JP3058275B1 - Search device - Google Patents

Abstract

An object of the present invention is to accurately perform a natural sentence search using a search sentence with a simple configuration. SOLUTION: The search sentence of a natural sentence is a search sentence input area 4
2 (S21), the special words of each entry of the special words phrase table are sequentially read, and the input search text is full-text searched using the read special words (S22). When a hit is found, information including the special term is output as a search result (S23). Next, words and phrases are automatically extracted from the search sentence (S24). The automatically extracted words are tabulated using the automatically extracted words table, and the information contained in the hit entry is output as a search result (S2).
5). The category information of the hit information is displayed in the hit information list display area 44 (S26). The user looks at the display in the list display area 44, selects desired information, and displays detailed information (S27).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search device for performing a search using a search sentence composed of natural sentences, and more particularly, to a simple and accurate search without performing complicated processing such as syntax analysis and semantic analysis. This makes it possible to search well.

[0002]

2. Description of the Related Art It has long been performed to create a search formula for one or more keywords and perform a search. However, in consideration of the usability of the user, it is desired that a natural sentence can be input as a search sentence and the search can be performed as it is. Such a search is also called a fuzzy search. Conventionally, when performing a search by entering such a natural sentence, it is necessary to parse the search sentence and, if necessary, perform a semantic analysis to extract appropriate keywords and create a search expression. There was a heavy load on the computer. Further, at present, it is impossible to perform a search with sufficiently high accuracy even if such complicated processing is performed.

[0003] As a technology related to the present invention,
JP-A-61-11846, JP-A-63-2174
No. 18 and Japanese Patent Application Laid-Open No. 64-28770. The descriptions disclosed in these publications are intended to narrow down keywords to be registered based on character types from the data or a part of the data when registering the data in the database. However, these techniques do not search using natural sentences as search sentences.

[0004] It should be noted that, here, a sentence in which a series of words are meaningfully connected is not called a natural sentence only a sentence in which the subject, predicate, etc. are completed, and a sentence in which nouns are stopped,
Natural sentences include various types of expressions, such as a sentence stopped, a subject only, and a predicate only.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide a retrieval apparatus capable of performing a retrieval of a natural sentence by a simple process with high accuracy by a simple process. And

[0006]

According to the present invention, in order to achieve the above object, a search device includes: a means for inputting a search sentence composed of natural sentences;
Means for extracting words and phrases based on the character type and the continuous length of the same character type; and means for searching a database based on the extracted words and phrases are provided.

In this configuration, since a phrase is mechanically extracted based on a character type and a length of a character string without performing syntax analysis or semantic analysis, natural amounts can be easily calculated with a small amount of calculation. You can search by text search sentence. In addition, a highly accurate search could be performed with a simple configuration.

Further, according to the present invention, in order to achieve the above object, a search apparatus includes: for each piece of search target information, a search target information body, and a character type and the same character type from a text acquired from the search target information body. Database means for holding a first word extracted based on the continuous length and a second word specified by the registrant of the search target information body; means for inputting a search sentence consisting of natural sentences Means for extracting a word from the input search sentence based on the character type and the continuous length of the same character type; means for searching for search target information in which the extracted word is registered as the first word Means for searching the search sentence for each of the second phrases, and searching for search target information in which the second phrase included in the search sentence is registered.

[0009] Also in this configuration, it is possible to perform a natural sentence search with a high accuracy while using a simple sentence. In particular, the registrant registers a special word as a second word based on his / her know-how, etc., and uses this special word to search the input search sentence in full text to determine whether or not it is applicable. A fine-grained search matching the target can be performed. In addition, since there are not many special words and phrases, even if the full text search is performed on the input search sentence, the calculation amount does not increase much.

In this configuration, the first phrase can be a character string composed of two or more Chinese characters or a character string composed of three or more katakana characters. Such a setting is preferably changed according to the characteristics of the information targeted by the database.

[0011] The search device may further include means for displaying a list of search results. Of course, the number of searches may be displayed.

[0012] When a plurality of entries in the list of search results are selected, the search target information may be displayed in sub windows arranged on tiles. In this way, a plurality of pieces of information can be compared and viewed.

[0013] Further, the retrieval apparatus further comprises means for associating one retrieval target information with another retrieval target information,
When the other search target information is searched and displayed, the above 1
May be displayed in a linked manner.

[0014] In this way, information that should be registered as one entity can be combined, information can be organized and handled, and the amount of information can be reinforced gradually.

Further, the other search target information and the first search target information may be displayed in a vertically connected area.

Further, the associating means may automatically perform the association based on the category information of the search target information. Of course, the information may be matched according to a user's instruction. It should be noted that the information may be merged with the user, and the merged information itself may be a separate entity in database management. Of course, it may be combined into one entity for database management.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to an image information retrieval system will be described below. This image information search system is designed to search for product brochures, product advertisements, magazine articles, newspaper articles, and the like as image information.

FIG. 1 shows a computer processing environment in which this embodiment is realized. In this figure, an information search system according to this embodiment includes a search server 11, a search client 12, and a database system 13 connected to a LAN 10. , Access server 14 and the like.
The search client 12 requests registration and search of image information to the search server 11.
Operates the data of the database system 13 for registration and search of image information based on the request of the search client 12. The database system 13 also holds text information and the like generated in association with the image information. This will be described later. The access server 14 is for realizing access from outside the network configured by the LAN 10. The access server 14 allows a salesperson to access the search server 11 from a customer or the like using a telephone line or the like. Desired image information can be obtained.

The information retrieval system of this embodiment registers and retrieves information. Hereinafter, registration and search of information will be described step by step. The information retrieval system according to the present embodiment automatically extracts words and phrases at the time of registering information and registers the words automatically. The user (registrant) manually designates and registers the information. The procedure for registering special words is performed. Further, the user registers information category information (attribute information). At the time of information search, a natural sentence search sentence is input. Then, for each of the special words, the search sentence is full-text searched. When the special words are detected, the corresponding information is first selected as a search result (see FIG. 8). Further, words are automatically extracted from the search sentence, and information to be used as the automatically extracted words in the search sentence is selected as a search result (FIG. 9).
reference).

FIG. 2 shows a procedure for registering information. For registering information, for example, a user interface as shown in FIG. 3 is used. That is, the image display area 21, the text display area 22, the category information registration area 23, and the special word / phrase registration area 24 are displayed as a user interface. In addition, the user who can both register and search information,
You may make it separate from the user who can only search information.

In FIG. 2, an original on a paper to be registered is scanned and formatted into a predetermined format (for example, JPEG), and then image data is stored (S11, S12). Thereafter, the text contained in the document is recognized and stored using the OCR (optical character recognition system) (S1).
3, S14). Of course, when an electronic document can be obtained from the beginning, the electronic document may be used as it is. In this state, the image and the text are displayed to the user in the image display area 21 and the text display area 22 of the user interface shown in FIG. The user looks at the text in the text display area 22 and transcribes a word to be registered as a special word into the special word registration area 24 as a special word (S15). Further, the user inputs the category information to the category information registration area 23 (S16). The category information varies depending on the type of information to be handled, and includes, for example, a manufacturer name, a series name, and a product name.

In the text, words are automatically extracted and registered based on the character type (S17). In this example, two or more Chinese characters and three or more katakana characters are automatically cut out and registered as words. Numerals, unit characters, and special characters can be used when searching for low territory with qualitative information or searching with common sense values, and these are also saved.

A phrase containing one kanji character is often a verb, and usually carries only general information. Therefore, by using two or more Chinese characters, only important information can be extracted. By the way, if a phrase containing only one kanji character is exceptionally important, it can be registered without exception as the above-mentioned special phrase.

FIG. 4 shows the relevance of each element of the registered information. In this figure, the information entity 30 includes image information 31, text information 32, category information 33, special word information 34, automatic extraction. It is configured to include the phrase information 35. The image information 31 is obtained by scanning a paper document. The text information 32 includes image information
It was obtained by CR processing. The category information 33 is input by the user regarding the attribute of the information. The special word / phrase information 34 is also input by the user. The automatically extracted word / phrase information 35 is automatically extracted. Image information 3
1 and text information 32 are managed as individual files for each information entity, and category information 3
3. The special word information 34 and the automatically extracted word information 35 are the category information table 33a and the special word table 34, respectively.
a and the automatically extracted phrase table 35a.

FIG. 5 shows an example of the configuration of the special expression phrase table 34a, in which information entity IDs including the special expression phrases are listed. For example, information A and C include special words a. FIG. 6 shows a configuration example of the automatic extraction word table 35a, and information including the automatic extraction word table is listed for each automatic extraction word table.

FIG. 7 shows a procedure for searching for information.
FIG. 8 shows an interface at the time of search. The search interface includes a category designation area 41, a search sentence input area 42, a hit count display area 43, and a hit information list area 44. In FIG. 7, when a search sentence of a natural sentence is input to the search sentence input area 42 (S21), the special words of each entry of the special words table are sequentially read, and the input search sentence is converted into the full text by the read special words. Search (S22). When a hit is found, information including the special term is output as a search result (S23). FIG. 9 schematically shows an example of a search using special words. In this example, there are a, b, c, e, g, and r as special words, and a full-text search is sequentially performed on the search sentences for these. In this example, b, e, and a are included in the search sentence in this order, and as a result, information A, B, and C are hit.

Next, words and phrases are automatically extracted from the search sentence (FIG. 7, S24). In this example, 2
Automatically extract more than Chinese characters and three or more katakana characters. Then, the automatically extracted words are tabulated using the automatically extracted word table (FIG. 6), and the information contained in the hit entry is output as a search result (S25). FIG. 10 schematically shows a search example using an automatically extracted phrase. In this example, words a, b,
e and g are automatically extracted. Then, the corresponding information is selected with reference to the table.

Note that the registrant can arbitrarily determine what special words are to be selected. In extreme cases, words that match the automatic extraction rules may be selected as special words.
In this case, the merit of using the special words is reduced. As the special words, words that are leaked by the extraction rule, for example, words composed of hiragana and words that include only one kanji character are selected. Further, a compound word composed of a word and a word may be registered as one word. In some cases, a search using only special words or a search using only automatically extracted words may be performed.

The logical sum of the search results output in steps S23 and S25 is obtained, and the category information of the corresponding information is displayed in the hit information list display area 44 (S
26). For example, a manufacturer, a series name, a product name, a sales start time, a use, a price, and the like are displayed for each hit entry. The number-of-hits display area 43 displays the number of hits. If the number of hits is large and not all entries can be displayed in the list display area 44, scrolling is performed.

The user looks at the display of the list display area 44 and selects and displays desired information (S27). Information can be displayed in a plurality of modes by selecting a display mode. For example, only the image information may be displayed, the text information or the category information may be added and displayed, or the special information phrase or the like may be added and displayed.

In this search, it is possible to narrow down the search range by designating category information using the category designation area 41. For example, it is possible to narrow down by manufacturer, product name, sales start time, application, price, and the like. The range of the price and the sales start time can be specified using an inequality or the like.

Further, since a search for an automatically extracted word can be performed using a number or a special symbol, it is possible to search for unusual information matching the number or a special symbol.

Next, additional features of this embodiment will be described.

FIG. 11 shows a related item dictionary. This dictionary is for mutually converting synonyms and related information included in a thesaurus and the like. When the search sentence is input and the words are divided, the dictionary is looked up, receives the conversion result together with the original word as an extracted word, and absorbs the fluctuation of the input search sentence. The special words are also converted in the same manner, and the conversion result is also subjected to full-text search of the search sentence along with the original special words.

In this embodiment, information can be merged with information. Alternatively, other information can be added to the predetermined information. FIG. 12 shows a procedure for automatically adding information based on category information.
In this example, when the categories match, image information and text information are merged. Actually, a mark indicating that the additional information is included in the additional information and the destination of the additional information are displayed. By doing so, related information can be treated as if it were one entity. In this case, the images are connected vertically. In this way, more images will be viewed by scrolling up and down. The user can handle up and down scrolling more easily than left and right scrolling.

The level of the category, or the item to be merged when the items match, differs depending on the type of information.

Further, there is a case where it is desired to add information even when the category information does not match. Therefore, information may be added according to a user's instruction. For example, when registering predetermined information, information of an addition destination may be displayed in advance, an addition instruction may be given, information to be registered may be input, and addition may be performed. Information may be searched for by operation, and the hit information may be added as an addition destination. In addition, addition of text and addition of an image may be individually specified. Of course, both may be added together. Such a designation can be set in advance even when it is automatically added.

Information can be added not only one but also a plurality of pieces sequentially. That is, three or more pieces of information may be linked.

It should be noted that the present invention is not limited to the above-described embodiment, and various changes can be made without departing from the gist of the present invention. For example, in the above-described embodiment, the search is performed for image information, but the search may be performed for only text. In this case, the text information (summary,
Automatic phrase extraction may be performed only from the top part, the end part, etc. (a part containing a lot of information). Also, voice,
It may be used for searching for a motion picture. In the above-described embodiment, the present invention is applied to a client-server environment. However, a standalone search system may be used. Also, the restrictions on the character type and the character string length of the same character type are not limited to katakana of two or more Chinese characters and three or more characters, and various changes are possible.

[0040]

As described above, according to the present invention, a search using a natural sentence search sentence can be performed with high accuracy while having a simple structure.

[Brief description of the drawings]

FIG. 1 is a system illustrating a computer processing environment in which an embodiment of the present invention is realized.

FIG. 2 is a flowchart illustrating a procedure of information registration according to the embodiment.

FIG. 3 is a diagram illustrating an information registration interface according to the embodiment.

FIG. 4 is a diagram illustrating elements registered as information according to the embodiment.

FIG. 5 is a diagram illustrating a special-word phrase table according to the embodiment.

FIG. 6 is a diagram illustrating an automatically extracted word / phrase table according to the embodiment.

FIG. 7 is a flowchart illustrating a procedure of an information search according to the embodiment.

FIG. 8 is a diagram illustrating an information registration interface according to the embodiment.

FIG. 9 is a diagram illustrating a search example using special words in the above embodiment.

FIG. 10 is a diagram illustrating an example of a search using an automatically extracted phrase in the above embodiment.

FIG. 11 is a diagram illustrating a related item dictionary according to the embodiment.

FIG. 12 is a diagram for explaining information addition in the above embodiment.

[Explanation of symbols]

 Reference Signs List 10 LAN 11 search server 12 search client 13 database system 14 access server 21 image display area 22 text display area 23 category information registration area 24 special words / phrase registration area 30 information entity 31 image information 32 text information 33 category information 33a category information table 34 special notes Phrase information 34a Special word table 35 Automatically extracted words 35a Automatically extracted word table 35a 41 Category designation area 42 Search sentence input area 43 Hit number display area 44 Hit information list display area

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-7-182370 (JP, A) JP-A-7-230468 (JP, A) JP-A 8-221440 (JP, A) JP-A 8- 30627 (JP, A) JP-A-10-187736 (JP, A) JP-A-10-11460 (JP, A) "Smart saving of telephone charges and connection fees for Internet surfing" Nikkei Personal Computer, October 21, 1996 JP, p. 204-209 (58) Field surveyed (Int. Cl. ⁷ , DB name) G06F 17/30 JICST file (JOIS)

Claims (7)

(57) [Claims] For each piece of search target information, a search target information body, a first phrase extracted from a text acquired from the search target information body based on a character type and a continuous length of the same character type, and the search target information Database means for holding the second phrase specified by the registrant of the main body; means for inputting a search sentence composed of natural sentences; and, based on the input search sentence, a character type and a continuous length of the same character type. Means for extracting a search term, a search means for searching for search target information in which the extracted word is registered as the first word, a search for the search sentence for each of the second words, Means for searching for search target information in which a second phrase included in the search term is registered. Wherein said first word search device according to claim 1 wherein the string of strings or 3 or more characters katakana consisting of two or more characters Kanji. 3. A means for displaying a list of search results.
The search device according to claim 1 . 4. The search device according to claim 3 , wherein when a plurality of entries in the list of the search results are selected, the search target information is displayed in sub windows arranged on tiles. 5. A means for associating one search target information with another search target information, and when the other search target information is searched and displayed, the one search target information is also displayed in a linked manner. The retrieval device according to claim 1, 2, 3, or 4, wherein 6. The claims and the other search target information and the search target information for the 1 displayed on the linked vertically region 5
Search device as described. Claim 7. wherein said correlating means performs automatic association based on the category information of the search target information 5
Or the search device according to 6 .