for temporarily storing data to be processed when the CPU 100a executes a program.

  The hard disk 100d is a non-volatile storage unit that stores a table storing various data and dictionary data such as an English-Japanese dictionary, a Japanese language dictionary, or an encyclopedia. Note that the search device 100 may include a flash memory instead of the hard disk 100d.

  The media controller 100e reads various data and programs from recording media including a flash memory, a CD (Compact Disc), a DVD (Digital Versatile Disc), and a Blu-ray Disc (Blu-ray Disc) (registered trademark).

  The video card 100g draws (ie, renders) an image based on the digital signal output from the CPU 100a, and outputs an image signal representing the drawn image. The LCD 100h is a display device that displays an image according to an image signal output from the video card 100g. The search device 100 may include a PDP (Plasma Display Panel) or an EL (Electroluminescence) display instead of the LCD 100h. The speaker 100j outputs sound based on the signal output from the CPU 100a.

  As illustrated in FIG. 3, the search device 100 functionally includes a generation unit 120, an acquisition unit 130, a search unit 140, an evaluation unit 150, an output unit 160, and a display unit 170. The CPU 100a reads the program stored in the ROM 100b into the RAM 100c, and executes and controls the program to function as these units. Further, the search device 100 functionally includes a data storage unit 110 and a display unit 170. The data storage unit 110 is constructed in a storage area of the RAM 100c or the hard disk 100d. The display unit 170 is a display device such as the LCD 100h.

  The data storage unit 110 stores dictionary data, which is an example of an electronic document to be searched. As shown in FIG. 4, the dictionary data includes a text representing a headword (hereinafter, referred to as “headword text”) and a text representing an explanation of the headword (hereinafter, referred to as “description text”). , Example sentences, phrases, compound words, and the like (hereinafter, referred to as “idioms”) (hereinafter, referred to as “example texts”).

  Hereinafter, the description text and the example text are collectively referred to as a body text. In other words, the text included in the dictionary data is classified into a text belonging to the headword category (headword text) and a text belonging to the body text category (body text). The headword category is called a first category, and the body category is called a second category. In the dictionary data, a portion occupied by the headword text is referred to as a heading portion, a portion occupied by the explanation text is referred to as an explanation portion, and a portion occupied by the example text is referred to as an example portion.

  The dictionary data is composed of a set of a heading part, a description part, and an example part as one constituent unit, and these constituent units are connected to each other. For example, if the dictionary data is an English-Japanese dictionary, the constituent units are arranged in alphabetical order of the headword text. In each of the constituent units, an explanation part for explaining the headword is arranged immediately after the heading part, and an example part showing an example of the headword is arranged immediately after the heading part.

  In the dictionary data, a plurality of explanation units are arranged according to the arrangement order determined by the editor of the electronic dictionary. More specifically, the explanation section for explaining the more general meaning content of the headword is stored at a position earlier than the explanation section for explaining the more specific meaning content of the headword. Alternatively, the explanation unit that explains the meaning content that is used more frequently may be stored at a position after the explanation unit that explains the meaning content that is used less frequently. As for the example section, there are cases where an example section in which an example corresponding to the explanation section is described is described for each corresponding explanation section, and a case where a plurality of example sections are described collectively after one explanation section. There is.

  The generation unit 120 generates a search index in which a plurality of N-grams included in dictionary data, which is an electronic document to be searched, are associated with appearance positions of the plurality of N-grams in the dictionary data. The N-gram is a character string of N consecutive characters (N is a natural number). For example, an N gram when N is 1 is a monogram, an N gram when N is 2 is a bigram, and an N gram when N is 3 is a trigram. The generation unit 120 is realized by the CPU 100a cooperating with the RAM 100c or the hard disk 100d or the like. The generation unit 120 functions as a generation unit. Further, the generation unit 120 functions as an index acquisition unit that acquires a search index by generating a search index.

  The generation unit 120 has functions of a relocation data generation unit 121 and a search index generation unit 122 to generate a search index.

  The rearrangement data generation unit 121 reads the dictionary data stored in the data storage unit 110 and generates rearrangement data by rearranging text included in the dictionary data. More specifically, in the dictionary data, as shown on the left side of FIG. 4, the description part and the example part are alternately arranged for each headword. The rearrangement data generation unit 121 rearranges the explanation units and the example units that are alternately arranged so as to be arranged collectively, and generates the rearrangement data illustrated on the right side of FIG. 4.

  The search index generation unit 122 generates a search index from the relocation data generated by the relocation data generation unit 121. More specifically, the search index generation unit 122 sequentially extracts N-grams included in the relocation data, and generates a table in which each extracted N-gram is associated with its appearance position as a search index. The search index generator 122 uses monograms and bigrams as N-grams.

  5 and 6 show the structure of the search index. As shown in FIG. 5, the search index includes a heading part index, an explanation part index, and an example part index.

  The heading index associates a plurality of N-grams (specifically, monograms and bigrams) included in the headword text in the dictionary data with the appearance positions of the plurality of N-grams in the headword text. It is an index. The heading index includes (1) a file (pattern_hdl.idx) related to the character string pattern of the N-gram included in the heading, and (2) a file (position_hdl.idx) related to its appearance position and frequency. .

  The explanation part index is an index that associates a plurality of N-grams included in the explanation text in the dictionary data with the appearance positions of the plurality of N-grams in the explanation text. The explanation part index includes (3) a file (pattern_bdy.idx) related to the character string pattern of the N-gram included in the explanation part, and (4) a file (position_bdy.idx) related to its appearance position and appearance frequency. .

  The example section index is an index that associates a plurality of N-grams included in the example text in the dictionary data with the appearance positions of the plurality of N-grams in the example text. The example section index includes (5) a file (pattern_exp.idx) related to the character string pattern of the N-gram included in the example section, and (6) a file (position_exp.idx) related to its appearance position and frequency. .

  The search index generation unit 122 separately extracts monograms and bigrams from the headword text, explanatory text, and example text while shifting one character from the beginning, and derives the appearance position and frequency of the extracted monograms and bigrams. I do. Then, the search index generation unit 122 generates such three types of indexes by associating the extracted monograms and bigrams with their appearance positions and appearance frequencies.

  Note that the heading section index is referred to as a first index, and the description section index and the example section index are collectively referred to as a second index.

  As shown in FIG. 6, the search index includes (7) a file related to a heading number (number.idx), (8) a file related to a start position of a heading part (headline.idx), and (9) an example section. File about the start position (example.idx). The file (number.idx) relating to the heading number contains information on the starting position of the heading part, the number of heading words, the starting position of the explanation part, the starting position of the example part, the number of examples, the dictionary number, and the heading number in heading word units. Contains. The dictionary number is a number for identifying dictionary data. The heading number is a number for identifying each of the plurality of heading parts.

  Note that the appearance position of the N-gram character string pattern included in the heading index and the starting position of the heading in the file related to the heading number are only the headword text from the first character of the heading of the relocation data (FIG. 4). Are positions derived by counting the number of characters in the heading portion 1 to the heading portion M). Similarly, the appearance position of the N-gram character string pattern included in the explanation part index and the start position of the explanation part in the file related to the index number are only the explanation text from the first character of the explanation part of the relocation data (FIG. 4). Are the positions derived by counting the number of characters in the explanation units 1 to M). Similarly, the appearance position of the N-gram character string pattern included in the example part index and the start position of the example part in the file related to the index number are only the example text from the first character of the example part of the relocation data (FIG. 4). Is a position derived by counting the number of characters in the example section 1 to the example section M).

  Returning to the description of the functional configuration of the search device 100 shown in FIG. The data storage unit 110 stores the search index generated by the generation unit 120. The data storage unit 110 functions as a storage unit.

  The acquisition unit 130 acquires a search keyword. The user can input a desired search keyword to the search device 100 by operating the keyboard 100i. When a plurality of electronic documents are registered in advance in the search device 100, the acquisition unit 130 operates the keyboard 100i to specify the electronic document to be searched from among the plurality of electronic documents. be able to. The acquisition unit 130 acquires the search keyword and the electronic document specification thus input. The acquisition unit 130 is realized by the CPU 100a cooperating with the keyboard 100i and the like. The acquisition unit 130 functions as an acquisition unit (search keyword acquisition unit).

  The search unit 140 searches the dictionary data, which is the electronic document to be searched, for the search keyword acquired by the acquisition unit 130. More specifically, the search unit 140 reads, from the search index, the appearance positions of a plurality of N-grams included in the search keyword in the dictionary data, and evaluates the continuity of the read appearance positions. Search for. The search unit 140 is realized by the CPU 100a cooperating with the RAM 100c and the like.

  More specifically, the search unit 140 includes functions of an incremental search unit 141, a full-text search unit 142, and an ambiguity search unit 143. The search unit 140 can execute these three types of searches.

  The incremental search unit 141 performs an incremental search on the heading part. The incremental search is a head-on search that is executed each time a search keyword is input by the user one character at a time. Each time the acquisition unit 130 acquires a search keyword one character at a time, the incremental search unit 141 uses, as a search keyword, a character string in which at least one character already acquired by the acquisition unit 130 and one newly acquired character are added. Search for.

  Specifically, an example in which the incremental search is executed will be described with reference to FIG. When the user inputs a search keyword one character at a time via the keyboard 100i, the obtaining unit 130 obtains the search keyword one character at a time in accordance with the input. When the user inputs the first character of the search keyword (“w” in the example of FIG. 7), the incremental search unit 141 refers to the index of the heading part and starts with “w” in the dictionary data to be searched. Search for a headword that starts. As a result, as shown on the left side of FIG. 7, a list of headwords beginning with "w" is generated.

  When the next one character (“h” in the example of FIG. 7) is input, the incremental search unit 141 compares the one character “w” already acquired by the acquisition unit 130 with the one character “h” newly acquired. The headword having the character string “wh” at the beginning of the character string is further searched with reference to the heading index. As a result, as shown in the center of FIG. 7, a list of headwords beginning with "wh" is generated. Further, when the next one character (“i” in the example of FIG. 7) is input, the incremental search unit 141 compares the two characters “wh” already acquired by the acquisition unit 130 and the newly acquired one character “i” "Is further searched for with reference to the index of the heading, which has the character string" whi "at the beginning. As a result, as shown on the right side of FIG. 7, a list of headwords beginning with "whi" is generated.

  As described above, the incremental search unit 141 searches the headline text belonging to the headword category for a character string starting with the search keyword by referring to the headline index each time the search keyword is input one character at a time by the user. Perform a prefix search.

  The full-text search unit 142 performs a full-text search on the entire dictionary data. The full-text search is an exact match search that is executed after the search keyword is determined by, for example, pressing the return key by the user. When the search keyword is determined, the full-text search unit 142 executes a full-text search that searches the entire dictionary data for a character string including the search keyword with reference to the index of the index, the description, and the example.

  For example, when a character string “while” is acquired as a search keyword, the full-text search unit 142 cuts out two characters in order from the beginning to the end of the search keyword “while”, and outputs “wh”, “hi”, The four bigrams “il” and “le” are extracted. Then, the full-text search unit 142 reads the appearance position of each bigram extracted from the heading index, the explanation index, and the example index stored in the data storage unit 110, and evaluates the continuity. More specifically, the full-text search unit 142 determines whether the extracted four bigrams “wh”, “hi”, “il”, and “le” appear at positions shifted by one character in this order. I do. As a result of the evaluation, when it is evaluated that the appearance position of each bigram has continuity, the full-text search unit 142 determines that a character string corresponding to the search keyword exists at the appearance position.

  FIG. 8 shows an example in which a full-text search is performed for a search keyword “while”. As illustrated in FIG. 8, the full-text search unit 142 performs an exact match search of the character string with “while” for each of the heading part, the description part, and the example part. As a result, a text including the search keyword (“while” in the example of FIG. 8) is found in each of the heading section, the explanation section, and the example section.

  The fuzzy search unit 143 performs a fuzzy search on the heading part. The fuzzy search is a partial match search that can search for a character string that partially matches but does not completely match the search keyword. The fuzzy search unit 143 executes a fuzzy search when the user operates the keyboard 100i to input a search keyword and selects the fuzzy search mode. The details of the fuzzy search are disclosed in Patent Document 2 described above, and will be briefly described here.

  The fuzzy search unit 143 reads, from the heading index, the appearance positions of the plurality of N-grams included in the search keyword in the text belonging to the headword category, and the read out appearance position is continuous with the positions of the plurality of N-grams in the search keyword. A search keyword is searched from the headword text based on the number of matching words (the number of consecutive matching N-grams). The number of consecutive matching N-grams is the number at which the appearance position in the headword text of each N-gram extracted from the search keyword continuously matches the position of each corresponding N-gram in the search keyword. The fuzzy search unit 143 calculates the number of consecutive matching N-grams, and determines that the higher the number of consecutive matching N-grams, the higher the evaluation (the higher the probability that the result is a result desired by the user).

  FIG. 9 shows an example in which the search keyword “saled” is ambiguously searched. As shown in FIG. 9, the fuzzy search unit 143 performs a fuzzy search to find a character string close to the character string “saled” from the headword. As a result, a headword such as "salad" is found.

  The search targets of the incremental search unit 141 and the fuzzy search unit 143 are only texts belonging to the headword category (first category) in the dictionary data. In other words, the incremental search unit 141 and the fuzzy search unit 143 each function as a first search unit that searches for a search keyword from text belonging to the first category. On the other hand, the search target of the full-text search unit 142 is the entire dictionary data, in other words, both the headword category (first category) and the body category (second category). In other words, the full-text search unit 142 functions as a second search unit that searches for a search keyword from texts belonging to the first category and texts belonging to the second category.

  The evaluation unit 150 calculates a ranking evaluation value for the search result obtained by the search unit 140. The ranking evaluation value is a value indicating a priority when outputting a search result. The evaluation unit 150 is realized by the CPU 100a cooperating with the RAM 100c and the like. The evaluation unit 150 functions as an evaluation unit.

  For example, the evaluation unit 150 evaluates the ranking higher as the appearance position of the search keyword obtained by the search is closer to the head of the heading part, the explanation part, or the example part. In addition, when there are a plurality of headwords in one heading, the evaluation unit 150 evaluates higher as the headword closer to the head is obtained as a search result. The same applies to the explanation section and the example section. Further, in the fuzzy search, the evaluation unit 150 ranks higher as the number of characters matching the search keyword increases. Further, in the case of an AND search in a full-text search, the evaluation unit 150 evaluates the rank higher as the appearance positions of the two search keywords are closer.

  The output unit 160 outputs a search result obtained by the search unit 140. Specifically, when the search is executed by the incremental search unit 141, the full-text search unit 142, or the fuzzy search unit 143, the output unit 160 calculates the appearance position of the search keyword obtained by the search by the evaluation unit 150. The sorted ranking evaluation values are sorted in descending order. Then, the output unit 160 displays information indicating the rearranged appearance positions on the display unit 170 as a search result. Alternatively, the output unit 160 may output the search result by voice from the speaker 100j. The output unit 160 is realized when the CPU 100a cooperates with the video card 100g, the LCD 100h, the speaker 100j, and the like. The output unit 160 functions as an output unit.

  More specifically, when the incremental search is performed, the output unit 160 displays, for example, a list of headwords as shown in FIG. 7 on the display unit 170 every time the user inputs a search keyword one character at a time. . On the other hand, when the full-text search is executed, the output unit 160 divides the search result into a heading section, an explanation section, and an example section and displays the search result on the display section 170 as shown in FIG. 8, for example. When the fuzzy search is executed, the output unit 160 displays the fuzzy search result on the display unit 170, for example, as illustrated in FIG.

  The flow of processing executed by the search device 100 configured as described above will be described with reference to the flowcharts shown in FIGS.

  First, the flow of the search index generation process executed by the generation unit 120 will be described with reference to the flowchart shown in FIG. The search index generation processing is divided into relocation data generation processing and search index generation processing. The search index generation process illustrated in FIG. 10 starts when, for example, the user operates the keyboard 100i and inputs an instruction to generate a search index.

  When the search index generation process starts, the relocation data generation unit 121 reads out the dictionary data stored in the data storage unit 110 (Step S1). Then, the relocation data generation unit 121 extracts a headword text, an explanation text, and an example text from the read dictionary data (step S2).

  The rearrangement data generation unit 121 adds a delimiter to each of the head and end of the extracted headword text (step S3). The delimiter character is a character that functions as a mark for indicating a delimiter between headwords. A character that does not appear in the headword text, such as a space or a special character, is used as a delimiter.

  When a delimiter is added to the headword text, the rearrangement data generation unit 121 generates rearrangement data as shown on the right side of FIG. 4 in which the description text and the example text are combined for each headword, and the data storage unit The data is stored in the file 110 (step S4).

  Next, the search index generation unit 122 generates a search index from the relocation data generated by the relocation data generation unit 121 (Step S5). More specifically, the search index generation unit 122 extracts monograms and bigrams for each of the heading part, the description part, and the example part in the rearranged data while shifting one character from the beginning, and the appearance of the extracted monograms and bigrams The position and appearance frequency are derived. Then, the search index generation unit 122 generates a search index as shown in FIGS. 5 and 6 by associating the appearance position and the appearance frequency of each of the extracted monogram and bigram, and stores the search index in the data storage unit 110. I do. Thus, the search index generation processing in FIG. 10 ends.

  Next, the flow of a search process executed by the search unit 140 will be described with reference to flowcharts shown in FIGS. As described above, the search unit 140 performs an incremental search process, a full-text search process, and an ambiguous search process as the search processes.

  First, the flow of the incremental search process will be described with reference to FIG. Each time the user operates the keyboard 100i to specify a target dictionary and inputs a search keyword one character at a time, the incremental search unit 141 starts the incremental search processing shown in FIG. Hereinafter, a case where a character string “whi” is input as a search keyword will be described as an example.

  When the incremental search process starts, the incremental search unit 141 generates a search character string from an input character string input as a search keyword (Step S11). More specifically, the incremental search unit 141 generates a search character string by adding a delimiter to the head of the input character string. For example, when the delimiter is represented as “•”, the incremental search unit 141 generates a search character string “• whi” by adding “•” to the beginning of the search keyword “whi”.

  After generating the search character string, the incremental search unit 141 extracts a bigram from the search character string (Step S12). More specifically, the incremental search unit 141 cuts out two characters in order from the beginning to the end of the search character string “· whi”, and divides the three bigrams of “· w”, “wh”, and “hi”. Extract. Then, the incremental search unit 141 sets the search character string as a reference character string (Step S13).

  When the bigram is extracted, the incremental search unit 141 refers to the index part index ((1) pattern_hdl.idx and (2) position_hdl.idx shown in FIG. 5) and displays the appearance position of each bigram extracted from the reference character string. Is read, and its continuity is evaluated (step S14). More specifically, the incremental search unit 141 determines whether the three extracted bigrams “· w”, “wh”, and “hi” appear at positions shifted by one character in this order in the headword text. judge.

  As a result of the evaluation, when it is determined that there is no continuity (step S15; NO), the incremental search unit 141 similarly evaluates the continuity of the other appearance positions read from the heading index in step S14.

  On the other hand, if it is determined that there is continuity (step S15; YES), the incremental search unit 141 determines the appearance position of the leading bigram as the appearance position of the reference character string (step S16). Then, the incremental search unit 141 refers to the file relating to the index number of the search index and the files relating to the start position of the index part ((7) number.idx and (8) headline.idx shown in FIG. 6), and The dictionary number and the index number are derived as search results from the appearance position of the character string (step S17). Then, the evaluation unit 150 calculates a ranking evaluation value for the search result (step S18).

  Next, the incremental search unit 141 determines whether an unevaluated appearance position exists (step S19), and when it determines that it exists (step S19; YES), returns to step S14. On the other hand, when it is determined that there is no unevaluated appearance position (step S19; NO), the output unit 160 arranges the search results based on the ranking evaluation value (step S20). Then, the output unit 160 outputs the search result to the display unit 170 in headword units (step S21). The output unit 160 displays the display screen illustrated in FIG. 7 on the display unit 170, for example. Thus, the incremental search processing ends.

  Second, the flow of the full-text search process will be described with reference to the flowcharts shown in FIGS. When the user operates the keyboard 100i to specify the target dictionary and finalize the search keyword, the full-text search unit 142 starts the full-text search processing shown in FIGS.

  When the full-text search process starts, the full-text search unit 142 generates a search character string from an input character string input as a search keyword (step S31). More specifically, the full-text search unit 142 generates a search character string by adding delimiters to both ends of the input character string. For example, when a character string of “while” is input as a search keyword, the full-text search unit 142 adds “•” to the beginning and end of the search keyword “while”, and thereby searches for “• while ·”. Generate a search string.

  After generating the search character string, the full-text search unit 142 determines whether the number of search character strings is one (Step S32). In other words, the full-text search unit 142 determines whether the current search is an AND search using a plurality of search keywords. If the number of search character strings is one (step S32; YES), full-text search unit 142 advances the process to step S33, while if the number of search character strings is plural (step S32; NO). ) And the process proceeds to step S35.

  When there is only one search character string (step S32; YES), the full-text search unit 142 extracts a monogram and a bigram from the search character string (step S33). More specifically, when the search character string is a three-character character string (· + 1 character input character string + ·), the full-text search unit 142 outputs one character string corresponding to one character input character string. Extract a monogram and two bigrams including the delimiters at both ends. On the other hand, when the search character string is a character string of four or more characters (. + An input character string of two or more characters +.), The full-text search unit 142 extracts bigrams in order from the beginning to the end of the search character string. For example, six bigrams of ".w", "wh", "hi", "il", "le", and "e." Are extracted from the search character string ".while."

  Next, the full-text search unit 142 sets the search character string as a reference character string (Step S34). Then, the full-text search unit 142 proceeds to a process of evaluating the continuity of the extracted bigram in the order of the heading, the explanation, and the example, that is, for each category.

  Referring to FIG. 13, the full-text search unit 142 determines whether the number of characters in the reference character string is three (step S51). If it is determined that the number of characters of the reference character string is three (step S51; YES), the full-text search unit 142 determines the index of the heading part ((1) pattern_hdl.idx and (2) position_hdl. With reference to (idx), the appearance position of one monogram extracted from the reference character string is read, and this is specified as the appearance position of the reference character string (step S52).

  When the appearance position of the reference character string is specified, the full-text search unit 142 derives a search state (Step S53). The search state is information indicating whether or not the search keyword is the headword itself. More specifically, the full-text search unit 142 reads the appearance positions of the two bigrams including the delimiter from the index of the index, and evaluates the continuity. If there is continuity, a search state that the search keyword is the headword itself is derived.

  On the other hand, when it is determined in step S51 that the number of characters of the reference character string is four or more (step S51; No), the full-text search unit 142 extracts the character string from the reference character string with reference to the index of the heading part. The appearance position of the bigram that does not include the delimiter character in the bigram is read and its continuity is evaluated (step S54). As a result of the evaluation, when it is determined that there is no continuity (step S55; NO), the full-text search unit 142 similarly evaluates the continuity of the other appearance positions read from the index of the index in step S54.

  On the other hand, when it is determined that there is continuity (step S55; YES), the full-text search unit 142 determines that the appearance position of the second bigram is the appearance position of the reference character string (step S56). Then, the full-text search unit 142 derives the search state by reading the appearance positions of the two bigrams including the delimiter from the index of the heading part and evaluating the continuity thereof (step S57).

  When the search state is derived in step S53 or step S57, the full-text search unit 142 compares the file ((7) number.idx) regarding the index number of the search index and the file ((8) headline.idx) regarding the start position of the index part. , A dictionary number, a heading number, and an example number are derived as search results from the appearance position of the reference character string (step S58). Then, the evaluation unit 150 calculates a ranking evaluation value for the search result (step S59).

  Next, the full-text search unit 142 determines whether an unevaluated appearance position exists (step S60). If an unevaluated appearance position exists (step S60; YES), the process returns to step S51. On the other hand, when there is no unevaluated appearance position (step S60; NO), the full-text search unit 142 determines whether an unevaluated category exists (step S61). If there is an unevaluated category (step S61; YES), the full-text search unit 142 returns to step S51, and executes the same process for the unevaluated category in the order of the explanation part and the example part.

  Finally, when it is determined that there is no category that has not been evaluated (step S61; NO), the output unit 160 rearranges the search results based on the ranking evaluation value (step S62). Then, the output unit 160 outputs the search result to the display unit 170 (Step S63). At this time, as shown in FIG. 8, the output unit 160 displays the search results from the heading part, the explanation part, and the example part on the display part 170 in headword unit, explanation part unit, and example part unit, respectively. Thus, the full-text search process in the case where the number of search character strings is one is completed.

  Returning to FIG. 12, in step S32, when the number of search character strings is plural (step S32; NO), the full-text search unit 142 extracts a monogram and a bigram from each search character string (step S35). . In other words, the full-text search unit 142 performs the same processing as in step S32 for each of the plurality of search character strings.

  Next, the full-text search unit 142 extracts the appearance frequency of each of the extracted monogram and bigram with reference to the heading index, the explanation index, and the example index (step S36). Then, the full-text search unit 142 specifies the minimum appearance frequency among the monograms and bigrams constituting each search character string (step S37).

  After specifying the minimum appearance frequency, the full-text search unit 142 determines a reference character string and a verification character string from a plurality of search character strings (step S38). More specifically, the full-text search unit 142 determines one search character string including a monogram or a bigram with the lowest appearance frequency as a reference character string, and determines at least one other search character string as a verification character string. Set. This is because, in the AND search, performing a search based on a search character string having a low appearance frequency can reduce the processing load and increase the search speed. Alternatively, in the case of a search in consideration of the input order of the search character strings, the full-text search unit 142 determines the search character string input first as the reference character string, and determines the search character string input second or later as the verification character string. It may be determined as a column.

  Thereafter, the full-text search unit 142 proceeds to a process of evaluating the continuity of the extracted bigram in the order of the heading, the explanation, and the example, that is, for each category.

  Referring to FIG. 14, the full-text search unit 142 determines whether the number of characters in the reference character string is three (step S71). Thereafter, the full-text search unit 142 specifies the appearance position of the reference character string and derives the search state (steps S71 to S77). The processing from step S71 to step S77 is the same as the processing from step S51 to step S57 described in the case where the number of search character strings is one, and therefore the description is omitted.

  When the appearance position of the reference character string is specified and its search state is derived, the full-text search unit 142 determines a file ((7) number.idx) relating to the index number of the search index and a file ((8)) relating to the start position of the index part. With reference to (headline.idx), a dictionary number, a heading number, and an example number are derived as search result candidates from the appearance position of the reference character string (step S78).

  When the evaluation processing of the reference character string ends in this way, the full text search unit 142 proceeds to FIG. 15, and executes the evaluation processing of the verification character string. In FIG. 15, the full-text search unit 142 determines whether the number of characters in the verification character string is three (step S81).

  When it is determined that the number of characters of the verification character string is three (step S81; YES), the full-text search unit 142 reads the appearance position of the extracted one monogram from the index of the heading part, and verifies this. The appearance position of the character string is specified (step S82). Then, the full-text search unit 142 derives the search state by reading the appearance positions of the two bigrams including the delimiter from the index of the heading part and evaluating the continuity thereof (step S83).

  When the search state is derived, the full-text search unit 142 evaluates whether the verification character string exists within the range specified by the dictionary number, index number, and example number of the search result candidate derived in step S78 (step S84). ). This range is a range where the reference character string and the verification character string are hit as an AND search. For example, if both the reference character string and the verification character string exist in the same headword constituent unit (including the description part and the example part) in the same dictionary data, the AND search is set to hit. Alternatively, if both the reference character string and the verification character string exist in the same example section, the setting can be set so that the AND search is hit.

  As a result of the evaluation, when it is determined that the appearance position of the verification character string is not more than the minimum position of the specified range (Step S85; NO), the full-text search unit 142 returns the process to Step S82. On the other hand, when determining that the appearance position of the verification character string is equal to or greater than the minimum position (step S85; YES), the full-text search unit 142 determines whether the appearance position of the verification character string is equal to or less than the maximum position within the specified range. Is evaluated (step S86). When determining that the appearance position of the verification character string is not less than the maximum position (step S86; NO), the full-text search unit 142 shifts the processing to step S103 in FIG. On the other hand, when determining that the appearance position of the verification character string is equal to or less than the maximum position (step S86; YES), the full-text search unit 142 shifts the processing to step S94.

  If it is determined in step S81 that the verification character string has four or more characters (step S81; NO), the full-text search unit 142 extracts the second bigram of the bigram extracted from the verification character string from the heading index. The appearance position of the bigram is read and identified as the appearance position of the verification character string (step S87).

  When the appearance position of the verification character string is specified, the full-text search unit 142 evaluates whether the verification character string exists within the range specified by the dictionary number, heading number, and example number of the search result candidate derived in step S78. (Step S88).

  As a result of the evaluation, when it is determined that the appearance position of the verification character string is not greater than or equal to the minimum position of the specified range (Step S89; NO), the full-text search unit 142 returns the process to Step S87. On the other hand, when determining that the appearance position of the verification character string is equal to or greater than the minimum position (step S89; YES), the full-text search unit 142 determines whether the appearance position of the verification character string is equal to or less than the maximum position within the specified range. Is evaluated (step S90). If it is determined that the appearance position of the verification character string is not less than the maximum position (step S90; NO), the full-text search unit 142 shifts the processing to step S103 in FIG.

  On the other hand, if it is determined that the appearance position of the verification character string is equal to or less than the maximum position (step S90; YES), the full-text search unit 142 determines the delimiter character of the bigram extracted from the verification character string from the heading index. The appearance position of the bigram that is not included is read and its continuity is evaluated (step S91). As a result of the evaluation, when it is determined that there is no continuity (step S92; NO), the full-text search unit 142 returns the process to step S87 and executes the same process for another occurrence position of the second bigram.

  On the other hand, when it is determined that there is continuity (step S92; YES), the full-text search unit 142 reads the appearance position of the bigram including the delimiter from the index of the heading part, and evaluates the continuity. Then, a verification state is derived (step S93). Then, the full-text search unit 142 shifts the processing to step S94.

  In step S94, the full-text search unit 142 determines whether the evaluation has been completed for all verification character strings (step S94). If it is determined that the evaluation has not been completed for all verification character strings (step S94; NO), the process returns to step S81, and the same processing is executed for all verification character strings.

  On the other hand, when it is determined that the evaluation has been completed for all the verification character strings (step S94; YES), the full-text search unit 142 proceeds to FIG. 16 and compares the search result candidates obtained by the above processing with the AND search. A search result is determined (step S101). In other words, the full-text search unit 142 determines the appearance position of one reference character string and the appearance position of at least one verification character string existing in the specified range as the positions where the AND search has been performed. The evaluation unit 150 calculates a ranking evaluation value for the search result of the AND search (step S102).

  Next, the full-text search unit 142 determines whether an unevaluated appearance position exists (step S103). If it determines that an unevaluated appearance position exists (step S103; YES), the full-text search unit 142 proceeds to FIG. The process returns to step S71 shown in FIG. On the other hand, when determining that there is no unevaluated appearance position (step S103; NO), the full-text search unit 142 determines whether an unevaluated category exists (step S104). If there is an unevaluated category (step S104; YES), the full-text search unit 142 returns to step S71, and executes the same processing for the unevaluated category in the order of the explanation unit and the example unit.

  Finally, when it is determined that there is no unrated category (step S104; NO), the output unit 160 sorts the search results by category based on the ranking evaluation value (step S105). Then, the output unit 160 outputs the search result to the display unit 170 (Step S106). At this time, as shown in FIG. 8, the output unit 160 displays the search results from the heading part, the explanation part, and the example part on the display part 170 in headword unit, explanation part unit, and example part unit, respectively. As described above, the full-text search processing when the number of search character strings is plural is completed.

  Third, the flow of the fuzzy search process will be described with reference to the flowchart shown in FIG. When the user operates the keyboard 100i to input a search keyword and selects the fuzzy search mode, the fuzzy search unit 143 starts the fuzzy search processing shown in FIG.

  When the fuzzy search process starts, the fuzzy search unit 143 generates a search character string from the input character string input as a search keyword (step S121). More specifically, the fuzzy search unit 143 generates a search character string by adding delimiters to both ends of the input character string. For example, when the character string “saled” is input as a search keyword, the fuzzy search unit 143 adds “•” to the beginning and end of the search keyword “salaed”, so that Generate a search string.

  After generating the search character string, the fuzzy search unit 143 extracts a monogram and a bigram from the search character string (Step S122). More specifically, assuming that the verification character string is M characters, the fuzzy search unit 143 uses the verification character string to include two bigrams each including delimiters at both ends and no delimiter (M-2 ) Extract monograms. For example, from the search string ".salaed.", Two bigrams of ".s" and "d." At both ends and "s", "a", "l", "a", " e ”and six monograms“ d ”are extracted.

  Next, the fuzzy search unit 143 sets the verification character string as a reference character string (step S123). After setting the reference character string, the fuzzy search unit 143 proceeds to a process of evaluating the continuity of the extracted monogram and bigram for each headword included in the dictionary data.

  First, the fuzzy search unit 143 selects a first headword from a plurality of headwords included in the dictionary data, and derives the appearance range of the headword (step S124). More specifically, the fuzzy search unit 143 refers to a file related to the index number of the search index and files related to the start position of the index part ((7) number.idx and (8) headline.idx shown in FIG. 6). Then, the range in which the headword appears in the dictionary data is specified.

  After deriving the appearance range of the headword, the fuzzy search unit 143 refers to the headline index ((1) pattern_hdl.idx and (2) position_hdl.idx shown in FIG. 5) and extracts each of the extracted character strings from the reference character string. The appearance positions of the monogram and the bigram are read (step S125).

  When the appearance position is read, the fuzzy search unit 143 calculates the number of consecutive matching N-grams based on the appearance position of each monogram and bigram existing within the appearance range of the headword text derived in step S124 (step S126). . More specifically, for example, when the search character string is “· salaed ·”, the fuzzy search unit 143 uses “· s”, “s”, “a”, “l”, It is determined whether eight monograms or bigrams of “a”, “e”, “d”, and “d ·” appear one character at a time in this order in the headword text. Then, the fuzzy search unit 143 calculates the number of monograms and bigrams whose appearance positions appear consecutively as the number of consecutive matching N-grams.

  After calculating the number of consecutive matching N-grams, the fuzzy search unit 143 generates a frequency distribution of the number of consecutive matching N-grams based on the calculated number of consecutive matching N-grams (step S127). Then, the fuzzy search unit 143 calculates an fuzzy search evaluation value from the generated frequency distribution (Step S128). As a method of calculating the fuzzy search evaluation value, the method disclosed in Patent Document 2 can be used. More specifically, the fuzzy search unit 143 refers to the generated frequency distribution and calculates a larger value as the number of consecutive matching N-grams is larger as the fuzzy search evaluation value. In addition, the fuzzy search unit 143 calculates, as the fuzzy search evaluation value, a larger value as the frequency of appearance of successively matching N-grams is higher, as the fuzzy search evaluation value.

  Next, the evaluation unit 150 calculates a ranking evaluation value (Step S129). Then, the fuzzy search unit 143 determines whether an unevaluated headword exists (step S130). If it is determined that there is an unevaluated headword (step S130; YES), the fuzzy search unit 143 returns the process to step S124, and performs the same processing for the unevaluated headword.

  Finally, when it is determined that there is no unevaluated headword (step S130; NO), the output unit 160 rearranges the search results based on the fuzzy search evaluation value and the ranking evaluation value (step S131). ). Then, the output unit 160 outputs the search result to the display unit 170, for example, as illustrated in FIG. 9 (Step S132). Thus, the fuzzy search process ends.

  As described above, the search device 100 according to the present embodiment stores a search index that associates a plurality of N-grams included in dictionary data, which is a search target electronic document, with their appearance positions in the dictionary data. Are generated separately for the text belonging to the headword category and the text belonging to the body category. Then, at the time of the incremental search and the ambiguous search only for the headword text in the dictionary data, the search device 100 searches the search keyword by referring only to the heading index, and targets the entire dictionary data. In a full-text search, a search keyword is searched by referring to the entire search index.

  In other words, the search device 100 generates an index by summarizing the appearance positions of the N-grams for each category, and generates the index not in the entire dictionary data but in the headword category unit in the incremental search and the ambiguous search. Search using an index with a small data size. As a result, the search efficiency at the time of the incremental search and the ambiguous search is improved, so that the search time can be shortened and a desired electronic document can be efficiently searched.

  In particular, since the incremental search is executed every time a search keyword is input one character at a time, by shortening the search time by the incremental search, it is possible to greatly reduce the processing time of the entire search process including the full-text search. it can. Further, since the ambiguous search is a search having a relatively long processing time, the effect of reducing the search time is great.

(Modification)
Although the embodiment of the present invention has been described above, the above embodiment is an example, and the scope of the present invention is not limited to this. That is, the embodiments of the present invention can be applied in various ways, and all embodiments are included in the scope of the present invention.

  For example, in the above-described embodiment, the search device 100 includes the generation unit 120 that generates a search index. However, in the present invention, the search device 100 may not have a function of generating a search index. If the search index generated by the external device is acquired in advance and the acquired search keyword is stored in the data storage unit 110, the search device 100 can execute the above-described search processing. For example, at the time of factory shipment, search indexes corresponding to a plurality of dictionary data mounted on the search device 100 may be stored in the data storage unit 110.

  In the above embodiment, dictionary data has been described as an example of an electronic document to be searched. However, the electronic document to be searched is not limited to a dictionary, and may be any document including text classified into a plurality of categories. For example, the electronic document to be searched may be a patent specification including texts classified into categories such as "Title of Invention" and "Claims". Alternatively, in the electronic document to be searched, a category (function name category) into which a text representing a name of a function of a certain product is classified and a category (operation) into which a text representing an operation method for using the function is classified (Method category).

  In the above-described embodiment, the search device 100 includes the incremental search unit 141 and the fuzzy search unit 143 as the first search unit, and includes the full-text search unit 142 as the second search unit. However, in the present invention, the search device 100 may include only one of the incremental search and the fuzzy search as the first search means. Alternatively, the search device 100 may include not only these searches but also other types of search functions as the first search means or the second search means. Further, a specific method, particularly a ranking method, in an incremental search (a forward match search), a full-text search (a perfect match search), and an ambiguous search (a partial match search) can be arbitrarily set. Further, in the above embodiment, a monogram and a bigram are used as the Ngram, but another Ngram such as a trigram may be used.

  In the above embodiment, in the search device 100, the CPU 100a functions as each of the generation unit 120, the acquisition unit 130, the search unit 140, the evaluation unit 150, and the output unit 160 by executing the program stored in the ROM 100b. However, in the present invention, the search apparatus 100 includes dedicated hardware such as an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or various control circuits instead of the CPU 100a. The hardware may function as each of the generation unit 120, the acquisition unit 130, the search unit 140, the evaluation unit 150, and the output unit 160. In this case, each function of each unit may be realized by individual hardware, or the function of each unit may be realized by a single piece of hardware. Further, a part of the function of each unit may be realized by dedicated hardware, and the other part may be realized by software or firmware.

  It should be noted that the present invention can be provided not only as a search device having a configuration for realizing the function according to the present invention but also as a search device according to the present invention by applying a program to an existing personal computer or information terminal device. You can also. That is, by applying a text search program for realizing each functional configuration by the search device 100 described in the above embodiment so that a CPU or the like for controlling an existing personal computer, information terminal device, or the like can execute the text search program. It can function as the search device 100 according to the invention. Further, the search method according to the present invention can be implemented using the search device 100.

  The method of applying such a program is arbitrary. The program can be stored in a computer-readable recording medium (Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD), Magneto Optical disc (MO), etc.), and applied, for example, to the Internet, etc. It is also possible to store the program in a storage on the network and download the program to apply. Further, a configuration is also possible in which part or all of the data required for the above processing such as dictionary data, relocation data, and search index is stored in an external server, and these data are obtained using a communication function to execute the above processing. It is possible.

  As described above, the preferred embodiments of the present invention have been described, but the present invention is not limited to the specific embodiments, and the present invention includes the invention described in the claims and the equivalents thereof. included. Hereinafter, the invention described in the claims of the present application is additionally described.

(Appendix 1)
A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. Storage means for storing the second index;
Acquisition means for acquiring a search keyword;
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. First search means;
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. Second search means for performing
Output means for outputting a search result by at least one of the first search means and the second search means;
A search device comprising:

(Appendix 2)
The acquiring means acquires the search keyword one character at a time,
The first search means, every time the search keyword is obtained one character at a time by the obtaining means, outputs a character string obtained by adding at least one character newly obtained to the at least one character already obtained by the obtaining means. Search as the search keyword,
3. The search device according to claim 1, wherein

(Appendix 3)
The first search means reads out appearance positions of a plurality of N-grams included in the search keyword from the first index, and the read out appearance position is continuous with positions of the plurality of N-grams in the search keyword. Searching for the search keyword from texts belonging to the first category based on the number of matches.
3. The search device according to claim 1 or 2, wherein

(Appendix 4)
The first search means searches the text belonging to the first category for a character string starting with the search keyword with reference to the first index,
The second search means searches the electronic document for a character string including the search keyword with reference to the first index and the second index;
4. The search device according to any one of supplementary notes 1 to 3, wherein:

(Appendix 5)
Generating the first index by associating the plurality of N-grams included in the text belonging to the first category with the appearance positions of the plurality of N-grams in the text belonging to the first category; Generating the second index by associating a plurality of N-grams included in the text belonging to the second category with the appearance positions of the plurality of N-grams in the text belonging to the second category Means, further comprising:
The storage unit stores the first index and the second index generated by the generation unit,
5. The search device according to any one of supplementary notes 1 to 4, wherein:

(Appendix 6)
The electronic document is dictionary data,
The text belonging to the first category is a text representing a headword in the dictionary data,
The text belonging to the second category is text representing an explanation or an example of the headword,
6. The search device according to any one of supplementary notes 1 to 5, characterized in that:

(Appendix 7)
A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. An index obtaining step of obtaining an index of 2;
A search keyword obtaining step for obtaining a search keyword;
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. A first search step;
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. A second search step,
An output step of outputting a search result by at least one of the first search step and the second search step;
Search method including.

(Appendix 8)
Computer
A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. Index acquisition means for acquiring the second index;
Search keyword acquisition means for acquiring a search keyword,
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. A first search means,
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. A second search means,
Output means for outputting a search result by at least one of the first search means and the second search means;
Program to function as

100 search device, 100a CPU, 100b ROM, 100c RAM, 100d hard disk, 100e media controller, 100g video card, 100h LCD, 100i keyboard, 100j speaker, 110 data storage unit, 120 ... Generating unit, 121 ... Relocation data generating unit, 122 ... Search index generating unit, 130 ... Acquisition unit, 140 ... Search unit, 141 ... Incremental search unit, 142 ... Full-text search unit, 143 ... Fuzzy search unit, 150 ... Evaluation Unit, 160: output unit, 170: display unit

Claims (8)

A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. Storage means for storing the second index;
Acquisition means for acquiring a search keyword;
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. First search means;
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. Second search means for performing
Output means for outputting a search result by at least one of the first search means and the second search means;
A search device comprising: The acquiring means acquires the search keyword one character at a time,
The first search means, every time the search keyword is obtained one character at a time by the obtaining means, outputs a character string obtained by adding at least one character newly obtained to the at least one character already obtained by the obtaining means. Search as the search keyword,
The search device according to claim 1, wherein: The first search means reads out appearance positions of a plurality of N-grams included in the search keyword from the first index, and the read out appearance position is continuous with positions of the plurality of N-grams in the search keyword. Searching for the search keyword from texts belonging to the first category based on the number of matches.
The search device according to claim 1, wherein The first search means searches the text belonging to the first category for a character string starting with the search keyword with reference to the first index,
The second search means searches the electronic document for a character string including the search keyword with reference to the first index and the second index;
The retrieval device according to claim 1, wherein: Generating the first index by associating the plurality of N-grams included in the text belonging to the first category with the appearance positions of the plurality of N-grams in the text belonging to the first category; Generating the second index by associating a plurality of N-grams included in the text belonging to the second category with the appearance positions of the plurality of N-grams in the text belonging to the second category Means, further comprising:
The storage unit stores the first index and the second index generated by the generation unit,
The search device according to any one of claims 1 to 4, wherein: The electronic document is dictionary data,
The text belonging to the first category is a text representing a headword in the dictionary data,
The text belonging to the second category is text representing an explanation or an example of the headword,
The search device according to any one of claims 1 to 5, wherein: Computer
A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. An index obtaining step of obtaining an index of 2;
A search keyword obtaining step for obtaining a search keyword;
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. A first search step;
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. A second search step,
An output step of outputting a search result by at least one of the first search step and the second search step;
Search method to perform . Computer
A first N-gram in which a plurality of N-grams included in the text belonging to the first category in the electronic document to be searched are associated with the appearance positions of the plurality of N-grams in the text belonging to the first category; An index, and a plurality of N-grams included in the text belonging to the second category of the electronic document, and an appearance position of the plurality of N-grams in the text belonging to the second category. Index acquisition means for acquiring the second index;
Search keyword acquisition means for acquiring a search keyword,
The search keywords are retrieved from the text belonging to the first category by reading the appearance positions of the plurality of N-grams included in the search keywords from the first index and evaluating the continuity of the read appearance positions. A first search means,
The search keywords are retrieved from the electronic document by reading the appearance positions of a plurality of N-grams included in the search keywords from the first index and the second index and evaluating the continuity of the read appearance positions. A second search means,
Output means for outputting a search result by at least one of the first search means and the second search means;
Program to function as

Images