JP2017117109A - Information processing device, information processing system, information retrieval method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To discriminate between a word phrase of a different meaning having a spelling similar to a keyword and a clerical error of the keyword or a word phrase of notation inconsistency.SOLUTION: An information processing device includes a keyword acquisition part for acquiring a keyword, a first retrieval part for retrieving similar words in which a character string is similar to the keyword from data of a retrieval object, and a second retrieval part for retrieving candidates for words which are estimated to be a clerical error of the keyword or notation inconsistency from among the similar words on the basis of the appearance frequency of the keyword and the appearance frequency of the similar words in the data of the retrieval object.SELECTED DRAWING: Figure 3

Description

  The present invention relates to an information processing apparatus, an information processing system, an information search method, and a program.

  2. Description of the Related Art Conventionally, there is a demand for a technique for efficiently retrieving what a searcher requests from a large amount of language information and text information stored in an electronic medium or the like. In order to perform such a search, a transposed index (transposed file) is created by reading linguistic information and text information, performing morphological analysis, etc., and using words and phrases appearing in the linguistic information and text information as index words. A technique for searching language information and text information using a transposed index for a keyword input by a searcher is known.

  Further, similar words of the input keyword are detected by using the similarity between the indexed character string and the input keyword (matching degree of N-gram, editing distance, and other feature information extracted from the character string). By searching with the input keyword as typographical error or notation, the keyword entered by the searcher, or the stored language information or text information can be prevented from being missed even if the typographical error or notation is shaken. A technique is known (see, for example, Patent Document 1).

  However, the conventional technique for detecting similar words has a problem that not only erroneous writing or shaking of notation but also a character string having another meaning that is simply similar in spelling is detected as a similar word.

  Accordingly, it is an object of the present invention to provide a technique capable of discriminating a word having another meaning similar in spelling with a keyword and a word having a misprinted keyword or a wobbling word.

  In the information processing apparatus, a keyword acquisition unit that acquires a keyword, a first search unit that searches for similar words in which the keyword and a character string are similar, from search target data, and the search target data A second search unit for searching for a candidate word that is presumed to be an error in the keyword or a fluctuation of the notation from the similar words based on the number of appearances of the keywords and the number of appearances of the similar words; .

  According to the disclosed technology, it is possible to discriminate between another meaning word / phrase similar in spelling to the keyword and a keyword error / notation fluctuation word / phrase.

It is a figure which shows the operation | use form of the information search system which concerns on embodiment. It is a figure which shows the hardware constitutions of the information search device which concerns on embodiment. It is a functional block diagram which shows the function structure of the information search device which concerns on embodiment. It is a figure which shows an example of word feature information. It is a figure which shows an example of the sequence of an information search process in an information search system. It is a figure which shows the example of an input screen of a keyword. It is a figure which shows the example of a screen at the time of selecting the keyword for a search from a candidate. It is a figure which shows the example of a display screen at the time of requesting a search with the selected keyword for a search. It is a figure which shows the example of a display screen of a search result. It is a flowchart of a keyword candidate search process. It is a conceptual diagram of the process which estimates formal notation from the input error etc.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  In the present embodiment, an information search system including an information search device for searching for patent documents will be described as an example.

  FIG. 1 is a diagram illustrating an example of an operation mode of the information search system according to the present embodiment. As illustrated in FIG. 1, the information search system according to the present embodiment includes an information search device 1, a client device 2, and a target information DB 200. The client device 2 is configured by a general information processing device such as a PC (Personal Computer). The information retrieval apparatus 1 is connected to the client apparatus 2 via a network, and is operated as a server that retrieves document information stored in the target information DB 200 in response to a retrieval request from the client apparatus 2.

  The target information DB 200 stores patent document information as search target information. That is, the search target information according to the present embodiment is patent document information stored in the target information DB 200. As shown in FIG. 1, in this embodiment, an example in which the target information DB 200 is provided separately from the information search apparatus 1 will be described. However, the target information DB 200 may be configured inside the information search apparatus 1. Is possible. The target information DB 200 is configured by a nonvolatile storage medium such as an HDD.

  Next, a hardware configuration of the information search apparatus 1 according to the present embodiment will be described. FIG. 2 is a block diagram illustrating a hardware configuration of the information search apparatus 1 according to the present embodiment. As shown in FIG. 2, the information search apparatus 1 according to the present embodiment has the same configuration as an information processing terminal such as a general server or a PC (Personal Computer). That is, the information search apparatus 1 according to the present embodiment includes a CPU (Central Processing Unit) 10, a RAM (Random Access Memory) 20, a ROM (Read Only Memory) 30, an HDD (Hard Disk Drive) 40, and an I / F 50. 80 is connected. Further, an LCD (Liquid Crystal Display) 60 and an operation unit 70 are connected to the I / F 50.

  The CPU 10 is a calculation means and controls the operation of the entire information retrieval apparatus 1. The RAM 20 is a volatile storage medium capable of reading and writing information at high speed, and is used as a work area when the CPU 10 processes information. The ROM 30 is a read-only nonvolatile storage medium and stores a program such as firmware. The HDD 40 is a non-volatile storage medium that can read and write information, and stores an OS (Operating System), various control programs, application programs, and the like.

  The I / F 50 connects and controls the bus 80 and various hardware and networks. The I / F 50 includes, for example, an Ethernet (registered trademark) connection interface and a USB (Universal Serial Bus) connection interface. The LCD 60 is a visual user interface for the user to check the state of the information search device 1. The operation unit 70 is a user interface such as a keyboard and a mouse for the user to input information to the information search apparatus 1. As described with reference to FIG. 1, the information search apparatus 1 according to the present embodiment is operated as a server. Therefore, user interfaces such as the LCD 60 and the operation unit 70 can be omitted.

  In such a hardware configuration, a program stored in a storage medium such as the ROM 30, the HDD 40, or an optical disk (not shown) is read into the RAM 20, and operates according to the control of the CPU 10, thereby configuring a software control unit. A functional block that realizes the function of the information search apparatus 1 according to the present embodiment is configured by a combination of the software control unit configured as described above and hardware.

  Next, functional blocks of the information search apparatus 1 according to the present embodiment will be described with reference to FIG. FIG. 3 is a block diagram showing a target information DB 200 that stores functional blocks of the information search apparatus 1 according to the present embodiment and document information to be searched by the information search apparatus 1. As shown in FIG. 3, the information search apparatus 1 according to the present embodiment includes a communication unit 101, a storage unit 102, an index creation unit 103, a word feature extraction unit 104, a similar phrase search unit 105, a keyword candidate search unit 106, a search A control unit 107 and a full-text search unit 108 are included. Each of these functional units is realized by a program loaded in the RAM 20 shown in FIG.

  The communication unit 101 communicates with the client device 2 and the target information DB 200.

  The storage unit 102 stores data such as word feature information 102A described later.

  The index creation unit 103 creates index information for each word included in the search target phrase data stored in the target information DB 200 and stores the index information in the storage unit 102.

  The word feature extraction unit 104 generates word feature information 102 </ b> A for each word for which index information has been created by the index creation unit 103 and stores it in the storage unit 102. FIG. 4 is a diagram illustrating an example of the word feature information 102A. The word feature information 102A includes data items such as a word, N-gram, character string length, and number of appearances. The N-gram is data obtained by dividing each word by a predetermined number of characters. For example, N-grams of two words “stationary” and “stationery” (when N = 2) are as follows.

stationary: st, ta, at, ti, io, on, na, ar, ry
stationery: st, ta, at, ti, io, on, ne, er, ry
The character string length indicates the number of characters of each word. The number of appearances indicates the number or frequency at which each word is included in the search target data stored in the target information DB 200.

  The similar phrase search unit 105, for example, sets “N-grams” of a plurality of partial character strings constituting each word included in the keyword input by the searcher and “N” of each word stored in the word feature information 102A. Based on the “gram” information, similar phrases are extracted.

  The similar word search unit 105 and the keyword candidate search unit 106 search for keyword candidates that are presumed to be erroneously written or fluctuated in the specified keyword from the words stored in the target information DB 200. I do.

  In the keyword candidate search process, the similar phrase search unit 105 searches for similar words having similar keywords and character strings from search target data stored in the target information DB 200. The similar phrase search unit 105 calculates, for example, the degree of coincidence between the N gram of the keyword and the N gram of each word in the word feature information 102A, and searches the target information DB 200 for a phrase including a word having a degree of coincidence of a predetermined threshold or more. , And extracted as a similar phrase (similar word). For example, if there are five N-grams of keywords and four N-grams match, the degree of coincidence is calculated as 4/5 = 80%. Note that the similar phrase search unit 105 may be configured to extract similar phrases using other well-known methods.

  In the keyword candidate search process, the keyword candidate search unit 106 uses the similar phrase search unit 105 based on the number of occurrences of the keyword in the search target data and the number of similar words searched by the similar phrase search unit 105. From the extracted words and phrases, a keyword candidate that is a word candidate that is estimated to be a misprint of the keyword or a fluctuation of the notation is searched.

  For example, when the ratio between the number of occurrences of a keyword and the number of appearances of a similar word is within a predetermined range, the keyword candidate search unit 106 determines that the similar word is erroneously written in the keyword or shakes the notation. to decide. In addition, the keyword candidate search unit 106 searches for a word that is a misprint of the keyword or a fluctuation of the notation based on the edit distance between the keyword and the similar word with a predetermined weight added to the first character. In addition, when the keyword includes a plurality of words, the keyword candidate search unit 106 uses a keyword error or notation based on the matching rate of the edit distance between each word included in the keyword and each word included in the similar word. Search for words that are shaking. A detailed example of keyword candidate search processing will be described later.

  When the search control unit 107 receives a keyword from the client device 2 via the communication unit 101, the search control unit 107 uses the keyword and the keyword candidate searched using the similar phrase search unit 105 or the keyword candidate search unit 106 for search. It transmits to the client apparatus 2 as a keyword candidate. Note that the search control unit 107 uses the index information created by the index creation unit 103 to transmit, for example, data on the applicant name of a patent document, including keyword candidates, to the client apparatus 2 as search keyword candidates. It is good also as composition to do.

  When the full-text search unit 108 receives one or more search keyword groups from the client device 2 via the communication unit 101, the full-text search unit 108 searches the target information DB 200 for a document including the keyword group, and the search result is sent to the client device. 2 to send.

  Next, the operation of the information search system will be described with reference to FIG. FIG. 5 is a diagram illustrating an example of a sequence of information search processing in the information search system.

  The client device 2 receives a keyword input operation from the searcher (step S101). The client device 2 transmits a search request including the input keyword to the information search device 1 (step S102).

  The information search apparatus 1 performs a keyword candidate search process for searching for a keyword candidate that is estimated to be a typographical error or notation of the received keyword from the search target data held in the target information DB 200 (step S103). . The information search device 1 transmits a keyword and a phrase including the keyword candidate extracted based on the keyword to the client device 2 as a search keyword candidate (step S104).

  The client apparatus 2 displays the received search keyword candidates to the searcher (step S105), and accepts an operation for selecting one or more keyword groups from the search keyword candidates from the searcher (step S106). . The client device 2 transmits one or more keyword groups selected by the searcher to the information search device 1 (step S107).

  The information search device 1 searches the document held in the target information DB 200 based on the received one or more keywords (step S108), and transmits the search result to the client device 2 (step S109).

  The client device 2 displays the received search result to the searcher (step S110).

  The keyword may be a single word or a phrase including a plurality of words separated by a space or the like.

  Next, an example of a display screen displayed when the client apparatus 2 accepts a keyword input operation from a searcher in step S101 of FIG. 5 will be described with reference to FIG. FIG. 6 is a diagram illustrating an example of a keyword input screen. The searcher inputs a keyword in the keyword input field 501 and presses a “search” button 502. Then, the client device 2 transmits a search request including the input keyword to the information search device 1.

  Next, referring to FIG. 7, the client device 2 displays the received keyword candidates for search in step S105 and step S106 in FIG. 5 to the searcher, and selects one or more keywords from the searcher. An example of a display screen that is displayed when receiving a message will be described. FIG. 7 is a diagram illustrating an example of a selection screen from search keyword candidates. For example, since all search keyword candidates are selected in the default state, they are masked in gray as in 503, and the searcher clicks on a word or phrase excluded from the search keyword candidates. When an operation such as this is performed, the word / phrase is not selected, and a gray mask is removed as in 504. The “search formula” column 505 displays a list of keyword candidates for search that are currently selected.

  Next, an example of a display screen displayed when the client apparatus 2 transmits one or more keywords selected by the searcher to the information search apparatus 1 in step S107 of FIG. 5 will be described with reference to FIG. FIG. 8 is a diagram illustrating an example of a display screen when a search is requested using a selected search keyword. For example, the searcher copies and pastes text data indicating a list of selected search keyword candidates displayed in the “search expression” field 505 in FIG. 7 into a predetermined search condition field, A “search execution” button 507 is pressed. Thereby, the client apparatus 2 transmits one or more keywords selected by the searcher to the information search apparatus 1.

  Next, a display screen example in which the client device 2 displays the received search result to the searcher in step S109 of FIG. 5 will be described with reference to FIG. FIG. 9 shows an example of a search result display screen.

  Next, with reference to FIG. 10, a detailed example of keyword candidate search processing by the keyword candidate search unit 106 for searching for a keyword candidate estimated as a keyword error or notation in step S103 in FIG. 5 will be described. . FIG. 10 is a flowchart of keyword candidate search processing.

  The similar phrase search unit 105 extracts a keyword similar phrase from the search target phrases stored in the target information DB 200 (step S201).

  The keyword candidate search unit 106 narrows down the words extracted by the similar word search unit 105 based on the character string length (step S202). When the keyword includes a plurality of words, the keyword candidate search unit 106 extracts each word separated by a space or the like from the keywords and similar words, and each word included in the keyword and each word Narrowing is performed based on the character string length of each word included in. For example, when at least one of the difference in the character string length and the ratio between the word in the keyword and the word in the phrase corresponding to the word in the keyword is equal to or greater than a predetermined threshold, the keyword candidate search unit 106 selects the word or phrase. Remove from keyword candidates. Thereby, for example, when the word in the keyword is “AAA” and the word in the phrase of the similar word corresponding to the word is “A”, the phrase of the similar word is removed from the keyword candidates.

  Subsequently, the keyword candidate search unit 106 performs narrowing down based on the ratio (relative frequency) of the number of appearances of the keyword and the number of appearances of each phrase of the similar word that has been narrowed down (step S203). For example, if the ratio between the number of occurrences of a keyword and the number of occurrences of each narrowed phrase is not within a predetermined range (for example, 20 or less to 1/20 or more), the keyword candidate search unit 106 selects the word or phrase. Remove from keyword candidates.

  Here, the keyword candidate search unit 106 has a relatively large ratio (relative frequency) between the number of appearances of the keyword and the number of appearances of the other word for another word that is simply spelled similar to the keyword. Alternatively, the swing of the notation uses a feature based on an empirical rule that the ratio between the number of appearances of the keyword and the number of appearances of the misprint or the notation swing is relatively small.

  Returning to FIG. 10, the description of the flowchart of the keyword candidate search process will be continued.

  Subsequently, the keyword candidate search unit 106 narrows down based on the weighted edit distance (step S204). The edit distance is a numerical value for quantifying the degree of similarity between two character strings, and is the minimum number of steps required to transform one character string into the other by insertion / deletion / replacement of characters. .

  The keyword candidate search unit 106 calculates an edit distance between the keyword and each word / phrase of the similar word that has been narrowed down, and if the edit distance is equal to or greater than a predetermined threshold (for example, 4), the keyword / phrase is searched from the keyword candidate. remove. Here, a predetermined weight is added to the editing distance between the keyword and the first character of each phrase. If the first character has been inserted, deleted, replaced, etc., the edit distance of the first character is multiplied by, for example, four times the weight. This is because there is little experience of writing errors or notation with respect to the first character. Also, a predetermined weight may be added for each operation of replacement, insertion, and deletion. Moreover, it is good also as a structure which reduces a weight about a specific character (for example, symbol). In addition, the replacement distance may be calculated with an edit distance of 1. For example, “AICOH” and “AICHO” replace “H” with “O” and “O” with “H”, so the original editing distance is 2, but “H” and “O” ", The edit distance is counted as 1. Thereby, it becomes easy to detect the replacement of characters as an error.

  Subsequently, the keyword candidate search unit 106 narrows down each word included in the keyword and each word included in each word / phrase of the narrowed-down similar word based on the editing distance in units of words (step S205). When the keyword includes a plurality of words, the keyword candidate search unit 106 extracts each word separated by a space or the like from the keyword and each phrase, and includes each word included in the keyword and each phrase. Narrowing is performed based on the word-by-word editing distance of each word. For example, when the matching rate of the edit distance between the word in the keyword and the word in the word corresponding to the word is equal to or less than a predetermined threshold (for example, 50%), the keyword candidate search unit 106 selects the word from the keyword candidate. remove.

  In the above-described example, the narrowing based on the edit distance with high calculation cost is performed after narrowing based on various conditions in the keyword candidate search process. Thereby, keyword candidate search processing can be executed at higher speed.

  Here, an example of keyword candidate search processing will be described using the word feature information shown in FIG. 4 as an example.

  No. of the word feature information shown in FIG. The word “station” in FIG. 7 is also a partial character string of the word “stationary” and has a high N-gram match, so that the similar phrase search unit 105 searches for a similar phrase. However, the keyword candidate search unit 106 can remove keywords from the keyword candidates by narrowing down based on the character string length in step S202 in FIG. 10 or narrowing down based on the edit distance in step S204 in FIG.

  The word feature information No. 1 shown in FIG. 7 words “stationary” and No. 7 The word “stationery” of 8 is different by one character as a character string. Therefore, since the matching degree of N-grams is high, the similar phrase search unit 105 searches for similar phrases. Moreover, it cannot be removed from the keyword candidates by narrowing down based on the character string length in step S202 of FIG. 10 and narrowing down based on the editing distance in step S204 of FIG. However, the keyword candidate search unit 106 uses the narrowing-down based on the relative frequency in step S203 in FIG. No. 7 appearance frequency, and No. 7 It is determined that the ratio (for example, 39/25) with the number of appearances of the word 8 is not within a predetermined range. As a result, it can be determined that it is another word with similar spelling, not an error or shaking of the notation, and can be removed from the keyword candidates.

  The word feature information No. 1 shown in FIG. 8 words “stationery” and The 9 word “stationery” is the only difference between the letter i and o. Therefore, since the matching degree of N-grams is high, the similar phrase search unit 105 searches for similar phrases. Further, the narrowing based on the character string length in step S202 in FIG. 10 and the narrowing based on the editing distance in step S204 in FIG. In addition, the keyword candidate search unit 106 can perform No. 3 even when narrowing down based on the relative frequency in step S203 of FIG. No. 8 appearance frequency and No. 8 It is determined that the ratio (for example, 25/1) with the number of appearances of the word 9 is not within a predetermined range. As a result, it can be determined that it is an error or a shaking of the notation, and can be prevented from being removed from the keyword candidates. Note that the word feature information No. 1 shown in FIG. 1 word “acceleration” and 2 words “accelaration” and No. 2 4 words “commercial” and no. The same applies to the second word “comercial”, and the keyword candidate search unit 106 can determine that it is a typographical error or a shake of the notation and prevent it from being removed from the keyword candidates.

<Summary>
As described above, in the embodiment, when collecting words from large-scale electronic data and extracting the characteristics of the words, large-scale data such as the number of appearances (frequency) of words as well as spelling similarity The statistical information obtained from is also stored as word feature information. Then, after a similar word search based on the similarity of spelling, based on statistical information such as the number of appearances of the word, an erroneously written or spelled word is distinguished from another word that is similar in spelling but different.

  In addition, in the conventional method of creating a correct answer dictionary in advance and searching for a word most similar to the input keyword, for example, in the case of a correct answer dictionary of a word such as a company name or product part number, the search target Every time the company name increases, maintenance work is required to reflect it in the dictionary. On the other hand, in the above-described embodiment, since it is not necessary to create a correct dictionary, such trouble does not occur.

  In the above-described embodiment, in the search target data, most of the search target words are officially written, and in rare cases, there is a typographical error or a fluctuation of the typographical expression, so that the typographical error or the fluctuation of the typographical expression is particularly accurately searched. it can. Therefore, this embodiment is applicable not only to the name of the applicant of a patent document, but also to an apparatus that searches for misprints or fluctuations in notations of various types of words such as product model numbers, person names, and place names.

<Modification>
The keyword candidate search unit 106 may be configured to change the range of relative frequencies determined as keyword candidates according to the number of appearances of the keyword. In that case, for example, the following configuration may be adopted. The keyword candidate search unit 106 determines that the ratio between the number of appearances of a keyword and the number of appearances of similar words is within a first range (for example, 20 or less to 1 when the number of appearances of the keyword is equal to or greater than a predetermined threshold (for example, 100). / 20 or more), it is determined that the similar word is a keyword error or a notation fluctuation. When the keyword appearance count is less than the predetermined threshold, the keyword candidate search unit 106 has a second range (for example, a ratio between the keyword appearance count and the similar word appearance count wider than the first range). , 50 or less to 1/50 or more), it is determined that the similar word is a keyword error or a notation fluctuation. As a result, for example, in the case of keywords such as small and medium-sized enterprises that have a small number of applications, the number of appearances in the applicant's name of patent documents is small, so if the detection accuracy is relatively low, a wider range of similar terms It can be determined that the keyword is erroneously written or the notation is shaken. Accordingly, a broader range of synonyms can be presented to the searcher as candidates for keyword error or notation fluctuation, and search omissions can be suppressed.

  If the data to be searched for keyword candidates is a natural language such as the full text of a patent specification, the data is divided into units of words, terms, etc. by well-known morphological analysis, etc. The process for searching for keyword candidates may be performed.

  When the keyword itself input by the searcher is a miswritten or written word, a keyword with a relatively large number of appearances among the keyword candidates searched in step S203 of FIG. 10 is a formal written word. It can be estimated that. Therefore, the search control unit 107 may be configured to perform the following processing. First, it is determined whether the ratio between the number of appearances of a keyword and the number of appearances of a phrase having the highest number of appearances among the keyword candidates of the keyword is equal to or less than a predetermined threshold (for example, 1/20). If it is equal to or less than the predetermined threshold, the client device 2 is notified of the most frequently appearing phrase among the keyword candidates of the keyword as a formal expression of the keyword via the communication unit 101 and displayed to the searcher. . Thereby, it is possible to point out an error or the like to the searcher.

  Next, with reference to FIG. 11, the concept of processing for estimating a word with a formal notation when the keyword itself input by the searcher is a miswritten or written word will be described. FIG. 11 is a conceptual diagram of processing for estimating a formal notation from an input error or the like. FIG. 11 shows an image with the vertical axis representing the number of appearances of each character string on a plane representing the distance of the similarity of each character string, the distance being smaller as each character string is more similar. Here, when the keyword input by the searcher is “AICHO”, among the keyword candidates searched by the similar phrase search unit 105 and the keyword candidate search unit 106, “AICOH” which is the phrase having the highest appearance frequency is Presumed to be an official notation and displayed (suggested) to the searcher.

  The search control unit 107 may be configured to select keyword candidates based on feedback statistical information on search results of keyword candidates from the searcher. In that case, for example, the following configuration may be adopted. The search control unit 107 is selected by the searcher received in step S107 of FIG. 5 in association with the keyword received in step S102 of FIG. 5 and the search keyword candidate transmitted in step S104 of FIG. One or more keyword groups are stored. In step S104 of FIG. 5, the search control unit 107 removes the search keyword candidates that have not been selected continuously for a predetermined number of times, for example, and transmits them to the searcher.

  The information search device 1 may be realized by cloud computing using an information processing system including one or more information processing devices (computers), for example. For example, the index creation unit 103, the word feature extraction unit 104, the full-text search unit 108, and the like may be realized by separate computers. Further, the information search device 1 and the target information DB may be configured as an integrated device. The information search device 1 and the client device 2 may be configured as an integrated device.

  Although the preferred embodiment of the present invention has been described in detail above, the present invention is not limited to the specific embodiment, and various modifications, within the scope of the gist of the present invention described in the claims, It can be changed. Moreover, it is also possible to combine a part or all of each embodiment mentioned above.

1 Information retrieval device (an example of "information processing device")
101 communication unit 102 storage unit 103 index creation unit 104 word feature extraction unit 105 similar phrase search unit (an example of “first search unit”)
106 Keyword candidate search unit (an example of “second search unit”)
107 Search control unit (an example of “keyword acquisition unit”)
108 Full-text search unit 2 Client device 200 Target information DB

JP 2005-011078 A

Claims (8)

A keyword acquisition unit for acquiring keywords;
A first search unit for searching for similar words having similar character strings to the keyword from data to be searched;
Based on the number of appearances of the keyword and the number of appearances of the similar word in the search target data, search for candidate words that are presumed to be erroneously written or fluctuated in the keyword from the similar words. A second search unit;
An information processing apparatus comprising: When the ratio between the number of appearances of the keyword and the number of appearances of the similar word is within a predetermined range, the second search unit is an error in writing the keyword or shaking the notation of the keyword. To judge,
The information processing apparatus according to claim 1. The second search unit is
If the number of appearances of the keyword is equal to or greater than a predetermined threshold, if the ratio between the number of appearances of the keyword and the number of appearances of the similar word is within the first range, the similar word is mistakenly written for the keyword. Or judge that it is shaking of the notation,
When the number of appearances of the keyword is less than the predetermined threshold, the ratio between the number of appearances of the keyword and the number of appearances of the similar word is within a second range wider than the first range, Determining that the similar word is a typographical error or a shaking of the notation of the keyword,
The information processing apparatus according to claim 1 or 2. The second search unit further searches for a word that is a typographical error or a notation of the keyword based on an edit distance between the keyword and the similar word with a predetermined weight added to the first character. To
The information processing apparatus according to any one of claims 1 to 3. The second search unit further searches for a word that is a typographical error or a notation of the keyword based on a matching rate of each word included in the keyword and an edit distance of each word included in the similar word. To
The information processing apparatus according to any one of claims 1 to 4. An information processing system including one or more information processing devices,
A keyword acquisition unit for acquiring keywords;
A first search unit for searching for similar words having similar character strings to the keyword from data to be searched;
Based on the number of appearances of the keyword and the number of appearances of the similar word in the search target data, search for candidate words that are presumed to be erroneously written or fluctuated in the keyword from the similar words. A second search unit;
An information processing system comprising: Information processing device
Obtaining a keyword;
Searching for similar words in which the keyword and the character string are similar from search target data;
Based on the number of appearances of the keyword and the number of appearances of the similar word in the search target data, search for candidate words that are presumed to be erroneously written or fluctuated in the keyword from the similar words. Steps,
Execute the information retrieval method. On the computer,
Obtaining a keyword;
Searching for similar words in which the keyword and the character string are similar from search target data;
Based on the number of appearances of the keyword and the number of appearances of the similar word in the search target data, search for candidate words that are presumed to be erroneously written or fluctuated in the keyword from the similar words. Steps,
A program that executes

Images