JPH10283355A - Method for analyzing enterprise name and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To separate an enterprise name even when the ending word of the enterprise name is not included, to describe an optimal expression, and to easily change a rule at the time of dividing the enterprise name inputted without any separation of an organization into a main name and a lower organization name such as a branch office and section name, and using it for a task for retrieval with the main name or the lower organization name. SOLUTION: A word dividing part 201 divides an inputted enterprise name into words by using a word dictionary 202 which registers a word and the meaning constituting the enterprise name, and applies the meaning of the word. A rule collating part 203 collates the divided enterprise name with a separation rule for applying the separation of the organization, and analyzes the separation position. A separation applying part 205 applies the separation of the organization to the analyzed separation position. A rule selecting part 401 selects one of each rule 402-405 from a rule storage part 204, and extracts and applies it.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a customer system for handling company names, which divides a company name entered without division of an organization into a main name and a subordinate organization such as a branch or a department name, and the like. More particularly, the present invention relates to a method and apparatus for analyzing a company name which can be used for a task of retrieving by organization name, and in particular, can make a table of delimiter rules and easily change the rule.

[0002]

2. Description of the Related Art In general, a customer system handling company names, for example, a public customer service handling company names at a city hall or a post office, or a private customer service handling company names at an advertising company, a bank, an insurance company, etc. It handles various forms of names, such as names that last from the main name to the subordinate organizations, names in the name of the main name only, and names in a form that omits departmental staff. Then, if necessary, it is necessary to extract only the unified names from the names of these various styles. When searching, only the main name or the name divided into the names up to the department name It is necessary to search by. In such a case, a character string representing the end of the main name included in the company name (for example, “city hall”, “library”, “university”, etc.) or a specific character string representing the subordinate organization (for example, “department”, “section”) ], “Room”,
"Branch" etc.).

[0003]

However, if the delimiter is determined only by including a specific character string such as "city hall" or "department" as described above, the following problem occurs. The last words of the company name are "University", "Agricultural Cooperative", "Shop"
Not only in the case of a single word such as "promotion / meeting" or "shopping / center", there are also cases in which there are two or more words, and it is impossible to consider all of them from the beginning. As in the case of "@Super City Hall", if the word indicating the end of the main name is followed by "before", the meaning of the word changes to indicate the target. Embedding such cases in a program one by one is extremely difficult and time-consuming. In the name of a chain store or the like, it may not be possible to judge where to divide based only on the meaning of a word. For example, "@rental car Yokohama / Isogo store" and "@ mart / Yokohama Isogo store" have the same place names, but are separated at different positions. In some cases, such as "ABC / Shibuya store", there is no clue to the separation in the name of the main person. Therefore, an object of the present invention is to solve such a conventional problem, and a company name input without division of an organization can be separated by a division rule and used for a search. It is an object of the present invention to provide a method and apparatus for analyzing a company name that can be correctly separated even if there is no clue to the delimitation or a delimiter posted in the main name.

[0004]

In order to achieve the above object, a company name analyzing apparatus according to the present invention employs notations such as words themselves, meanings of words and negative signs, delimiters, symbols representing arbitrary words, and the like. A rule interpreting unit that interprets rules described by combining the rules, a rule matching unit that matches a company name with a delimiter rule, and a storage unit that stores a rule that gives a delimiter. , The meaning of the word is given, the break position is analyzed by the rule matching unit, and the break of the analysis result is given. In the present invention, since the break rules are tabulated,
Rules can be easily changed, and rules can be written using symbols representing word meaning, negation, word list, and the like, so that the expressiveness of rules increases. In addition, rules to check from the beginning and rules to judge the main name from the lower organization are set, so even if there is no clue to the division, regardless of the lower organization name, such as a convenience store, there is no clue to the separation posted in the main name Can be separated correctly.

[0005]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a diagram for explaining the principle of the present invention. The method for analyzing a company name according to the present invention is roughly divided into two.
It can be divided into two step operations. Step 1
Is a process of dividing a character string into words and assigning meanings to those words. The next step 2 is a process of referring to the rule storage unit to search for a matching rule and assigning an organization break. In the first embodiment of the present invention, the break rule of step 2 is tabulated in the rule storage unit, thereby making it easy to change the rule. Further, in the second embodiment of the present invention, not only the word entity but also the meaning of the word, words that must not be placed before and after the word, a list of words, and the like can be described in the rule, thereby increasing the expressive power of the rule. Let it. Further, in the third embodiment of the present invention, in the delimiter rule of step 2, not only the rule for determining and terminating the last word of the company name, but also the rule to be collated from the head and the part of the subordinate organization are determined. By providing a rule for delimiting with, a company name that cannot be delimited by only the last word can be delimited.

FIG. 2 is a block diagram of a company name analyzing apparatus showing a first embodiment of the present invention, and FIG. 5 is an operation flowchart showing a company name analyzing method in FIG. First
In the embodiment, the company name analysis device divides the input company name into words using a word dictionary, and gives a word meaning, a word itself and inflectional change, meaning, A word dictionary 202 storing data necessary for analysis, a rule matching unit 203 for matching a company name divided into words with a segmentation rule, and a segmentation rule storage unit 204 for storing a rule for giving an organizational segmentation When,
The rule collating unit 203 includes a delimiter assigning unit 205 that assigns an organization delimiter to a delimiter position analyzed.

[0007] As shown in FIG.
Divides the input character string into words and assigns meanings to the words (step 501). Next, the rule matching unit 203 repeats the following processing while shifting the rule matching position from the beginning to the end of the character string before division (step 502). Rule matching unit 203
Retrieves rules one by one from the delimiter rule storage unit 204 (step 503), and checks whether or not the input sentence matches the rule at the collation position (step 50).
4). In other words, the operation of the rule matching unit 203 is as follows. For the character string (divided into words) output from the word division unit 201, rules are sequentially extracted from the division rule storage unit 204 from the first word to the end, and Check if the rule matches the sentence matching position. If the pattern matches (step 506), it is determined whether the next break position is at the end of the publication (step 50).
7) If it is at the end, the next rule is taken out without a break (step 508). If the delimiter position is not the end of the publication, it is delimited at the designated position (step 505). For example, as a result of word division for an input of "Tama agricultural cooperative Musashino branch", the result is "Tama / Agricultural cooperative / Musashino / branch". For this result, since there is no rule that matches with the first word “Tama”, a rule that matches with the next word “Nokyo” is searched. As a result, because the rule matches with the rule that is delimited by “Agricultural Cooperative” in the rule that matches by the last word, the delimiter is provided with a delimiter of “Tama agricultural cooperative / Musashino branch” (step 505). However, if the collated position is at the end, it is meaningless to insert a break there, so the process proceeds to the next rule. For example, in the case of “Tama agricultural cooperative”, it is meaningless to put a break after “no agricultural cooperative”, so the process proceeds to the next rule without a break here.

FIG. 3 is a block diagram of a company name analyzing apparatus showing a second embodiment of the present invention, and FIG. 6 is an operation flowchart showing a company name analyzing method in FIG. FIG. 7 is a diagram of a table showing characters that can be used in the rules, and FIG. 8 is a diagram of a table showing the meaning of half-width alphanumeric characters. In the second embodiment, as shown in FIG.
A "rule interpreting unit 301" is added to the configuration of the first embodiment. That is, in order for the rule matching unit 203 to perform matching,
A rule is extracted from the delimiter rule storage unit 204, and the rule interpreting unit 301 interprets the rule, thereby facilitating the process of the rule matching unit 203. The operation flow of the second embodiment is the same as the flow of FIG. 5, which is the same as that of the first embodiment. The operation (step 504) of checking whether or not the input sentence matches the rule at the matching position is detailed. However, steps 505 to 508 below this are the same as those in FIG.

The rule is stored in the delimiter rule storage unit 204 as 1
When it is determined whether the rule matches the rule at the collation position, first, one item is extracted from the top of one rule selected from the rule storage unit 204 (step 601). Here, it is assumed that an item delimiter is a single-byte character. As shown in FIG. 8, the meaning of the half-width alphanumeric characters is as follows: in the case of the meaning symbol 1, the surname / company name (for example, Suzuki), and in the case of the meaning symbol 2, the name (for example,
In fact, the meaning symbol 3 represents an organization and a post (for example, section manager, government building). As shown in the destination of the item selection 601 in FIG. 8, when the extracted item is a full-width character string, it is compared with the word notation at the collation position and proceeds to the next step (step 602). On the other hand, if the extracted item is a half-width numeral / upper case, it is compared with the meaning of the word at the collation position, and if they match, the process proceeds to the next step (step 60).
3). For symbols and single-byte lowercase letters, step 6 in FIG.
Steps 04 to 607 are performed. For example, when the opening symbol "[" appears in the brackets, the words up to the closing symbol "]" in the brackets are stored as a list of words to be matched (step 6).
05,606). In the case of a period ".", The matching position is set to the next word. If the symbol is an asterisk "*", the matching position is shifted until the next rule item is matched (step 604). In addition, in the case of the small English character “x”, the matching condition is that it does not match the next appearing word or word list. That is, the collation conditions are set to disagree (step 607). Finally, the collation result is returned to the caller, and the process proceeds to the next process.

As shown in FIG. 7, characters that can be described in a rule include a kanji character string (eg, NTT), a% symbol (organization separator) (eg, NTT% Shibuya / branch), half-width alphanumeric characters (word meaning) ) (For example, 9 branch offices), period "."
(Arbitrary one word) (for example, NTT%. Branch), asterisk “*” (arbitrary 0 or more words) (NTT% * branch), small English character “x” (word (list) described next)
Must not come (e.g., NTT% x software), "[""]" (list of words) (e.g., NTT
% X [software, docomo, personal]), ","
(Word list break), "/" (word break) (for example, NTT / software), and the like.

FIG. 4 is a block diagram of a company name analyzing apparatus showing a third embodiment of the present invention, and FIG. 9 is an operation flowchart of the company name analyzing method in FIG. The configuration of the third embodiment is similar to that of FIG.
Is connected to the rule selection unit 401, and the delimiter rule storage unit 204 is connected to the rule selection unit 401. In the delimiter rule storage unit 204, a rule 402 to be collated from the beginning,
A rule 403 for referring from an arbitrary position, a rule 404 for separating the main name from the lower organization, and a rule 405 for separating the lower organization are provided. The flow of FIG. 9 shows in detail the step 503 of extracting one from the delimiter rule storage unit 204 in the flow shown in FIG. 5, and the other steps are the same as those of FIG. When extracting rules from the delimiter rule storage unit 204, a rule group to be collated from the beginning is selected (step 902), a rule group delimited by corporation type is selected (step 903), or a rule group is delimited by an end word at an arbitrary position Whether to select a rule group (step 904), to select a rule group to put a main name in front of the subordinate organization (step 905), or to select a rule group to separate the subordinate organization (step 90)
6), select one of them. The collation process applies the classified rules in order.

FIG. 10 is an explanatory diagram showing an example of a break rule in the present invention. For the type that attempts to match from the beginning, at the time of registration, for example, 9 branch office% x before, NTT
%. Register as a branch, and at the time of analysis, for example, Musashino City Hall%
It is separated from the General Affairs Division and the NTT% Kichijoji Branch. For example, in order to accurately delimit a chain store such as a convenience store, the precision of delimitation can be improved by a rule for collation from the beginning. Next, in the corporation type, for example, C% is registered at the time of registration, and, for example, at the time of analysis, it is separated from Nippon Telegraph and Telephone Corporation% General Affairs Department. A symbol indicating a corporation type such as (Shares) Ltd. almost certainly represents a division of an organization, and therefore a separate rule for the corporation type is separately provided. Next, for the type that attempts to match with the last word, for example, at the time of registration,
Registered as a university, hospital, or agricultural cooperative, it is divided into, for example, the University of Tokyo, Faculty of Science, Musashikan Hospital, Internal Medicine, and Tama Agricultural Cooperative, Musashino Branch. Next, for the type that separates the main name in front of the lower organization, for example, at the time of registration, it is registered as, for example,% 9 stores, and at the time of analysis, for example, it is separated as, for example, ABC% Shibuya store. Unlike the ABC Shibuya store, the main name "ABC" has no clue to separate the organization. Even in such a case, if "% 9 stores" is registered, a separator can be inserted before the place name + store. Next, in the rule type for dividing the subordinate organization, for example, it is registered as a department at the time of registration, and is separated from, for example, an NTT% general affairs department% general affairs section at the time of analysis. By analyzing the subordinate organization and using the rule immediately before the main name as a delimiter, it is possible to separate even when there is no clue to the delimitation in the main name. For example, if "N / automobile / Yokohama" is registered, even if there is a division between place names such as "N automobile Yokohama Isogo store", it is possible to accurately separate.

[0013]

As described above, according to the present invention,
Separation rules are tabulated so that meanings, words, negations, word lists, etc. can be described, so that optimal expressions can be described according to the case, and rules can be easily changed by tabulation. it can. In addition, since a rule for matching from the beginning and a rule for judging the delimitation of the main name from the lower organization are provided, division can be easily performed even when the last word of the company name is not included.

[Brief description of the drawings]

FIG. 1 is a flowchart showing the operation principle of the company name analysis method of the present invention.

FIG. 2 is a block diagram of a company name analyzer according to the first embodiment of the present invention.

FIG. 3 is a block diagram of a company name analyzer according to a second embodiment of the present invention.

FIG. 4 is a block diagram of a company name analyzing apparatus showing a third embodiment of the present invention.

FIG. 5 is an operation flowchart showing a company name analyzing method according to the first embodiment in FIG. 2;

FIG. 6 is an operation flowchart showing a company name analyzing method according to the second embodiment in FIG. 3;

FIG. 7 is a diagram showing characters that can be described in a rule in the present invention.

FIG. 8 is a diagram showing the meaning of half-width alphanumeric characters in the present invention.

FIG. 9 is an operation flowchart illustrating a company name analyzing method according to a third embodiment in FIG. 4;

FIG. 10 is an explanatory diagram showing an example of each rule in FIG. 9;

[Explanation of symbols]

201: word division unit, 202: word dictionary, 203: rule collation unit, 204: delimiter rule storage unit, 205: delimiter assignment unit, 301: rule interpretation unit, 401: rule selection unit, 402: rule to be collated from the beginning, 403: rule to be referenced from an arbitrary position, 404:
Rule for separating main name from subordinate organization, 405 ... Rule for separating subordinate organization.

Claims (4)

[Claims] 1. A company name analysis apparatus for giving a division of an organization to a company name inputted without a delimiter, wherein the input company name is converted into a word using a word dictionary in which the words constituting the company name and the meaning are registered. A word segmentation unit that assigns the meaning of the word; a segmentation rule storage unit that stores a rule that assigns an organization segmentation; and a rule matching unit that matches the company name segmented into words with the segmentation rule. And a delimiter for assigning a delimiter of an organization to a delimiter position analyzed by the rule collating unit. 2. A rule interpreting unit for interpreting a rule described by combining notations of the word itself, meanings of the word, negation symbols, delimiters, and symbols such as symbols representing arbitrary words. 2. The company name analysis device according to claim 1, wherein: 3. The rule storage unit stores a rule that refers to a company name from the beginning, a rule that defines a word that should come to the end of a main name, a rule that defines a main name by a word that indicates a subordinate organization, and a rule that defines a subordinate organization. 3. The company name analyzing device according to claim 1, wherein the device is classified into rules and stored, and the partitioning unit includes a rule selecting unit that sequentially applies the classified rules. 4. 4. A company name analysis method for giving an organization break to a company name input without delimiters, wherein the input company name is converted to a word using a word dictionary in which words and meanings constituting the company name are registered. A first step of assigning the meaning of the word; and a collation rule for assigning a division between the company name and the organization divided into words in the first step, and analyzing the position of the division. 2. A company name analysis method, comprising: a second step; and a third step of giving a division of an organization to a division position analyzed in the second step.