TW469386B - Document retrieval and classification method and apparatus - Google Patents

Abstract

In a document retrieval and classification system, a retrieving operation is performed on a database of documents in accordance with retrieval conditions entered by a user so as to pick up retrieved documents as intended. The user is allowed to input classification standards of a plurality of classifications in response to the retrieved documents picked up by the retrieving operation. The classification standards are converted into retrieval conditions. The similarity between the converted retrieval conditions resultant from the classification standards and retrieved documents picked up by the retrieving operation is calculated. And, an attribute of each retrieved document picked up by the retrieving operation to each classification is calculated with reference to the similarity, thereby classifying each retrieved document into a classification having a highest attribute.

Description

Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

Invention Field I The present invention relates to a document retrieval and classification method for searching a document as desired from a database storing a bundle of electronic document data. Furthermore, the present invention relates to a document retrieval and classification system that implements the document retrieval and classification method of the present invention. SUMMARY OF THE INVENTION The present invention is suitable for storing various kinds of file information in a database such as a memory device installed in a word processor, an office electronic version, a personal computer, or the like, and storing a medium in which they can be loaded. Information. Recent developments in fields including e-mail, electronic catalogs, and electronically published data links have provided users with a wealth of document information that can be accessed and obtained. Furthermore, the number of Internet users has increased significantly. Therefore, 'the need to search or collect documents from a huge database if desired. At the same time, the need for classifying picked files as desired increases. However, according to the conventional document retrieval and classification system, the search conditions and classification criteria are usually determined in advance or according to the preferences of users. In this respect, 'the conventional document retrieval and classification system is static from the standpoint of retrieval conditions and classification criteria.' Summary of the present invention_ One of the objects of the present invention is to improve the flexibility in document retrieval and classification. Another object of the present invention is to allow users to arbitrarily change search conditions and classification criteria. Another object of the present invention is to allow users to perform search and classification operations in response to search results based on their instantaneous judgments * This paper size applies the China Store Standard (CNS) A4 (210 X 297 mm) 4 —— : ----: ---. 1 ''! Exhibition * nn ^ " J. I line, C Please read the notes on the back before filling this page) A7 B7 ΗMember of Intellectual Property Bureau of the Ministry of Economic Affairs ΗConsumption Cooperation 钍Printed 5. Description of the invention (2 Another object of the present invention is to achieve an automatic classification to assist the user's mental activity. In order to accomplish the above and other related purposes. The present invention provides a first text peer retrieval and classification system 'contained use An input, output device to allow a user to enter search conditions and classification criteria: A search device is set to perform a = search operation on a database of documents according to a search condition containing arbitrary strings and character strings, and is used to Calculate the similarity between the retrieved files picked up by the retrieval operation and the retrieval conditions = a retrieval result storage device is set to store the retrieved files picked up by the retrieval operation ε a classification standard conversion device is The classification criteria are converted into search conditions. The classification criteria are represented as a set of arbitrary characters or strings. A retrieval result classification device is configured to classify the retrieved documents picked up by the retrieval operation according to a plurality of classification criteria. Therefore, the present invention can provide a flexible document retrieval and classification system to assist mental activities during document retrieval and classification. According to a preferred embodiment of the present invention, the retrieval device is responsive to a retrieval condition input by a user via the input / output device. The search operation is performed on the document database according to the search conditions input by the user. The search result storage device stores the retrieved file type 1 standard conversion device picked up by the search operation of the search device-response in response to the The input / output device enters multiple classification criteria, and generates a search and retrieval search bar accompanied by the input classification criteria with a search device between the converted search conditions and the process request file obtained from the request.柘 Likeness and store it in 礆 nYou Af Same as 4. 桧 索佩 果 V type device reference * Retrieve> ^ 1 cl I- t ^^ 1 i ^ · 1 «1 I * n · n I n I n IBP J Aft ^^ 1 I line (please read the precautions on the back before filling in this page) '： Ί0 · 4 69 38 6 ΚΙ ___ ____ Β7 ___ V. Description of the invention (3) The similarity calculated by the device is calculated by the retrieval The attributes of each retrieved document picked up by the operation are sorted based on the attributes of each classification criterion, thereby implementing a document classification. In this configuration, the user can enter search conditions arbitrarily when such a group or character string is in mind during the retrieval operation. The user can arbitrarily classify the search results as desired. Based on the preferred embodiment of the present invention, 'the input / output device allows the user to input multiple classification criteria each containing a set of arbitrary words or arbitrary character strings. 'And the classification standard conversion device converts the set of arbitrary words or strings into search conditions. With this configuration, the user can input any word group or any character string emerging in their mind as a classification criterion (that is, a classification position). Therefore, a great deal of flexibility is given to the stance of setting classifications. According to a preferred embodiment of the present invention ', the document retrieval and classification system further includes a keyword detection device for extracting keywords from an arbitrary sentence or document. In this case, the 'keyword detection means is responsive to a plurality of classification criteria expressed by a user via an arbitrary sentence input by the input / output means' and extracts keywords from the input sentence. At the same time, the classification standard conversion device converts a set of extracted keywords into search conditions. With this configuration, the user can directly input any sentence belonging to a desired field as a classification criterion. This makes it possible to express complex classification positions. Therefore, the setting of classification positions can be implemented flexibly with multiple levels. According to a preferred embodiment of the present invention, the input / output device allows a user to specify a plurality of files for use as a plurality of classification criteria. The designated document is selected from the retrieved documents picked up by the retrieval operation. Keyword detection device This paper standard is applicable to national standards of the country < CNS) A4 specification (210 X 297 public love) < Please read the precautions on the back before filling this page) Cooperative Du Printing 6 Member of the Intellectual Property Bureau of the Ministry of Economic Affairs and Consumer Cooperatives Print A7 B7 V. Description of the invention (4) Extract keywords from the designated text companion. At the same time, a 1-category standard conversion device converts a group of keywords out into search conditions; with this configuration, after the user confirms the retrieved document picked up by the retrieval operation, the user can select the retrieved document itself or a part thereof As a position for expressing classifications, the setting of classification positions can be easily implemented. Furthermore, the present invention provides a second document retrieval and classification system including an input / output device for allowing a user to input retrieval conditions. A retrieval device is provided to perform a -retrieval operation on a database of documents according to retrieval conditions containing arbitrary strings and character strings, and is used to calculate the similarity between the retrieved files and retrieval conditions picked up by the retrieval operation. A retrieval result storage means is provided to store the retrieved documents picked up by the retrieval operation ... The keyword detection means is arranged to retrieve keywords from the retrieved documents picked up by the retrieval operation-an automatic keyword classification means is arranged to retrieve the retrieved documents The extracted keywords are automatically classified into a plurality of clusters. A classification criterion conversion device is set to convert the classification criteria into search conditions. Each taxonomy is a set of keywords classified into clusters. A retrieval result sorting means is arranged to sort the set of retrieved documents picked up by the retrieval operation according to the sorting criteria. Therefore, the present invention can provide an automatic document retrieval and classification system to assist mental activities during document retrieval and classification. ^ According to a preferred embodiment of the present invention, the retrieval device responds to the retrieval input by the user via the input and output device. Conditions and according to the search conditions entered by the user to perform a retrieval operation on the f database of the file: the retrieval operation retrieved by the retrieval operation of the retrieval device of the 5th storage device to store the retrieved text 迮 rr 冼Seize and retrieve the retrieved text of the operation to extract I Μ, ---------------- lacquer (please read the precautions on the back before filling this page) Economic Village Intellectual Property Bureau Consumer Goods Cooperative Yinle 4 6 9 38 6 A7 _ B7 V. Description of Invention (5) Take keywords. The automatic keyword classification device automatically classifies the extracted keywords into a plurality of clusters. The classification standard conversion means generates a converted search condition caused by the classification criteria of a set of keywords in which each department is classified into each cluster. The retrieval means calculates the similarity between the converted retrieval condition and the retrieved document picked up by the retrieval operation, and stores it in the retrieval result storage means. At the same time, the retrieval result classification device refers to the similarity calculated by the retrieval device to calculate the attributes of each retrieved file picked up by the retrieval operation for each classification criterion, thereby implementing a document classification. With this configuration ', it becomes possible to automatically take the stand of the classification inherent in the search results independently of the classification criteria input by the user. Users can automatically obtain classification positions that they did not expect. No special effort is required. As a result, it becomes possible to effectively assist in document classification. Furthermore, the present invention provides a first document retrieval and classification method including a step of performing a retrieval operation on a document database according to retrieval conditions input by a user to pick up the retrieved document as desired; allowed A step in which a user inputs a classification criterion of a plurality of classifications in response to a retrieved document picked up by the retrieval operation; a step of converting the classification criteria into a search condition; A step of converting a search condition and a similarity between the retrieved files picked up by the search operation; and a step of calculating an attribute of each category of each retrieved file picked up by the search operation with reference to the similarity, whereby Each retrieved file is classified into a category with the highest attribute. In this way, the user can arbitrarily enter the search conditions when there is such a group in mind during the search operation. Furthermore, the user can retrieve the paper size applicable to the National Standard (CNS) A4 specification (210 X 297 public Chu II 丨 -------- installation ---- I nn amme ϋ 1 . ^ 1 t line-, (Please read the note on the back to fill out this page first) A7 Economic Guo. Printed by the Intellectual Property Bureau Staff Expense Cooperatives V. Invention Description (6) The results can be arbitrarily classified as desired: Therefore, „The present invention can assist mental activity during document retrieval and classification = according to a preferred embodiment of the present invention. When the user 釺 for each classification: class standard enters a set of eight arbitrary words or any character string. The arbitrary or arbitrary strings are converted into search conditions, and the similarity between the converted search conditions and the retrieved files picked up by the search operation is calculated. This method allows the loser to lose eight Arbitrary word group :: Italian character string as a classification criterion (ie, the position of classification therefore gives great flexibility in the position of classified classification. According to a preferred embodiment of the present invention, when a user enters an arbitrary sentence for use When making the classification criteria for each classification, the keywords are extracted from the sentence M and a group Keywords are converted into search conditions, and the similarity between the converted search conditions and the retrieved files picked up by the M cable is calculated: In this way, the user can directly enter the Arbitrary sentences are used as classification criteria. This makes it possible to express complex classification positions. Therefore, the setting of the 'classification position' can be flexibly implemented in multiple levels. According to a preferred embodiment of the present invention, a user specifies a retrieved document picked up by the retrieval operation. Multiple files in the specified file are used as the classification criteria for each classification: Then. Keyword is extracted from the extracted file-the keywords of the group are converted into search conditions. At the same time, after the search conditions are converted and the search operation is performed, The similarity between the retrieved retrieved files is calculated. With this configuration, the user retrieves the retrieved file itself or a part of it as a fida classification after using the retrieved search result to confirm the retrieval operation picked up by the retrieval operation. Field; the setting of the position of the classification of g and ^ yi jin, -------- order -------- (please read the second notice on the back before filling this page) Α7 θ-β4 δ9 386 ~ __Β7 ___ V. Description of the invention (7) Furthermore, the present invention provides a second document retrieval and classification method, which includes: performing a retrieval operation on the data frame of the document according to the retrieval conditions input by the user, such as A step of picking up retrieved files desirably; a step of extracting keywords from the retrieved files picked up by the retrieval operation; a step of classifying the extracted keywords into multiple clusters; a group of grapefruits belonging to each cluster A step of converting a keyword into a search condition; a step of calculating a similarity between the converted search condition caused by the extracted keyword and the retrieved file picked up by the search operation; and referring to the similarity to One step of calculating the attributes of each retrieved document picked up by the retrieval operation for each classification, thereby classifying each retrieved document into a classification having the highest attribute. In this way, it becomes possible to automatically extract the stance of the inherent classification of the search result without relying on the classification criteria input by the user. Users can automatically obtain classification positions that they did not expect. No special effort is needed, as a result, it becomes possible to effectively assist in document classification. The above-mentioned and other objects, features, and advantages of the present invention will become more apparent from the following detailed description of the J formula read from the accompanying drawings, in which: Figure 1 shows the first according to the present invention A functional block diagram of the structural configuration of a document retrieval and classification system according to the embodiment: FIG. 2 is a view showing a view of a retrieval result obtained by the document retrieval and classification system according to the first embodiment of the present invention; FIG. 3 is a display A view of the search results according to the classification standard according to the first embodiment of the present invention; The fourth circle shows the calculation of the attributes according to the first embodiment of the present invention. The paper size is applicable to the Chinese standard (CNS> A4 specification ( 210 x 297 mm> — — — — — 1111111 a 5, I — — — — — — — < Please read the notes on the back before filling out this page) Department of Economics, Intellectual Property, Employees, Consumer Cooperatives, Department of Printing 10 5 、 Explanation of invention (A: B7 View of employee co-operation. I ± fp ¥ e of Intellectual Property Bureau of the Ministry of Economic Affairs; Figure 5 is a view showing the result of document classification according to the first embodiment of the present invention: No. 6 Is a functional block diagram showing the structural configuration of a document retrieval and classification system according to the second embodiment of the present invention: and FIG. 7 is a diagram showing the structural configuration of a document retrieval and classification system according to the third embodiment of the present invention Functional block diagram. _Following the detailed description of the embodiment, hereafter, the preferred embodiment of the present invention will be explained with reference to the accompanying drawings: the same elements of the comprehensive drawing are designated by the same reference numerals. The embodiments of the present invention are based on the This character file. Therefore, the following explanation includes Chinese characters and or katakana 'followed by an English translation in parentheses. First Embodiment The first example shows that the first embodiment of the present invention is used to carry out-case retrieval and jealousy. Method-like structure of a document retrieval and classification system. Functional block diagram. In the document retrieval and classification system shown in Figure 1+, an input / output section 21 allows the user to enter retrieval conditions and classification criteria, and also enters retrieval. In the end, the result section 24 stores the file, and the search section 23 determines the similarity between the retrieved file and the search conditions ... "The result storage section stores the retrieved text. The search result of a classification is entered into the conversion section, and the classification target supplied from the output section 2 丨 the M class material is transferred to the inspection chapter. The department can be searched—can be retrieved—lift the fruit section to “M 部 AGE | The criteria for the retrieval and retrieval of the tomb's photos (please read the precautions on the back before filling in this page) —installation-wire gauge 柊 A7 469386 ______B7______ Description of the invention (9) The searched documents are classified according to the classification criteria. The explanation will be explained later The details of the document retrieval and classification processing according to the first embodiment. First, the user enters a retrieval condition into the input / output section 21. For example, 'the following logical (Bolin) expression 1 can be given as a retrieval condition. (M OR u / OR policy) (1) where rice is the Chinese character for rice, 3 > is the katakana for rice, and policy is the Chinese character for policy. The retrieval section 23 retrieves documents stored in the document storage section 24 based on the retrieval conditions. The search section 23 may perform a search based on a search condition including an arbitrary word group or an arbitrary character string. Furthermore, the retrieval section 23 can calculate the similarity between the retrieval results and the retrieval conditions. As disclosed in Published Patent Application No. 9-3 19766, such a retrieval section may be combined with an entire sentence retrieval section capable of detecting all documents in which a finger group exists. The similarity between the search condition and the retrieved file (that is, the search result) can be expressed by the following formula. S (Dj) = 2 {fij X (l-log (di / N))} The sum of Σ is a relative variable “factory”, “fij ·”, which represents the occurrence of the document “ti” Frequency (or degree), 'di,' represents the number of documents in which the phrase ,, ti ,, appears, and "N" represents the total number of retrieved documents. The above expression takes the sum of the similarity of the individual words involved in the search conditions, ti ". This is based on the weighting of words according to the" TFIDF "method, and according to A7 B7 V. Description of the invention (10-scale similarity Degree calculation: Now suppose that in a retrieved document Dj. The search phrase in the current search condition has the following occurrence frequency ', that is, " fij ·· Given as follows = rice (rice (Chinese characters)) 3 口 〆 ( Daozhu (Katakana)) 2 Politics (Policy 丨 1 At the same time, among all the documents stored in the document storage section 24, the number of documents related to each search block is as follows: that is, "dr is given as follows {rice (Chinese characters)) 5000 π / (rice (Katakana)) 1250 Policy (policy) 2500 When N =] 0000, calculate the similarity S (Dj) of Dj in the following way (锖 first read the back of the > i (Please fill in this page for the matters needing attention) .— Install the Intellectual Property Bureau of the Ministry of Economic Affairs_Industrial and Consumer Cooperatives

SfDj) = 3 Xf] -log (5000/10000)) -2 x (i-! Ogf 1250/10000)) ~ 1 x (l-log (2500/10000)) = 6 + 8 + 3 = 17 2nd The figure shows the details of the search results, in which the number of files of each retrieved file is displayed together with the similarity and content of each retrieved file. According to Figure 2, a total of 0 files are picked up according to the above search conditions. The order of the ranks is specified by specifying the maximum value to 100 to normalize the similarity of each article, and the search results are stored in the search result storage section 2 "看 余 'output # 索索 纪: fruit after.' _Ϋ́_! _ 货 _ Search results 丨, ... new search ^ OJ- nnn ί I nnnnin -i if i8 >, _ · iaf If. Tfay-M. -1 ； ί: Λ; Regulation M 469386 Α7 Β7 Ministry of Economy Wisdom Printed by the Property Coordination and Consumer Cooperatives V. Description of the Invention (1) Sort the search results that have been passed. When the user wants to classify the search results shown in Fig. 2, the user inputs a plurality of classification criteria via the turn-in / output section 21. Based on the first embodiment, the classification criterion is a group of words expressing the position of classification. For example, the user can input the following words as classification criteria via the input / output section 21. Classification Standard 1: (Rice; Katakana), Rice Noodle (Rice Price), New Food Law (New Staple Food Control Act) Classification Standard 2: North Korea (North Korea), China (China), Rice North Korea Agreement (US-North Korea Negotiation) Classification criteria 3: Miguchi (USA), Mijun (US military) The classification criteria conversion section 22 converts the entered classification criteria into possible retrieval conditions in the retrieval section 23. For example, it is preferable to make a logical expression by appropriately combining the words input as the classification criteria and then using AND to connect the generated logical expression with the above-mentioned logical expression 1 for use as the final retrieval condition. The following are examples of search conditions that are converted. Search condition 1: (3〆OR rice price) AND (rice OR) 〆OR policy) Search condition 2: (North Korea OR China) AND (rice OR 3〆OR policy) Search condition 3: (rice country OR rice army) AND (Mi OR 3〆OR policy) Use AND to connect the last search condition (ie, logical expression 1) to the classification criteria to better reduce the scale of the number to be retrieved. Next, the search section 23 performs a search based on the search conditions 1 to 3 described above, and obtains the search results shown in the third step. This paper size applies Chinese national standard (CNS > A4 specification (210 X 297 male «)) 14 (Please read" Notes on the back before filling in the true ") I-SJ. A: B7 V. Description of the invention (l2… such as As shown in the figure, the search department obtained from the inspection document ... in Figure 2 has §-A plus phase ''% fruit. The sub-collection of the retrieved documents that belong to the entire collection is not equal to = ίνinch Yu Geshuo y + 々 ° ° 3 _ 'The similarity of the two retrieved files is displayed relative to — the corresponding piece of a score 4 / do battle' heart per cable \ 旳 similarity value. Now suppose saj) represents — file. Condition (ie, the similarity of the classification criterion ντ calculates a file for its next 'retrieval result classification section 26.' P is for M '; r' ,, asks $ 贝 标. 1 attribute Tn.jp For example, the following formula is used Calculate attribute D (丨, 丨). T <! 〇) -t * SfKj)-(lC) MOO ^ (S (ij) / IS (ik)) {2) where Σ is a relative variable, .k, · Sum 'and c is a constant in the range 0 < c <. The above formula 2 is only an example for obtaining attributes: the method for calculating the attribute I (i'j) is not limited to the formula 2. The figure 4 shows the conditions of 0. 5 applied to the file 1 and the classification t to 3. The attribute T (i, j) calculated by Equation 2 above. The search result classification section 26 uses the following formula 3 to identify a classification having the southernmost attribute T (丨, j) with respect to each file ". C (i) = max {T (i, j)} (3) Among them, "max" is the maximum value of a relative variable "j"; Finally, the output file of the search result classification section 26 belongs to the classification c (丨 丨 Conclusion: This conclusion is shown by the input eight / output section 2 丨Or notify the user of the < graph display; τ profile is shown in graph 4; f: an example of the finalization of the inverted π meals:% _} ¥ Ά: 广; 第: Κ 屮 顾 π's check. Cable knot tired 4 a; f. A set -t --- II I ------------ — Ini — In f (read the information on the back of the card before filling in this page) Intellectual Property of the Ministry of Economic Affairs BureaucratsΗConsumption

A7 4 69 386 B7__ 5. Description of the invention (13) The retrieved documents in combination) can be classified into multiple sub-sets according to the classification criteria 1 to 3 of the word groups input by the user. According to the above example,> Xunzi, M, is one of the given search elements in logical expression 1. However, the Japanese word, rice,. Has multiple meanings. Therefore, among the retrieved documents shown in Figure 2, some documents (documents Nos. 6 and 10) include `` rice '' expressing "rice", while other documents (documents 3, 4, 5, 7 and No. 9) includes "meters" expressing the United States (i.e., the United States of America). However, users can then divide these documents into different categories by entering the appropriate taxonomy. Furthermore, 'when the user enters a search condition or taxonomy, Users can arbitrarily select search groups without special attention, such as "New Food Law" and "Michao Agreement", etc. Combination words composed of multiple characters. Furthermore, the classification is finalized according to the given classification criteria Then, the user can designate a subset corresponding to a classification criterion as a new quantity for another classification (ie, precise classification). As mentioned above, the first embodiment of the present invention provides flexible document retrieval and classification Method and system. According to a first embodiment, a retrieval operation is performed on a database of documents according to retrieval conditions input by a user to pick up the retrieved document as desired. Allowing the user The classification criteria for multiple classifications shall be entered in the retrieved documents picked up by the retrieval operation. These classification criteria are converted into retrieval conditions 0 in the transformed retrieval conditions resulting from the classification criteria and the retrieval criteria picked up by the retrieval operation The similarity between the retrieved documents is calculated. At the same time, the attributes of each retrieved document picked up by the retrieval operation for each classification are calculated with reference to the similarity, thereby classifying each retrieved document into a classification having the highest attribute. This paper size applies to the national standard (CNS) A4 specifications (2) 0 X 297 mm. ≪ Please read the note on the back before filling out this page > ij- Printed by the Shellfish Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 16 V. Description of the invention (14 A7 B7 The cooperation between employees of the Intellectual Property Office of the Ministry of Economic Affairs and tears. Du J. Pi. The first embodiment of the present invention makes it possible to test in the heart the right time when you do this. Arbitrary input retrieval can check f condition 1, user file h '' Anyone who consistently controls can assist mental activity during the 'double-dong and classification', according to the embodiment of the classification, when the user Each person ^ ㈣ quasi-input ~ a set of arbitrary words or strings of characters, the two groups of words or strings of characters entered are converted into search conditions, and the converted 4 files retrieved by the search operation are converted 4 C. Legacy This makes it possible for users to enter any word group or string of Rensizi that they find in their minds as the classification criteria (ie, the position of the classification). Therefore, it gives great flexibility in setting the classification. c Second Embodiment 6 FIG. 6 shows a functional block circle of the structure configuration of the document retrieval and classification system according to the second embodiment of the present invention. In the document retrieval and classification system shown in FIG. 6, input / The output section 11 allows the user to input search conditions and classification criteria, and also outputs the search results and classification results—the file storage section 15 stores files. One search section # 4 Calculates the similarity between the retrieved files and the search conditions. Retrieve the results of the search. The search results of the search results of the retrieved files ... keyword check; 'The section i 2 receives the expression of the classification H field entered by the user through the input / output section 丨 丨 from the Receive Sentence detection keyword, inspection keyword m μ, use Ύ light and simple quasi. To choose. ¾, i 彳 * quasi conversion section 13 receives the classification standard factory and country 1: standard, specifications CiiO, -i ---- --------- Μ ------------- f Please read the notes on the back before filling this page) 4 69 38 6 A7 B7 V. Description of the invention (15) Economy The input keywords issued by the Ministry of Intellectual Property Bureau's Consumer Cooperatives are converted into possible search conditions in the search section 14. A search result classification section 17 refers to the similarity calculated by the search section 14 to classify the retrieved documents according to the classification criteria. Details of the document retrieval and classification processing according to the second embodiment will be explained later. First, the user inputs a search condition to the input / output section 2 j. As in the first embodiment, it is now assumed that a search condition defined by logical expression 1 is input. At the same time, the search results shown in Fig. 2 are obtained. When the user wants to classify the search results displayed in the second step, the user can input multiple classification criteria through the input / output section Π. According to a second embodiment, the classification criterion is a sentence expressing a standpoint of classification, 'identifying a reference number of a retrieved document, or an essential part of a retrieved document. For example, the user can input the following sentences via the input / output section 11 as a classification criterion. Taxonomy 4: 3 market jealous government 0 meter government (on rice harvesting market and government's rice price policy) Taxonomy 5: North Korea is jealous of China 々i, (c to 5 meters country to 応 (United States for North Korea and China) Classification standard 6: Korean shovel shogunate woman (10 § Mijun problem (question of the US military in Korea and Japan) In response to the input of such a sentence, the document retrieval and classification system of the second embodiment implements the following processing 》 Keyword detection section 12 uses a dictionary database (not shown) to extract the word groups that appear in each sentence according to the word formation analysis, and select each sentence, please «read the note on the back and fill in! Dimensions of this paper are in accordance with China National Standard (CNS) A4 (21〇κ 297mm) 18

V. Description of the invention (ί6) Keyword of the child (This poor word group is about the selection of the key words of each sentence. It is expected to check in advance the frequency (or degree) of the magic half in the document storage section 15 > and then select keywords based on the TFiDF .. block weighting method or the like: in, for example, Figure F and Information Science No. 26 (1988.) by Um_, based on the block weighting principle based on frequency of occurrence information, There is a method for weighting such words, and, for example, another method for selecting keywords. It is better to consider words such as katakana and hiragana when extracting words from a sentence (that is, characters for Japanese documents). Differences in font types such as Chinese characters, Chinese characters, etc. are effective for detecting new or combined characters that are not registered in the dictionary. Needless to say 'It is better to combine the above two methods appropriately. According to the second embodiment' reference dictionary data To select the keywords for each sentence. Now suppose the following words are extracted from the above classification criteria 4 to 6, classification criteria 4. mouth / (rice), market (.market), government (government), rice cloth policy (rice price Policy) points Criterion 5 ': North Korea (North Korea), China (China), Rice (uS) Classification Standard 6: Korea (Korea). Japanese < Japan), Rice Army (US Army) Question (Question) Subsequent Classification Standard Conversion Department Segment 13 converts the classification criteria 4 * to 6 into the retrieval segment in the same manner as the processing performed in the classification criterion conversion segment 22 of the first to fourth embodiments. Φ 4 possible search conditions ... Further users Possibly after viewing the retrieval statement shown in Figure 2 Φ :: r-疋 slowly cooling t .5 ,. 之 之 文 # 数 a as a classification criterion T.Tr -——_____ ___________ ”... Λ.; CN 丨 S); Vi Άϋ 楼), 0 I 1— tr nnnnn 1— I nn < 1 n—OJ_ I »ni (Please read the note on the back before filling this page) Wisdom of the Ministry of Economic Affairs Printed by the Bureau of Property, Industrial and Consumer Cooperation, A7 Λ 69 386 ___ Β7 ____ 5. Description of Invention (17) 〇 Classification Criteria 7: 1, 2 Classification Criteria 8: 4, 5 Classification Criteria 9: 9 In response to the input of such a reference number, The document retrieval and classification system of the second embodiment implements the following processing. The key detection section 12 reads from the file storage section μ by reference No. (that is, the 'classification criteria') of the text of the document specified, and extracting the keywords involved in the specified document ^ The keyword extraction can be implemented in the same manner as in the above example of extracting keywords from sentences. Alternatively, the comparison It is better to fetch keywords from each file in advance, and store the picked keywords along with the corresponding files in the file storage section 15. In this case, the keyword detection section 丨 2 refers to the designated file (that is, the 'classification' Standard) The reference number Ray reads the keywords stored in the file storage section 15. It is now assumed that the following blocks are extracted from the above-mentioned classification criteria 7 to 9. Classification criteria 7 ':: 7〆 (rice), reserve (inventory), food (food), rice 価 (rice price), Nonghyup (peasant association), production (product), farmhouse (farmer)' rice production (rice yield) 'Consumer (Consumer), Rice (Rice) Taxonomy 8': North Korea (North Korea), Talks (Conference), South Korea (Korea) 'Agreement (Talks), Rice (US),' Mihan (US-Korea) Question (Question) 'North-South (North-South), Korean Peninsula (Korean Peninsula), Rice Army (US Army) Classification Criteria 9': Okinawa (Okinawa), Rice (US), Lianbang (Federal), Investigation (Investigation) , Return (return) 'public documents (official documents), information (materials, paper size, applicable national standard (CNS) A4 specifications (210 X 297 male t > — II I) III J--i — — —— —— iusius I (please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 20 A: B7 Consumption cooperation of the employees of the Intellectual Property Bureau of the Ministry of Economic Affairs. 18) material). Anti-disposition (False sanctions>, local rulings (District Court decision (decision)) The classification criterion conversion section π subsequently classifies the classification criteria 7. to 9. in the same manner as the processing implemented in the classification criterion conversion section 22 of the first embodiment. Conversion into possible search conditions in the retrieval section 14: After Yuancheng converts the classification criteria into search conditions, the document retrieval and classification system of the second embodiment implements the same processing as disclosed in the first embodiment: as explained above The search results shown in Figure 2 (that is, the retrieved documents of the entire collection) can identify the reference number of the retrieved document, or the essence of the retrieved document according to the sentence that expresses the position of the user's turn classification. Part of the classification criteria 4 to 6 (or 7 to 9) to classify into multiple sub-sets. So 'it becomes possible for users to implement classification in multiple ways. For example, users can be flexible and selective on the sorted and retrieved documents Use a complex position or a simplified position; as described above, according to the second embodiment of the present invention, when a user inputs an arbitrary sentence for use as a classification criterion for each classification, a clause Extracted keyword, the extracted keyword set is converted into a search condition, and the similarity between the converted file retrieved by the retrieval condition and the retrieval operation or in the pickup 1 is

It is different. According to the second embodiment of the present invention, the user can directly input an arbitrary sentence + which belongs to the desired position + as a classification criterion; this makes it possible to express complex classifications. The field can therefore be implemented with I * multiple levels of flexibility. Setting of the classification position. 4 Contains the right winter. Hairpin: ϋ 泡 泡 # ;. User search 1 > · fi ·; -i ni -f ^ U, ...,., IEJ L- ., A tm ". · &Amp; ai 1:;; Λ;;.: ·:.:; J · ':-U-:-, It. ^ 1 n I nnn-* DJ. N. ^ 1 IT n I line < Please read the notes on the back before filling out this page) Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 469 38 6 A7 B7 V. Description of the invention (19) Designated in the retrieved documents picked up by operation Multiple documents, designated documents for use as classification criteria for each classification. Then extract keywords from the extracted files. A set of extracted keywords is converted into search conditions. At the same time, after the similarity between the transformed search condition and the retrieved document picked up by the retrieval operation is calculated, after the user has confirmed that the retrieved document picked up by the retrieval operation, the user can choose the retrieved document itself or a part thereof Share to express the position of classification. Therefore, setting of the classification stand can be easily performed. Third Embodiment A third embodiment provides a document retrieval and classification system, which is characterized in that classification criteria are automatically determined and the retrieved documents are automatically classified. Fig. 7 is a functional block diagram showing the configuration of a document retrieval and classification system according to a third embodiment of the present invention. In the document retrieval and classification system shown in Fig. 7, an input / output section 71 allows a user to input retrieval conditions and classification criteria, and also outputs retrieval results and classification results. A document storage section 76 stores documents. A search section 75 calculates the similarity between the retrieved documents and the retrieval conditions. A retrieval result storage section 77 stores retrieval results such as retrieved documents. A keyword detection section 72 detects keywords from the retrieved files stored in the search result storage section 77. An automatic keyword classification section 73 classifies a set of detected keywords into a plurality of clusters and the classified keywords in each cluster are used as classification criteria. The _classification criterion conversion section 74 receives the keywords (i.e., classification criteria) sent from the keyword classification section 73 and converts the input keywords into possible search conditions in the retrieval section 75. A search result classification section 78 refers to the paper size S Standard (CNS) A4 specification (210 X 297 Public Love) 22 — IJ — 1,1 — I — I — — — — — I —- ^ y · 1111111 (Please read the notes on the back before filling out this page) Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs, Consumer Cooperatives, Ainong A7 _____B7 ---- ------- V. Description of Invention (2G) The similarity classifies the retrieved documents according to the classification criteria. The details of the document retrieval and segmentation processing according to the third embodiment will be explained later. First, the 'enter' enters the retrieval conditions into the input :, and the output section 71-like the first In the embodiment, it is now assumed that the search condition defined by the logical expression 丨 is inputted = At the same time, the search results shown in Fig. 2 are obtained. The second embodiment is different from the first and second embodiments described above in that it does not need to rely on The user automatically determines the classification criteria by inputting eight classification criteria. The automatic classification according to the third embodiment will be explained in detail later. First, the keyword detection section 72 detects the relationship of each retrieved document stored in the retrieval result storage section 77. Keyword. The details of keyword extraction disclosed in the above-mentioned second embodiment of the present invention may also use the keyword extraction method disclosed in Published Japanese Patent Application No. 9-176822. Second, automatic keyword classification Section 73 classifies the keywords detected by the _ group into multiple sub-sets. For automatic keyword classification, the following method V is used; Now suppose that the file storage section 76 stores a total of n files D1 to Dn 'where a total of m groups of words W 1 to W m appear in this inertia respectively. The following n-th order internal quantities Vj can be cited for each block Wj.

Vj-iei.e2 „e3 ..eni, Equation 4 shows that each rsj quantity element ei (r: L., Η! Is true; r I Mi Λ j ί \ i〇g (n; W) ij! 4 : Zhang < I, Competitive π is JJ's House, NS) A4 Specifications--------------- Install i (Jing first read the $ items on the back before filling in this page} Order -_-· line. Intellectual Property Bureau of the Ministry of Economic Affairs, 8 Industrial Consumer Cooperatives, 9 9 6 6 A7 B7 V. Description of Invention (2i) where TFi (Wj) represents the frequency (or degree) of the word Wj in the file Di 'and DF (Wj) represents the number of files in which the word group Wj appears. Preferably, the vector Vj is normalized so that its length becomes 1. In this way, vectors 乂 to Vm can be obtained for each word group of the claw. Secondly, ' Consider multiple block groups G1 to Gp. Each block group contains a specific block group that often appears in a specific field. Each block group can be generated manually, or by using a dictionary or large-scale file. The occurrence distribution of the word group is automatically generated. In this case, the following n-th level vector VGk can be introduced for each word group group Gk. VGk = (e'l, e'2, e'3, shown in Equation 5 below Calculation of each vector element e'i (i = i, ..., n). E'i = TFi ( Gj) Xlog (n / DF (Gj)) (5) where TFi (Gj) represents the frequency (or degree) of the word group belonging to the group Gj in the file Di, and DF (Gj) represents the group of Dj The number of files in any block. Preferably, the vector VGk is normalized so that its length becomes 1. In this way, vectors VG1 to VGp can be obtained for p blocks, respectively. One of the vectors Vj and VGk can be obtained. The inner product is used to obtain the similarity Sjk between each word group Wj and the word group Gk. Using the above vector and similarity calculation makes it easy to realize automatic classification of keywords. For example, now suppose that there are three words often used in the following blocks. Groups G1, G2, and G3 This paper size applies to the Chinese National Standard (CNS) A4 (210 χ 297 mm) 24 II j III .1 — LI III · IIIIIII · 11111111 (please note the note on the back of IHtl before (Fill in this page) A: B: Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs f. 5. Description of the invention (22) G1: Beacon engine for automatic vehicles G 2: Aircraft accident G3: Internet retrieval section 75 Search for files related to the engine. Then, the keyword detection section 72 extracts the following keywords.

Office VU > (gasoline), accident (accident), W Ww, fuel (fuel consumption), search (retrieval), explosion (explosion), airport (airport) 'URL for individual block groups G 1. G2 and G3 The similarity of each word group is calculated in the following way: S (accident (accident): production (0.2.0.6, 0.3) S (WWW) = (〇.1.0.2.0.8) S (fuel cost (fuel consumption)) Gas 0.7.0 丄 0.2 > S (Retrieval (Retrieval)) = (0.0.0.2,0.6) S (Explosion (Explosion)) = (0.4.0.6,0.1) S (Airport (Airport)) = (0.0, 0.9,0,2) S (URL) = (0.L0.0,0.9) Then pay attention to each keyword belongs to the group with the highest similarity G1, G2, and G3 = G1: 々y _ 丨 > (Gasoline fuel cost (fuel consumption) G 2 accident [accident 丨 explosion potential {explosion; airport (airport) G 3 claim inspection (retrieval Bu [丨 Η1 so it 3 sets of action key sets; paragraphs' won the keyword group was (please read the precautions on the back before filling this page) --install -------- order-- ------- Line ------- 469 386 A7 B7 DuPont Printing 2. Description of invention (23) is input into the classification criterion conversion section 74. When the number of the word group G is large (for example, 100), or when the number of keyword groups used as classification criteria needs to be reduced (For example, 2), the automatic keyword classification section 73 operates in the following manner. * LSt step 1 uses the sum of the weighted values of the classified keywords for each block group 0, and then treats the obtained sum as The score of this word group is 0- * 2nd. Step 1 continuously considers the highest value of the score to select a predetermined number of groups. According to the above example, the score of G1: 0.8 + 0.7 = 1.5 The score of G2: 0.6 + 0.6 + 0.9 = 2.1 G3 score: 0.8 + 0.6 + 0.9 = 2.3 According to this, when the number of keyword groups needs to be reduced to 2, the automatic keyword classification section 73 considers the score of each word group to select a group. 2 And G3. When the automatic keyword classification section 73 implements the above processing, a group of keywords extracted from the retrieved files can be automatically classified into multiple groups. According to the above example, the following classification criteria are obtained. Classification criteria 10: Fang V 〇 > (Gasoline) 'Fuel (fuel consumption) classification Standard 11: Accident (accident), explosion (explosion) 'airport (airport) classification standard 12: WWW, search (retrieval)' URL Thereafter, the classification standard conversion section 74 is the same as the classification standard conversion section of the first embodiment. In the same way as the treatment implemented in 22, the classification standard is 10 to (please read the note on the back first »Item # Establish this page) Order----" This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297g t) 26 A. B7 Member of the Intellectual Property Bureau of the Ministry of Economic Affairs X. Consumption Cooperative Press 1 V. Description of Invention (24) 12 Translate into possible search conditions in the search section. After completing the conversion of the classification criteria into the search conditions, the text and classification system of the third embodiment is implemented as disclosed in the first embodiment, ^ as explained above. The third embodiment automatically determines the proper occurrences in the retrieved documents. Block of words. Detected stops were classified as classification criteria. So ‘it becomes possible to implement document classification based on the nature of the search results. In other words, the third example provides simplified document classification. The keyword group obtained by the automatic keyword data 73 can be entered by the input section 8. The output section M is shown to the producer_times. The user can modify or correct the group of related keywords indicated. Then, the classification criterion conversion section 74 sets the modified keyword group 卩 'the modified classification criterion' to the money condition. This process allows the user to know that the user is not aware of the classification position. As a result, the third embodiment effectively assists the document classification work. As mentioned above, the third embodiment of the present invention provides an automatic document retrieval and classification method and system for implementing a retrieval operation on a database of documents in accordance with retrieval conditions input by a user to pick up retrieved documents as desired- The keywords are extracted from the retrieved files picked up by the retrieval operation. The extracted keys + are classified into f clusters: a group of extracted keys belonging to each cluster are converted into search conditions: the converted search conditions caused by the extracted keywords and the retrieved by the retrieval operation The similarity between the documents is simultaneously: f is the relative degree to calculate the two disparities that are picked up by the retrieval operation, and the genus and membership of the class v. Classifies each retrieved document into a specification: 111IIIIII1 — — · * JII 1 III Sleep — — — — — — — (Please read the notes on the back before filling in this page) Λ 6 9 3 Β 6 Α7 ___ Β7 V. Description of the invention (25 According to the third implementation of the invention Example 'It becomes possible not to rely on the classification criteria turned on by the user to automatically extract the classification position inherent in the search results. Users can automatically obtain unintended classification positions. No special effort is required. Results' becomes possible to effectively assist the document Classified work. The invention can be implemented in several forms without departing from the spirit of its basic characteristics. The embodiments as described are therefore intended to be illustrative only and not restrictive, and thus the invention The scope is defined by the scope of the attached patent application, not by its previous description. All changes that fall within the boundaries and limits of the scope of the patent application, or equivalents of such boundaries and limits, are therefore intended to be governed by the scope of the patent application. Included: Component number comparison

Employees' Cooperative Cooperation of Intellectual Property Bureau of the Ministry of Economic Affairs Du Yin I 11,21,71… Input / output section 13,22,74… Classification standard conversion section 15,24,76… File storage section 17,26,78… Retrieval Result classification section 12,72 ... Keyword detection section 14,23,75 ... Search section 16,2 ^ 77 ... Search result storage section 73 people 'Automatic keyword classification sectionϋ ^ 1- I nnn «nn 1 * 9 (Please read the precautions on the back first. Write this page> This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) 28

Claims (1)

A8 B8 C8 D8 4 6 9 38 6 6. Scope of patent application 1 'A file retrieval and classification system, including: input / output devices (21, 11), used to allow a user to enter search conditions and classification criteria: search device (23, 14) for performing a search operation on a document database based on such search conditions containing arbitrary words or arbitrary substrings, and for calculating the retrieved files and such searches picked up by the search operation Similarity between conditions; search result storage means (25, 16) for storing the retrieved documents picked up by the search operation; classification standard conversion means (22, 13) for converting the classification criteria into Search conditions, the classification criteria being expressed as a set of arbitrary words or arbitrary character strings; and a retrieval result classification device (26, 17) for retrieving the retrieved items picked up by the retrieval operation according to multiple classification criteria File classification. 2. A document retrieval and classification system according to item 1 of the scope of patent application, wherein: the retrieval device (23) is responsive to the retrieval conditions entered by the user via the input / output device (21), and used in accordance with The search conditions entered by the reporter perform the search operation on the document database, the search result storage device (25) stores the searched files picked up by the search operation of the search device, and the classification standard conversion device (22)飨 Should be based on a plurality of classification criteria input by the user via the input / output device, and generate transformed search conditions generated by the input classification criteria, while the retrieval device (2 3) calculates the converted criteria The search conditions and Chinese paper standard (CNS) A4 specifications (210 X 297 mm) are applied to this paper size < Please read the notes on the back before filling in this page) i τ Printed by the Consumers ’Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs 29 Member of the Intellectual Property Bureau of the Ministry of Economic Affairs, “Consumer Hezhushe” " 1_ __ Si Application for Patent Scope This inspection operation picks up and stores such information in the search result storage device. The similarity between the retrieved files and the μ k cable cutter device (26 丨 refers to the similarity calculated by the retrieval device to calculate the attributes of each retrieved file picked up by the retrieval operation for each classification criterion, thereby implementing A file classification action c 3. Document retrieval and classification parameters according to item [Scope of patent application], where: edge input output device (2 η allows the user to enter each of which contains a set of arbitrary subgroups or arbitrary character strings Multiple classification standards, and the classification standard conversion device (22) converts the set of arbitrary words or strings into search conditions. 4. The document retrieval and classification system according to item 丨 of the scope of patent application, which further includes: Word detection device (12) for extracting keywords from an arbitrary sentence or file, wherein the keyword detection device (12) is responsive to a plurality of expressions expressed by an arbitrary sentence input by the user via the input / output device. Classification criteria, and extract keywords from these rounded sentences & A 77 type standard conversion device i 3) to convert the extracted keywords of the group into search conditions. According to the document retrieval and classification system of the first patent application scope, among which-the input \ .. surplus output device (n :! allows the messenger to designate multiple documents for use as multi-class classification umbrellas. These are designated The document is to select t Jv vj from the searched and retrieved Φ which should be replaced by the waiting t. — — — — — ——I—— — — 1 11 — 1 — 11 * lliJ — — · (( Please read the notes on the back before filling this page) ,,-:-: · -------- ---------------------- such as ~ __A test application! T ^^ Jia 蟫 push ws, A i specifications; .. 469 386 A8 B8 C8 D8 VI. Patent application scope The keyword detection device (12) extracts keywords from these designated documents, And the classification standard conversion device (13) replaces a set of extracted keywords with search conditions. 6. A document retrieval and classification system, comprising: an input / output device (71) for allowing a user to enter a search condition; a retrieval device (75) for using the The search condition performs a search operation on a document database and is used to calculate the similarity between the retrieved file picked up by the search operation and the search conditions; the search result storage device (77) is used to store the search operation by the search operation. The retrieved documents picked up; keyword detection means (72) for extracting keywords from the retrieved documents picked up by the retrieval operation; automatic keyword sorting means (73) 'for The extracted keywords are automatically classified into multiple clusters; a classification standard conversion device (74) is used to convert the classification standards into search conditions, and each of these classification standards is classified into a set of keywords of each cluster; and a search result classification device ( 78) for classifying a set of the retrieved documents picked up by the retrieval operation according to the classification criteria. 7. A document retrieval and classification system according to item 6 of the scope of patent application, wherein: the retrieval device (75) responds to the input / output of the paper size by the user via the Chinese National Standard (CNS) A4 specification (210 X 297 mm) — — — — — — — — — — — — — — — I) 11111111 I / {Please read the notes on the back before filling out this page) Printed by the Shelley Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs? Clothing 31 6. Scope of patent application: Member of the Intellectual Property Bureau of the Ministry of Economic Affairs, X Consumption, Printed Device, Type 1) Enter the search bar f3, and execute the operation on the document database according to the search conditions of the user ’s turn. The retrieval operation, the retrieval result storage device (77) stores the retrieved files picked up by the retrieval device of the retrieval device, * the keyword detection device (72) retrieves the retrieved files from the retrieval operation, Pomelo takes these keywords, and the automatic keyword classification device (73) automatically classifies the extracted key cars into the multiple clusters. * The classification standard conversion device (74) generates classifications that are classified into clusters by each department. -The converted test pieces generated by the classification criteria of the group of keywords, '..., i, the search device (75) calculates the converted check conditions and is retrieved by a search operation and stored in the search result storage device; The similarity between the retrieved documents and the retrieval result classification device (78) refers to the similarity calculated by the retrieval device to calculate each of the inspections picked up by the inspection operation ^ A file classification action based on the attributes of each classification criterion is used to implement a document classification action—a kind of document consultation and classification method, including a step of -τ column:, according to the search conditions entered by the user pair—the actual moment of the document database—the search Operation. The user who picks up the desired documents to be scanned responds to the selected jealousy companions picked up by the retrieval operation to turn multiple categories of jealousy criteria. ----------- ---------- Order -------- I (ttPlease read the notes on the back before filling in this page J A8B8C8D8 469 38 6 6. Apply for a patent scope and convert these classification criteria into a search Conditions; Calculate the similarity between the transformed search conditions generated from the classification criteria and the retrieved documents picked up by the retrieval operation: and refer to the similarity to calculate the various retrieved documents picked up by the retrieval operation The attributes of the retrieved documents for each category are used to classify each retrieved document into a category with the highest attributes. 9. The document retrieval and classification method according to item 8 of the scope of patent application, where: The taxonomy inputs a set of any When a word group or any character string is entered, any input word group or any character string is converted into a search condition, and the similarity between the converted search condition and the searched documents picked up by the search operation is changed. Calculate 10. According to the method of document retrieval and classification according to item 8 of the scope of patent application, wherein: when the user enters an arbitrary sentence for use as the classification criterion for each category, the keywords are taken out from the sentence and a group is extracted Keywords are converted into search conditions, and the similarity between the converted search conditions and the retrieved documents picked up by the retrieval operation is calculated. 11_ Document retrieval and classification according to item 8 of the scope of patent application Method, wherein: when the user designates a plurality of documents among the retrieved documents picked up by the retrieval operation to be used as classification criteria for each category, this paper size applies to the Chinese National Standard (CNS) A4 specification (210 X 297 Mm) — Kill — — IIIII t · 11--— — II I J. (Please read the notes on the back before filling out this page) Member of the Intellectual Property Bureau, Ministry of Economic Affairs Printed by Consumer Cooperatives 33 A8 B8cs D8 VI. The keywords of the patent application scope are extracted from these documents, and the 4 keywords are converted into search conditions and the converted search conditions and those picked up by the search operation The similarity between the retrieved documents in this temple is counted.-A method of document retrieval and classification, including: The following steps: Implementation of a document database based on the search conditions rotated by the user: search operation 'to pick the desired The retrieved documents: Therefore, the retrieved documents picked up by Ak are used to extract keywords; the extracted keywords are classified into multiple clusters; the ones belonging to each cluster-group of the extracted keywords Into search conditions; The similarity between the transformed search conditions generated by Shiyi Rong Xinrong and other keywords and the retrieved files picked up by the search operation: and the similarity is used to calculate the pick up by the search operation Each retrieved document has attributes for each category, thereby classifying each retrieved document into a category with the highest attributes = --------------- ^ ------- -^ --------- ^ (Read the precautions on the back before filling out this page) ___ Member of the Intellectual Property Bureau of the Ministry of Economic Affairs, Consumer Cooperation, -fic_ 1