JP2000331027A - Similar document search device and similar document search method - Google Patents

Abstract

(57) [Summary] [Problem] To realize a similar document search focused on the content of each item when an item of different content exists in a search key document. A document in a unit of an item is cut out from a search key document and a search target document, and a similarity between the search key document and the search target document is calculated in the unit of the item using a vector space method or the like. Based on the result, a search result (for example, a document ID) of a similar document is determined and output.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a similar document search apparatus and a similar document search method for automatically searching a document similar to the key document from a plurality of search target documents, using an arbitrary document as a search key.

[0002]

2. Description of the Related Art In recent years, a large amount of digitized document data has been distributed, and it has become important to classify such a large amount of document data according to a certain rule to enhance the usability. In order to classify document data, there is a similar document search apparatus that uses a certain document as a search key and extracts a document whose content is similar to the document from a search target document database.

[0003] This similar document search apparatus uses words constituting a document as a search key (hereinafter, referred to as a search key document) and individual documents (hereinafter, referred to as a search key document database) in a search target document database.
The search document is compared with the words constituting the search document, and the similarity between the search key document and each search target document is calculated based on the comparison result, and the similarity is calculated from the plurality of search target documents based on the similarity. Is being extracted. As a method of calculating the similarity based on the word comparison result, there is a method of calculating the similarity between the search key document and each search target document by a vector space method from the type, the number of appearances, the appearance place, etc. of the word commonly appearing. is there.

[0004]

In the conventional similar document search method, similar documents are extracted from similarities between documents, that is, similarities between the entire search key document and the entire search target document. If one document in which the contents are described is used as a search key, a search target document including the plurality of contents at the same time is extracted as a similar document. This means that there is a limitation that a document similar to each content included in the search key document, in other words, a document similar to only one content of the search key document cannot be searched.

[0005] The present invention is to solve such a problem. A similar document which can search for a document similar to a search key document from a plurality of search target documents in units of items constituting the document. A document search device and a similar document search method are provided.

That is, the present invention provides a similar document search apparatus and a similar document search method capable of performing a similar document search focused on the content of each item when an item having different contents exists in the search key document. The purpose is to provide.

[0007]

In order to achieve the above object, a similar document search apparatus according to the present invention provides a similar document search apparatus for searching a document similar to a search key document from a plurality of search target documents. An item extracting unit for extracting a document in an item unit from the search key document and the search target document, and calculating a similarity between the search key document and the search target document in the unit of the extracted item, and calculating the same. Calculating means for outputting a similar document search result based on the result.

According to the present invention, the similarity between a search key document and a search target document is obtained in units of items constituting each document. It is possible to perform similar document search focused on the content.

[0009] Further, a means for setting a priority for each item is added, and the similarity of the retrieval key document and the retrieval target document in item units is calculated in consideration of the set priority. For example, when a total sum of similarities of all items is obtained for each search target document and the result is output as a search result, a more optimal similar document search in which a similarity of an important part of the document is weighted is realized. Can be.

Further, means for unifying the sentence structure of each item of the search key document or the document to be searched extracted by the item extracting means is further added, and the search key is unified for each item having such unified sentence structure. By configuring the similarity between the document and the search target document on an item-by-item basis, the degree to which the difference in the sentence structure of each item affects the similarity can be reduced. Relevance can be increased.

[0011]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows a hardware configuration of a similar document search apparatus according to this embodiment. As shown in FIG. 1, the similar document search device includes a control device 1 including a CPU and a memory, an input device 2 such as a keyboard, a display device 3 for displaying similar search results, and an external device for storing search data and the like. It is constructed in a computer system environment such as the storage device 4.

FIG. 2 is a block diagram showing the configuration of the control unit of the similar document search apparatus by function. As described above, the control unit of the similar document search device includes the processing unit 11 and the memory unit 12.

The control section 11 is a section for executing calculations for various controls and processes. The main processing section 200, an initialization section 201, an input section 202, an output section 203, a search target document reading section 204, a search target Document item extraction unit 20
5. Search target document item generation unit 206, search target document word extraction unit 207, search target word appearance frequency calculation unit 208, search target word information calculation unit 209, search key document input unit 21
0, search key document item priority input unit 211, search key document item cutout unit 212, search key document item generation unit 21
3, a search key word extraction unit 214, a search key word appearance frequency calculation unit 215, a search condition setting unit 216, a common word extraction unit 217, a similarity calculation unit 218, a search result output unit 219, and the like.

The memory unit 12 includes a search condition setting buffer unit 229, a search target document storage buffer unit 230, a search target document item storage buffer unit 231, a search target document item generation storage buffer unit 232, and a search target word information storage buffer. Section 233, search key document item priority buffer section 2
34, search key document storage buffer 235, search key document item storage buffer 236, search key document item generation storage buffer 237, search key word information storage buffer 238, common word information storage buffer 239, similarity storage buffer Unit 240, search result output buffer unit 241,
It comprises a work buffer unit 242 and the like.

The details of each section of the processing section 11 will be described.

The initialization section 201 initializes each of the buffer sections 229 to 242 in the memory section 12.

The input unit 202 processes input of information for various settings from a user and a document serving as a search key.

The output unit 203 performs a process of displaying a search key document, a similar search result, various setting information, and the like to the user through the display device 3.

The search target document reading unit 204 converts the search target documents stored in the external storage device 4 into a database and stores them in the search target document storage buffer unit 230.

The search target document item cutout unit 205 analyzes the sentence structure of the search target document stored in the search target document storage buffer unit 230 and, based on the analysis result, extracts a document of a fixed item from the search target document. It is cut out and stored in the search target document item storage buffer unit 231.

The search target document item generation unit 206 regenerates the document of the item stored in the search target document item storage buffer unit 231 as necessary, and replaces the regenerated item document with the search target document item. The data is stored in the generation storage buffer unit 232.

The search target document word extraction unit 207 cuts out a word from a document stored in the search target document storage buffer unit 230 or the search target document item storage buffer unit 231, and selects the document (from the cut-out word group). Or a key word to represent the content of
The extracted words are stored in the search target word information storage buffer unit 233.
To be stored.

The search target word appearance frequency calculation unit 208 stores the key words extracted by the search target document word extraction unit 207 in the search target document item storage unit 231 or the search target document item generation storage buffer unit 232. The appearance frequency in the search target document is calculated for each word type, and the result is stored in the search target word information storage buffer unit 233.

The search target word information calculation unit 209 stores the appearance frequency of each word type in the search target document stored in the search target word information storage buffer unit 233 in the external storage device 4.

The search key document input unit 210 is connected to the input unit 20
2 stores the document of the search key input by the user in the search key document storage buffer unit 235.

The search key document item priority input unit 211
The priority of each item of the search key document is set in the search key document item priority buffer unit 234 by input from the user.

The search key document item cutout unit 212 analyzes the sentence structure of the search key document stored in the search key document storage buffer unit 235, and cuts out a document of fixed items from the search key document based on the analysis result. Then, it is stored in the search key document item storage buffer unit 236.

The search key document item generation unit 213 regenerates the document of the item stored in the search key document item storage buffer unit 236 as necessary, and converts the regenerated item document into the search key document. Item generation storage buffer unit 237
To be stored.

The search key document word extraction unit 214 cuts out words from the search key document stored in the search key document item storage buffer unit 236 or the search key document item generation storage buffer unit 237, and outputs the words from the extracted word group. , A key word representing the contents of the search key document (or item) is extracted, and the extracted word is stored in the search key word information storage buffer unit 238.

The search key word appearance frequency calculation unit 215 calculates the key word extracted by the search key word extraction unit 214
The appearance frequency in the search key document stored in the search key document storage buffer unit 238 is calculated for each word type, and the result is stored in the search key word information storage buffer unit 238.

The search condition setting section 216 sets search conditions for calculating similar documents based on input from a user.
The search conditions include the presence / absence of a process of generating a search key document item, the presence / absence of a process of extracting a search target document item, and the presence / absence of a process of generating a search target document item. The setting contents of these search conditions are stored in the search condition setting buffer unit 229.

The common word extracting section 217 extracts a word common to the search key document stored in the search key word information storage buffer section 238 and the search target document stored in the search target word information storage buffer section 233. Then, the common word information and the frequency information for each common word are stored in the common word information storage buffer unit 239.

The similarity calculation unit 218 includes an appearance frequency for each word type in the search key document stored in the search key word information storage buffer unit 238, and a search target stored in the search target word information storage buffer unit 233. Based on the appearance frequency of each word type in the document and the appearance frequency of each common word type stored in the common word information storage buffer unit 239, the similarity between the search key document and each search target document by a vector space method or the like. Is calculated, and the similarity value is stored in the similarity storage buffer unit 240. At this time, when the priority of the search key document item is set in the search key document item priority buffer unit 234, a weight corresponding to the priority is assigned to the similarity value of the corresponding item, and this weight is assigned. The stored similarity value is stored in the similarity storage buffer unit 240 again.

The search result output unit 219 stores a search result in the search result output buffer unit 241 based on the similarity value for each search target document stored in the similarity storage buffer unit 240. Then, the search result output unit 219 outputs the contents of the search result output buffer unit 239 to the display device 3.

Next, the operation of the similar document search apparatus according to this embodiment will be described.

FIGS. 3 and 4 show the flow of a similar document search process according to this embodiment. First, the initialization unit 201 is activated and initialization such as clearing of the memory unit 12 is performed (step 300).

Subsequently, the search condition setting section 216 is activated. The search condition setting unit 216 sets search conditions for calculating a similar document by inputting from the input device 2, and stores the set data in the search condition setting buffer unit 229 (step 301). FIG. 4 shows a setting example of the search mode.
As shown in the figure, the setting items of the search mode include the item generation of the search key document = Yes / No, the item extraction of the search target document = Yes / No, the item generation of the search target document = Yes / No, and the like. The user can select an arbitrary mode for each item. In this example, it is assumed that “Generate item of search key document = Yes” and “Cut out item of search target document = Yes” are selected. The setting information of these search modes is stored in the search condition setting buffer unit 229. FIG. 6 shows a storage mode of the search mode setting information in the search condition setting buffer unit 229.

Next, it is determined whether or not to input the item priority of the search key document (step 302). When the item priority is input, the search key document item priority input section 211 is input.
Is started. The search key document item priority input unit 211 inputs the priority of each item in the search key document from the user via the input unit 202, and stores information on the priority for each item in the search key document item priority buffer unit 234. (Step 303).

FIG. 7 shows a setting screen for setting the importance of each of the five standard items in the search key document. In the figure, the priority (= importance) of each item is represented in 10 levels, the priority of item number 1 is “10”, the priority of item number 2 is “5”, and the priority of item number 3 is The priority of “3” and the item numbers 4 and 5 are set to “1”.

After setting the item priority of the search key document,
Alternatively, if it is determined in step 304 that the item priority of the search key document is not input, the process proceeds to step 304.

In step 304, the retrieval key document input unit 2
10 is activated. When the user inputs a document serving as a search key, the search key document input unit 210 stores the input search key document in the search key document storage buffer unit 235 (step 304). FIG. 8 shows an example of the input search key document. Where

The retrieved search key document has a plurality of fixed items (for example,

).

Thereafter, the retrieval key document item cutout unit 21
2 is activated. Search key document item cutout unit 212
Analyzes the sentence structure of the search key document stored in the search key document storage buffer unit 235, cuts out all the document portions of the standard items from the search key document based on this analysis result, and The document is stored in the search key document item storage buffer unit 236 (step 305).
FIG. 9 shows an item-by-item document cut out from the search key document.

Next, the search key document item generation unit 213
With reference to the search condition setting buffer unit 229, it is checked whether or not “item generation of search key document = Yes” is set here (step 306). If “Generate item of search key document = Yes” is set, the search key document item generation unit 213 regenerates the document of the standard item stored in the search key document item storage buffer unit 236 (step). 307). The regenerated document of the standard item is stored in the search key document item generation storage buffer unit 237.

FIG. 10 shows an example of regenerating the document of the fixed item. Regeneration of the document of the standard item is performed according to a predetermined rule. The regeneration of the document of the standard item is performed for the purpose of absorbing the difference in the sentence structure and the notation between the item documents. For example, as shown in FIG. 10, the description portion of “Claim X” (where X is an integer of 1 or more) included in the item documents after item number 2 is replaced with the item document of item number X.

After the re-creation of the document of such a fixed item, or when “Generation of item of search key document = not set” is set in step 306, then step 3 is executed.
08 is executed.

In step 308, the retrieval key word extraction unit 2
14 is activated. The search key word extracting unit 214 cuts out a word from the search key document stored in the search key document item storage buffer unit 236 or the search key document item generation storage buffer unit 237, and extracts the search key document (or Key words that reflect the content of the item are extracted, and the extracted words are stored in the search key word information storage buffer unit 238. Here, the extraction of the word is performed by morphological analysis or the like, and the key word can be determined based on the part of speech of the word. For example, words such as "noun" and "sa-noun" are determined as key words.
The word information extracted in this way is stored in the search key word information storage buffer unit 238 as shown in FIG.

Subsequently, the search key word appearance frequency calculation unit 21
5 is activated. Search key word appearance frequency calculation unit 215
Is a search key document item generation storage buffer unit 237 or a search key document item storage buffer unit 236 for the word type stored in the search key word information storage buffer unit 238.
1 is calculated in the item document stored in FIG.
As shown in FIG. 7, the result is stored in the search key word information storage buffer unit 238 (step 309). In FIG. 12, "document database = 1" indicates that the word "document database" appears once.

Next, the retrieval target document reading section 204 is started. The search target document reading unit 204 determines whether there is a search target document which has not been processed yet in the external storage device 4 (step 310). If there is a search target document, one search target document is searched. Storage buffer unit 2
30 (step 311). FIG. 13 shows an example of the search target document.

Subsequently, the search target document item cutout unit 20
5 refers to the search condition setting buffer unit 229 and determines whether to extract items of the search target document (step 312). When performing the item extraction, the search target document item cutout unit 205 analyzes the sentence structure of the search target document stored in the search target document storage buffer unit 230, and based on the analysis result, extracts the fixed-form item from the search target document. Is cut out, and the cut out document in item units is stored in the search target document item storage buffer unit 231 (step 313). FIG. 14 shows an item-by-item document cut out from the search target document.

Next, the search target document item generation unit 206
With reference to the search condition setting buffer unit 229, it is checked whether or not "create item of search target document = Yes" is set here (step 314). If “Generate item of search target document = Yes” is set, the search target document item generation unit 206 regenerates the document of the standard item stored in the search target document item storage buffer unit 231 (step). 315).

FIG. 15 shows an example of regenerating the document of the fixed item. The item reproduction of the search target document is performed in the same manner as the above-described item regeneration of the search key document. The regenerated standard item document is stored in the search target document item generation storage buffer unit 232.

After regenerating the document of such a fixed item, or when “create item of search target document = not set” is set in step 314, step 316 is executed.
Is executed.

In step 316, the search target word extracting unit 2
09 is started. The search target word extracting unit 209 cuts out a word from the search target document stored in the search target document item storage buffer unit 231 or the search target document item generation storage buffer unit 232, and extracts the search target document (or Key words that reflect the contents of the item are extracted, and the extracted words are stored in the search target word information storage buffer unit 233. The extraction of the key word from the search target document is performed by morphological analysis or the like, similarly to the extraction of the key word from the search key document. The word information thus extracted is stored in the search target word information storage buffer unit 233 as shown in FIG.

Subsequently, the search target word appearance frequency calculation unit 20
8 is activated. Search target word appearance frequency calculation unit 208
Are the search target document item storage buffer unit 231 or the search target document item generation storage buffer unit 23 for the words stored in the search target word information storage buffer unit 233.
2 is calculated, and the result is stored in the search target word information storage buffer unit 233 as shown in FIG. 17 (step 317).

Thereafter, the common word extraction unit 217 is activated. The common word extraction unit 217 searches for a commonly stored word for each item of the same number in the search target word information storage buffer unit 233 and the search key word information storage buffer unit 238, and searches for a common word for each searched item. Is stored in the common word information storage buffer unit 239 (step 318).
FIG. 18 shows an example of a search result of the common word.

Next, the similarity calculating section 218 is started.
The similarity calculation unit 218 uses a search key document / search based on the storage information of the search target word information storage buffer unit 233, the search key word information storage buffer unit 238, and the common word information storage buffer unit 239 using a vector space method or the like. The similarity for each item between the target documents is calculated, and the similarity value is stored in the similarity storage buffer unit 240. FIG. 19 shows an example of the similarity value for each item.

When calculating the similarity value, the similarity calculating unit 218 searches the search key document item priority buffer unit 23
Refer to FIG. FIG. 20 shows an example of the priorities set in the search key document item priority buffer unit 234. If a priority is set here, the priority is added to the similarity for the item for which the priority is set. For example, if “priority = 10”, the similarity of the item is doubled,
If “priority = 5”, the similarity of the item is weighted by priority by, for example, multiplying the similarity of the item by 1.5, and the result is stored in the similarity storage buffer unit 240 again. (Step 319). FIG. 21 shows the similarity for each item to which a weight is assigned according to the priority.

Thereafter, the flow returns to step 310. If it is determined in step 310 that there is no search target document in the external storage device 4 that has not been processed, the search result output unit 219 is activated. As shown in FIG. 22, the search result output unit 219 determines the search target document having the highest similarity for each item of the search key word, and determines its document information (for example, document ID).
Is stored in the search result output buffer unit 240. And
The contents of the search result output buffer unit 241 are output to the display device 3 (Step 320). FIG. 23 shows an example of the output result.

As described above, according to the present embodiment, it is possible to search for a search target document having an item similar to each item of the search key document.

Further, as shown in FIG. 24, for each item of the retrieval key document, the sum of the similarities of the respective items of the retrieval target document is obtained, and as shown in FIG. 25, the result is stored in the retrieval result output buffer section. 240 and may be output together with the sum of the similarities. At this time, the IDs of the search target documents may be displayed side by side in descending order of the sum of the similarities.

Further, as shown in FIG. 26, the sum of similarities of all items may be obtained for each document to be searched, and the result may be output together with the sum of similarities as shown in FIG. Also in this case, it is preferable to display the IDs of the search target documents in the descending order of the total sum of the similarities.

The above-described similar document search function can be provided by storing it as a computer-executable program on a computer-readable CD-ROM or other storage medium.

[0065]

As described above, according to the present invention, the similarity between a search key document and a search target document is obtained in units of items constituting each document. If there is, a similar document search can be performed that focuses on the individual content of each item.

A means for setting a priority for each item is added, and the similarity of the retrieval key document and the retrieval target document is calculated in item units in consideration of the set priority. For example, when a total sum of similarities of all items is obtained for each search target document and the result is output as a search result, a more optimal similar document search in which a similarity of an important part of the document is weighted is realized. Can be.

Further, means for unifying the sentence structure of each item of the retrieval key document or the retrieval target document is further added.
By configuring such that the similarity between the search key document and the search target document is calculated for each item whose sentence structure is unified in this way, the difference in the sentence structure of each item affects the similarity. The degree can be reduced, and the validity of the calculated similarity can be increased.

[Brief description of the drawings]

FIG. 1 is a diagram showing a hardware configuration of a similar document search device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a control unit of the similar document search apparatus of FIG.

FIG. 3 is a flowchart showing the flow of processing of the similar document search device of FIG. 1;

FIG. 4 is a flowchart showing the flow of processing of the similar document search device of FIG. 1;

FIG. 5 is a diagram showing a search mode setting screen.

FIG. 6 is a diagram showing a storage example of search mode setting information.

FIG. 7 is a diagram showing a screen for setting the importance of each item in a search key document.

FIG. 8 is a diagram showing an example of a search key document.

9 is a diagram showing a document in item units cut out from the search key document of FIG. 8;

FIG. 10 is a view showing an example of a document regenerated from the item document shown in FIG. 9;

11 is a diagram showing an example of word information extracted from the document of each item shown in FIG. 10;

FIG. 12 is a view showing an example of information on the frequency of appearance of words extracted from the document of each item shown in FIG. 10;

FIG. 13 is a diagram showing an example of a search target document.

FIG. 14 is a diagram showing an item-based document cut out from the search target document in FIG. 13;

FIG. 15 is a view showing an example of a document regenerated from the item document shown in FIG. 14;

FIG. 16 is a view showing an example of word information extracted from the document of each item shown in FIG. 15;

FIG. 17 is a diagram showing an example of information on the frequency of appearance of words extracted from the document of each item shown in FIG. 15;

FIG. 18 is a diagram illustrating an example of a search result of a common word.

FIG. 19 is a view showing a calculation result of similarity for each item between a search key document and a search target document.

FIG. 20 is a diagram showing an example of priorities set for each item.

FIG. 21 is a diagram showing the similarity for each item to which a weight is assigned according to the priority;

FIG. 22 is a view showing a similar document search result for each item.

FIG. 23 is a diagram showing an output example of a similar document search result for each item in FIG. 22;

FIG. 24 is a view showing a similar document search result obtained by calculating the sum of similarities of the items of the search target document for each item of the search key document.

FIG. 25 is a diagram showing an output example of a similar document search result of FIG. 24;

FIG. 26 is a diagram illustrating a similar document search result obtained by calculating the sum of the similarities of all items for each search target document.

FIG. 27 is a view showing an output example of a similar document search result of FIG. 26;

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 control device 2 input device 3 display device 4 external storage device 200 main processing unit 201 initialization unit 202 input unit 203 output unit 204 search document reading unit 205 search target document item cutout unit 206: search target document item generation unit 207: search target document word extraction unit 208: search target word appearance frequency calculation unit 209: search target word information calculation unit 210: search key document input unit 211: search key document item priority input unit 212 ... search key document item cutout unit 213 ... search key document item generation unit 214 ... search key word extraction unit 215 ... search key word appearance frequency calculation unit 216 ... search condition setting unit 217 ... common word extraction unit 218 ... similarity calculation unit 219 search result output section 229 search condition setting buffer section 230 search target document storage buffer section 231 Search target document item storage buffer unit 232 ... search target document item generation storage buffer unit 233 ... search target word information storage buffer unit 234 ... search key document item priority buffer unit 235 ... search key document storage buffer unit 236 ... search key document item Storage buffer unit 237 search key document item generation storage buffer unit 238 search key word information storage buffer unit 239 common word information storage buffer unit 240 similarity storage buffer unit 241 search result output buffer unit 242 work buffer unit

 ──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Yukio Nakamoto 1-3-3 Shinmachi, Ome-shi, Tokyo Inside Toshiba Computer Engineering Co., Ltd. (72) Inventor Tsuyoshi Matsukuma 3-3-1 Shinmachi, Ome-shi, Tokyo Toshiba Computer Engineering Co., Ltd. (72) Inventor Takuya Nishina 3-3-1, Shinmachi, Ome-shi, Tokyo F-term within Toshiba Computer Engineering Co., Ltd. 5B075 ND03 NK02 NK31 PP24 PQ36 PR04 PR08 QM08

Claims (8)

[Claims] 1. A similar document search apparatus for searching a document similar to a search key document from a plurality of search target documents, comprising: an item cutout unit that cuts out a document of an item unit from the search key document and the search target document; Calculating means for calculating a similarity between the search key document and the search target document in units of the cut-out items, and outputting a similar document search result based on the calculation result. Similar document search device. 2. The method according to claim 1, further comprising: setting a priority for each of the items, wherein the calculation unit considers the similarity of the search key document and the search target document in item units in consideration of the set priority. The similar document search apparatus according to claim 1, wherein the degree is calculated. 3. The apparatus according to claim 1, further comprising a unit configured to unify a sentence structure of each item of the search key document or the search target document extracted by the item extraction unit, wherein the calculation unit unifies the sentence structure. 2. The similar document search apparatus according to claim 1, wherein a similarity between the search key document and the search target document is calculated for each item. 4. The calculation means includes: means for extracting a key word from an item cut out from the search key document, obtaining an appearance frequency for each item of the extracted key word, and information obtained by cutting out the search target document. Means for extracting a key word from an item, and calculating an appearance frequency of each item of the extracted key word; means for searching for a key word common to the search key document and the search target document; The degree of similarity of each of the search key document and the search target document is calculated based on the extracted key words from the search target document, the appearance frequency of each item of the key word, and the searched common key words. 2. A similar document search apparatus according to claim 1, further comprising: 5. The similar document search apparatus according to claim 1, wherein the calculation unit outputs a search target document having the highest similarity as a search result for each item of the search key document. 6. The calculation means calculates a sum of similarities of respective items of a search target document for each item of the search key document,
2. The similar document search device according to claim 1, wherein the result is output as a search result. 7. The similar document search device according to claim 1, wherein said calculation means calculates a total sum of similarities of all items for each search target document and outputs the result as a search result. 8. A similar document search method for searching a document similar to a search key document from a plurality of search target documents, wherein a document of an item unit is cut out from the search key document and the search target document. A similar document, comprising: calculating a similarity between a search key document and the search target document in units of the cut-out items; and outputting a similar document search result based on the calculation result. retrieval method.