JP4656330B2 - Synonym integration system - Google Patents

Description

  The present invention provides a synonym dictionary integration system that creates and integrates synonym dictionaries from a plurality of synonym groups obtained by extracting logically ORed word groups when searching a text database using a search logical expression. It is about.

2. Description of the Related Art A search system is used that searches a “text database” that is a collection of materials including characters, that is, texts, and finds materials required by a searcher.
A configuration example of this search system is shown in FIG. In FIG. 1, a text database system 20 built on a server or the like is provided with a display device 25 such as a liquid crystal and an input device 26 such as a keyboard / mouse via an electric communication line 27 such as the Internet or a LAN. Then, a search expression or the like is input from the search terminal 23, and a target document or the like is searched from the database. In this search system, a searcher creates and inputs a text search formula, and a material having a character arrangement that matches the content indicated by the search formula is extracted. “Text database” built on the server includes patent gazettes, newspapers and magazine articles, academic papers, text materials on the Internet, other text materials, secondary materials such as summaries, etc. These are the headwords and tags assigned. It is a set of textual explanations, headwords, tags, etc., attached to materials consisting of figures, images, videos, sounds, etc.

  In text search, for example, if you use “digital camera” in a text search expression, which represents the concept of “camera that electronically shoots without using silver halide film”, the computer will have exactly the same characters as “digital camera”. Extract only materials that contain columns. However, the expressions of the sentences described by each person are quite different. In addition, there are many expressions such as “electronic camera”, “CCD camera”, “still image camera”, “electronic still camera”, “digital camera”. The relationship between words that can be regarded as having almost the same concept is called a "synonym", the group of words is called a "synonym group", and the "synonym dictionary" is a collection of "synonym groups" for convenient use. It is.

Therefore, if only “digital camera” is used in the search expression, materials in which other expressions are used in the meaning of “digital camera” are not extracted, but by using synonyms in an OR combination, that is, the search expression is digital camera. + Electronic camera + CCD camera + Still image camera + Electronic still camera + Digital camera-(1)
It becomes possible to reduce extraction omission.
The creation of a synonym dictionary and the subsequent maintenance of corrections require a great deal of labor when done manually, so attempts have been made to use computers.
Patent Document 1 describes an attempt to automatically create a synonym dictionary using a computer for collecting and classifying words. From the search target document, structural analysis of the sentence using morphological analysis, examples of collecting word groups from the obtained relationships between words, examples of using associations, examples of performing semantic analysis, thresholds from co-occurrence relationships and appearance frequency An example of registering a word exceeding the dictionary is shown.
Patent Documents 2 to 5 show a configuration in which a search expression or log data input by a searcher is analyzed and stored in a synonym dictionary.

JP-A-8-161343 JP-A-6-314296 Japanese Patent Laid-Open No. 10-320419 JP 11-31168 A Japanese Patent Laid-Open No. 9-319767

Accumulation of synonyms by using natural language processing such as morphological analysis, structural analysis, association, semantic analysis, use of co-occurrence relationships, etc. described in the patent documents shown in the prior art is still used by humans We have not yet understood the diversity and complexity of words, and synonym dictionaries generated from them are not practical enough.
Also, a configuration in which a search expression or log data input by a searcher is analyzed and stored in a synonym dictionary is not dealt with regarding the problem of organizing and integrating word groups.
An object of the present invention is to provide a synonym dictionary integration system that creates a synonym dictionary from synonym groups extracted from a search expression.

In order to achieve the above-mentioned object, the present invention provides a synonym integration system for creating a synonym dictionary from a set of synonym groups whose rows are OR combination words of a search expression inputted at the time of search, and the synonym The synonym group importing means for processing the word frequency corresponding to each word of the group line to 1, storing the line and importing it into the synonym dictionary, one line of the synonym dictionary as a main line, and the other line as a sub line A main line sub-row determination means, a main line and a sub-line are compared, a line comparison means for determining whether a predetermined condition is satisfied and determining whether to perform integration processing, and the line comparison means, As a result of comparison, when a predetermined condition is satisfied, a word that is the same as the main row in the sub-row is added to the word frequency of the corresponding word in the main row, and a word that is not in the main row of the sub-row Is added to the main line, sub-lines are integrated into the main line, and words in the main line are rearranged in order of word frequency. Characterized in that it comprises a means.
Furthermore, after the synonym dictionary line integration processing is completed, it is possible to provide line rearranging means for rearranging in a fixed order using the first word of the line in the synonym dictionary.
When there is an integrated synonym part in the synonym dictionary, the synonym group capturing means distinguishes the imported synonym part from the integrated synonym part, and the main sub-determining means determines the main synonym part from the integrated synonym part. The sub-line should be the part of the newly imported synonym group. Further, the predetermined condition of the row comparison means can limit the number of comparison target words in the main row, and the number of words that match the sub-row can be a predetermined number or more. The row integration means can also delete the integrated sub rows.
The main row / sub row determination means may create a sub row list including at least two words to be compared in the main row in advance and select a sub row from the list.
Further, the synonym dictionary may be provided with synonym processing means for performing synonym processing for synthesizing synonyms into the same word before working on the synonym dictionary with the synonym group capturing means. .
A program for causing a computer system to realize each function of the above synonym integration system and a recording medium on which the program is recorded are also the present invention.

By arranging the synonym groups according to the configuration of the present invention described above, duplication is greatly reduced, and a synonym dictionary that is easy to see and understand can be created. Moreover, the quality of the dictionary increases with repeated reprocessing.
By using the present invention, it is possible to effectively use a combination of precious words, that is, a synonym group that has been conceived in each situation required by a searcher, which is discarded in most cases. became.

Embodiment of the Invention

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

Since it is considered that the OR group-connected word group can be used as a synonym from the search expression input by the searcher, the synonym group is extracted from the search expression and collected. Consider creating a synonym dictionary from this collected synonym group. Now, this extracted synonym group has the following problems.
1) There are many lines. The synonym group may be, for example, 10 million lines or more, depending on the collection location and period. There are duplicate rows that represent the same concept.
In such a situation where there are many duplications, it is very difficult to see the synonym group for display.
In addition, finding and integrating lines with the same concept requires a lot of computer resources. Development of a procedure that achieves the objective with a small amount of calculation is required.

2) There are many different items that the searcher wants to search, and there are various ways of expressing the searcher, so there are various types of words with the same concept, and what is the same concept? It is difficult to judge.
A synonym of “memory” will be described as an example. In the search formula, there are many examples that are used in an OR combination including the word “memory”, and the synonym group of “memory” in the synonym group from which the OR combination is extracted is used for integration processing. There are cases that make judgment difficult. They are shown below.

a) Combinations of exactly the same words, in word order (in this case, they can be integrated without problems)
Accumulation, storage, storage, memory, storage, recording ---- (2)
Accumulation, preservation, storage, memory, storage, recording ---- (3)

b) Word combinations that are the same as word group (2) but with different word order Memory, storage, recording, storage, storage, accumulation ---- (4)

c) Some words contain the same word group (2) but some words are different Memory, accumulation, storage, memory, import, write ---- (5)

d) Memory, storage, storage of only a part of words in the word group (2) ----- (6)

e) Things that are not synonyms, but contain words whose concepts are slightly separated Memory, memory, storage, storage, registration, file ----- (7)

f) Considering common sense, some words are considered synonyms, but some words are not considered synonyms. , Preservation, narration, conversation, voice ---- (8)

g) Different concepts and different word groups when used.
(In the following example, the word order is diverse, but “memory” is placed at the top for easy understanding.)
Memory, memory, RAM, ROM, HD, non-bola, ---- (9)
Memory, memory, fixed length, variable length ---- (10)
Memory, Memorial, Memories, Events (11)
Memory, register, buffer, latch ----- (12)
Memory, reset, preset, write, read ---- (13)

h) Other concepts. This example is the AD conversion concept line.
Memory, AD conversion, A / D conversion, A / D conversion ---- (14)
i) In the above b) to h), the word arrangement order is various.
In these examples, a group of words including “memory” has been described as an example, but it is also a problem which word to focus on in a row. It is necessary to develop an appropriate method for determining whether or not a certain line and a certain line have the same concept.

3) Handling synonyms.
A set of words such as “display, display, display, display, display, display, display” is called a synonym or notation. Although the character arrangement is different, the words are conceptually 100% identical in concept. Synonyms are special examples that are included in synonyms. Despite the exact same concept, it occupies a large number of cells in the display row and makes the display screen difficult to see.
Further, in the case of the following example, although it is a line to be integrated in the integration process described later, since there are many different words between the lines, it is erroneously determined that it should not be integrated.
Display, Display, Display, Display, Posting ---- (15)
Display, display, display, display, warning ---- (16)
There is also a need to solve this problem.

4) The synonym dictionary after completion of the integrated organization process should be easy to see and use. When creating a synonym search expression, it is required that it is easy to understand which line is important, which word is important, and which word should be used.

5) To perform the integration process, it is necessary to compare at least two lines to determine whether or not to integrate. When the size of the synonym dictionary increases, the number of combinations of two lines is proportional to the square of the number of lines of the synonym dictionary. Even the amount of calculation required for this determination is enormous. Therefore, it is also required to reduce the calculation amount.
The embodiment of the present invention described below solves all the above-mentioned problems.

First, one embodiment of the present invention will be described in detail with reference to FIGS.
FIG. 2 illustrates an outline of a hardware configuration example of the “synonym integration system” of the present invention.
In FIG. 2, a processing device 15 is a central device in a computer system and functions as a synonym integration system by installing a processing program 16 for synonym integration. The storage device 10 can exchange information with the processing device 15, and stores or can store various tables used in the synonym dictionary integration system. The storage device 10 stores a new synonym group 11, an integrated synonym dictionary 12, and other data 14 used for the synonym integration system. The new synonym group 11, the integrated synonym dictionary 12, other data 14, etc. will be described in detail later. In addition to the above, various types of tables and data used during the processing are stored in the storage device 10 as necessary. Note that the new synonym group is an OR combination of the search expressions input from the search terminal 23 of the search system described in the technical background, and is stored in the synonym dictionary. It is the previous one.
The display device 17 displays processing results and the like, and the input device 18 is for giving an instruction to the processing device 15, and is, for example, a keyboard or a mouse.
The configuration of the synonym integration system shown in FIG. 2 may be configured in the server 20 of FIG.

FIG. 3 is a flowchart showing integration processing performed in the system of FIG.
The flowchart in FIG. 3 will be described on the assumption that a newly collected synonym group is added to the “synonym dictionary” that has been integrated in the past and integrated. In the first integration process, the process may be performed assuming that there is no synonym dictionary on which the integration process has been performed.

Now, when displaying a synonym dictionary, it is easy to understand if a matrix-like table format is used. Here, the dictionary and the like will be described in the form of a table. In the table format used below, a synonym group of one concept is represented as one row of the table, and one word is described in one cell in the row. If the lines are different, it is assumed to be another synonym group.
In the present invention, a synonym extracted from an expression used in a search logical expression is handled. Although there may be a group of words that cannot be said to be synonyms, the group of words used in the OR combination in the search logical expression is treated as a synonym. The OR combination is a collection of words that are combined in the meaning of “or” such as +, blank, etc. in addition to OR.
The following is an example of a search logical expression.
Camera + Display + LCD + Shooting ---- (17)
(Camera + Display + LCD + Shooting) * (Guidance + Tourism)-(18)
(Camera + Display + LCD + Shooting) + (Guidance + Tourism) ---- (19)
Camera + Display + LCD + Shooting + Guidance + Sightseeing-(20)
(Camera + Display + LCD + Shooting) Neighborhood symbol (Guidance + Sightseeing) --- (21)
The synonym group is obtained by collecting and arranging a set of OR-linked portions as a group, that is, one row of synonyms from various expressions as described above. In the above search formula, in the case of formula (18), formula (19), and formula (21), it is collected as a synonym group of two lines.
FIG. 4 is a group of synonyms collected from an OR join search expression. This synonym group is created by collecting the search formulas sent from the search terminal 23 in the text search system 20 shown in FIG. In this pre-integrated synonym dictionary, the in-line word order is the order used in the search formula.

First, the synonym group is taken into the synonym dictionary (S110). The synonym dictionary after the import process is shown in FIGS. 5A and 5B.
FIG. 5A is a diagram illustrating synonyms, and a horizontal line indicates a synonym group.
In the synonym dictionary shown in FIG. 5 (when FIG. 5A and FIG. 5B are expressed as a unit, it is shown as FIG. 5. Hereinafter, the word frequency corresponding to each word is the same when there are A and B in other figure numbers). Are present and are shown in FIG. 5B. 5A and 5B are expressed as two tables for the sake of convenience, but if the correspondence between words and word frequencies is taken, the data structure is not in the table format shown here. May be.

Lines 1 to 9 in FIG. 5 are synonym dictionary parts that have been previously integrated, and lines 10 to 34 are parts of new synonym dictionaries added to the synonym dictionary shown in FIG. Is.
In FIG. 5A, column B is the old and new identification tags, 0 in the synonym dictionary part for which the integration process has been completed, and 1 in the newly imported synonym dictionary part for which the integration process has not yet been completed. The A column stores the row number, the F column stores the number of words in the row, and the J column stores the head word frequency. The K column is the position of the first word, and the synonyms of that row are arranged after that. In addition, since the data display of these columns is described so that the process can be grasped, it is not always necessary for the process.
The word frequency corresponding to each word in the synonym dictionary shown in FIG. 5B is set to 1 as an initial value in the newly imported synonym dictionary part. This word frequency is rewritten by the integration process and indicates the number of integrated words.

The processing so far is the preparation for performing the integration processing. From now on, the process of integrating the rows is performed.
The process of integrating lines proceeds from the upper line to the lower line of the synonym dictionary shown in FIG. Now, the process of merging lines is always performed on two lines, the main line and the sub-line. The main line is the main line of processing, and the line (sub-line) of the processing partner A mutual integration process is performed.
First, initial setting of a main row (mn: main row number) and a sub row (sn: sub row number) is performed (S150). In this process, the first main line is the first line (mn = 1) of the integrated synonym dictionary part, and the sub-line is the tenth line (first line of the new synonym dictionary part below it) ( sn = 10). This is because the necessity of performing the integration process on the integrated dictionary portion is low, and the integration processing of the newly incorporated portion and the integrated dictionary portion is mainly performed.

Then, the main line and the sub-line are compared, and it is determined whether or not integration processing is performed (S160).
The number of main line comparison words can be set for the main line. This number of main bank comparison words is set to mg. The main line words to be compared with the sub line words are from the first word to mg words. This is the “main bank comparison word”. When the number of words in the main line is less than mg, all the words are “main line comparison words”.
As will be seen later, the number of words in the main bank increases with integration. For this reason, “main line comparison word” is set.
The value of the main row comparison word number mg can be a constant value during the integration process, for example, mg = 6. Alternatively, this mg value may be all words, or may be adaptively changed according to the attribute of the main line. Examples of adaptive changes include a function of the number of words in a row and a function of the frequency of the first word.

Also in the sub-row, a “number of sub-row comparison words” for limiting the word used for comparison with the main row is set, and this is set as sg. The number of sub-row comparison words is also the number of words from the first word at the left end of the sub-row. If the number of words in the sub-row is less than sg, all words are used.
The number of sub-row comparison words can also be changed adaptively according to a certain value (for example, sg = 6), all words, and sub-row attributes.
In this way, when restricting sub-rows, it is necessary that the order of words in the row be arranged in a certain order of importance, for example, in order of frequency of use. However, in the sub-row where the old and new tags are 1 (new), the word order in the line is disordered and cannot be said to be the conceptual order of the line, so sg is preferably all the words in the line. The new synonym group line is also the main line, but at this time, all the words in the line may be the main line comparison word.

  A line composed of synonyms after integration is a group of words expressing a certain concept. In the present invention, as described later, the first word at the left end of a line is the most concerned line using word frequency. Arranged so that the degree of representing the concept increases, and as it goes to the right sequentially, the degree of representing the concept of the row not only decreases, but also the degree of representing another concept that is not the row concept increases. Yes. For this reason, if too many words are used for determining whether or not integration is possible, the determination may be wrong. This is the reason for limiting the number of words to be compared.

Whether or not integration processing is to be performed is determined by comparing each word in the main row and sub-row to determine the number of matching words. The conditions for whether integration is possible will be described in detail below. Here, the number of words having the smaller main row comparison word and sub row comparison word for each of the main row and the sub row to be compared is set as the comparison word number hn. The number of words that are matched by comparing the main row comparison word and the sub row comparison word is defined as in.
In the following conditions, if the head word frequency of the sub-row is 1, it is all, but this is a row where this row has never been integrated, and the inside of the row remains disordered. It is.
(Judgment condition example 1)
mg = 6, sg: First word frequency 1 is all words, and first word frequency 2 or more is sg = 6.
If the comparison word number hn is 4 words or more, in is 3 words or more, if the comparison word number hn is 3 words, in is 2 words or more, and if the comparison word number hn is 2 words, in is 2 words. Unless it is determined that integration is possible, it is determined that integration is not possible.

(Judgment condition example 2)
mg = 3, sg: first word frequency 1 is all words, first word frequency 2 or more is sg = 3
It is determined that integration is possible because in is 2 words or more (judgment condition example 3)
mg = 4, sg: first word frequency 1 is all words, first word frequency 2 or more is sg = 4
It is judged that integration is possible.

(Judgment condition example 4)
mg = 6, sg: head word frequency 1 is all words, head word frequency 2 or more is sg = 6
The decision to merge is
When the comparison word number hn is 5 words or more: in is 4 words or more When the comparison word number hn is 4 words: in is 3 words or more,
When the comparison word number hn is 3 words: in is 3 words When the comparison word number hn is 2 words: in is 2 words (judgment condition example 5)
mg = 7, sg: first word frequency 1 is all words, first word frequency 2 or more is sg = 7
The decision to merge is
When the comparison word number hn is 5 words or more: in is 4 words or more When the comparison word number hn is 4 words: in is 3 words or more When the comparison word number hn is 3 words: in is 2 words or more The comparison word number hn is 2 words Case: in is 2 words

In addition to the above example, the values of mg, sg, and in can be determined according to the characteristics of the synonym dictionary with reference to the above example of the judgment condition.
In the flowchart of FIG. 3, the determination is performed according to the above-described determination condition example 1.
If it is determined that the integration is impossible (= No = No) in the integration possibility determination (S160) according to the above-described determination condition example 1, the process proceeds to the process of updating the sub-row (S180). If it is determined that integration is possible (= Yes = Yes), the process proceeds to line integration processing (S170).

  In the integration possibility determination (S160), first, as the preliminary integration possibility determination, it may be first checked whether the main line comparison word exists in the sub-line comparison word. Since this process is simpler than the integration possibility determination based on the above-described conditions, the calculation amount for the possibility determination can be reduced by performing this process first. If the predetermined number of identical words does not exist in the main row comparison word and the sub row comparison word, it is immediately determined that integration is not possible (= no).

The line integration processing (S170) in FIG. 3 compares all the words in the main line and the sub line, adds the word frequency of the sub line of the matched word to the word frequency of the word in the main line, and matches the main line. For words in sub-rows that do not have a word, the word is added to the right end of the word row in the main row along with its frequency (the number of words in the main row increases). The word frequencies of the integrated sub-lines are all 0. Now all the words in the sub-line are integrated into the main line along with the word frequency.
The in-line words are rearranged in the order of word frequencies in the line for the integrated main line. In the main bank, there is no duplication of the same word.

Thereafter, the process proceeds to the process of “subrow update, subrow finished?” (S180). In the process of updating the sub-line and determining whether the sub-line is the last (S180), first, the sub-line is lowered to a line where the word frequency of the first word is not 0 (sn = sn + n, n: a positive integer). This is because the line where the word frequency of the first word is 0 is a line integrated with the main line.
Then, it is checked whether or not the sub-line is at the end of the synonym dictionary. If not, the process returns to the line integration possibility determination process (S160). If the integration processing of the sub-rows for one main row has been completed to the end of the synonym dictionary, the main row is updated and the processing of whether the main row is the last (S190).

In the processing of “update main line, end main line?” (S190), the main line is lowered to a line where the word frequency of the first word is not 0 (mn = mn + m, m: positive integer). At this time, the new synonym dictionary part also becomes the main bank. Then, if the main line is not the end of the synonym dictionary, the sub line is determined. At this time, the sub line is a line below the main line, and the first line in the line of the new synonym portion whose head word frequency is not 0 is set as the sub line. Then, the process returns to the row integration possibility determination process (S160).
If the main line is the last line of the synonym dictionary (YES in S190), the line integration process is completed for all lines, and the process goes to a process of deleting unnecessary words / lines (S200).
FIG. 6A and FIG. 6B (also FIG. 6) show synonym dictionaries when the row integration processing is completed for all rows (before going to the unnecessary word / row deletion processing (S200)).

In the unnecessary word / line deletion process (S200), the line with the word frequency 0 of the first word is deleted. This line of word frequency 0 of the first word is a line integrated with the main line. Therefore, it is necessary to delete.
In addition, if the size of the newly introduced synonym dictionary portion is large, the number of in-line words in the main line in which the sub lines are integrated becomes very large. Thus, if there are many words in one line, it may not be suitable as an easy-to-use dictionary. In such a case, it is preferable to adopt a process from the first word to the cn word and cut off unnecessary words. For example, cn = 36, and the 37th word and after are deleted. The word range cn for cutting off the words in the line may also be changed adaptively depending on the attribute, with no cutting, cn being a constant value. Examples of adaptive changes include a function of the number of words in a row, a function of the frequency of the first word and a word frequency, and the like.

The above-described deletion of the first word line with the word frequency 0 and the unnecessary word deletion process may be performed when the line integration process is performed, as will be described later. At this time, it is not necessary to perform the unnecessary word / line deletion process (S200).
It is also possible to delete lines where the first word is not used frequently. For example, a line whose usage frequency of the first word is 3 or less is deleted. Here, “usage frequency” refers to the frequency of use in the created similar word dictionary.
As described above, it is easier to use a dictionary by deleting “words that are less frequently used” and “lines where the first word is less frequently used”.

Then, in each row, the new head word is in order of frequency of use, for example, when the head word is the same, in the order of head word frequency. They are rearranged as follows (S210).
It should be noted that the usage frequency order table may be created not only from the synonym dictionary but also from various text databases. Examples of various text databases include patent gazette databases, newspaper databases, magazine databases, academic journal databases, and others. In the above description, an example of creating a word order table with importance by frequency is illustrated, but a word order table can also be used in order of importance according to other criteria. The order is not necessarily the frequency order, and may be a unique order such as a JIS code order.
When this process ends, the integration process ends.
FIG. 7A and FIG. 7B show synonym dictionaries when the integration process is completed with respect to the example shown in FIG. As shown in FIG. 7, a synonym dictionary having the most important words (words with the highest word frequency in the line) at the top of each line, compactly arranged.

A further feature of synonym dictionary integration according to the present invention is that different conceptual lines become separate lines even with the same head word. In FIG. 7A, since the size of the table is small, the feature is difficult to appear,
In line number 3, map, MAP, map, survey, measurement, coordinates, GPS, ...
Line number 4 is a map, sightseeing, destination, course line number 5 is a map, printing, and camera. Line number 3 is a surveying map itself, and line number 4 is a sightseeing guide No. map ”and line number 5 are lines of the concept of“ map printing related ”.
The synonym dictionary continues to evolve to a better one by repeatedly performing the integration process described above by incorporating it into the synonym dictionary each time a new synonym group is created.
Moreover, it can be expected that the synonym dictionary is further integrated by performing the above-described integration process with the synonym dictionary once created as a target and all lines being new. At this time, it is necessary to set an integration condition so that rows of different concepts are not integrated. For example, it is preferable that the number of comparison words is the same for both the main row and the sub-row, and the judgment conditions are strict. For example, the determination condition 5 described above may be used.

FIG. 8 is a diagram for explaining how the search system of FIG. 1 performs a search using the synonym dictionary created by the above-described processing. First, the synonym dictionary created by the above-described processing is stored in the server 20 or the search terminal 23 of the text search system of FIG. 1 so that the search terminal 23 can search.
FIG. 8 is an example of a screen for creating a search expression displayed on the display device 25 in the search terminal 23 of FIG. The search formula is created by inputting characters from the input device 26 into the “search formula creation area” 34.
Now, it will be described how to create a search expression word using the synonym dictionary created in the present invention on the search input screen shown in FIG.
First, a search word is input into the search word input area 31 for searching the synonym dictionary. For example, when the word “map” is input here and the “word search” button 32 is pressed, the synonym dictionary is searched, and the search result is displayed in the “synonym search result display area” 33. In FIG. 8, a line of a synonym dictionary having a map as the first word is displayed. It is exemplified that the frequency column 36 of the first word is 3 and 1. In the row, words are displayed from left to right from words with high frequency to words with low frequency. The line with the highest first word frequency is displayed in the upper line.
The user who performs the search can refer to the displayed synonyms, copy the necessary words to the “search expression creation area” 34 by cut and paste, and complete an appropriate search expression. Thereafter, a “material search button” 35 can be pressed to search for the material according to the input search formula.

Other embodiments

<Index table>
As described above, in the present invention, in the above-described embodiment, a combination of a main row and a sub-row, which is a target of row integration processing, is extracted by a kind of brute force method, and two rows are compared and integrated. Judgment is made. For this reason, the number of times integration is determined is large, and as the dictionary size increases, the amount of calculation increases and the time until the processing result is obtained increases.
In the processing described below, an index table is created, a combination of two rows (main row and sub row) is taken out using this index table, and row integration processing is performed. Thereby, processing cost can be reduced rather than processing by round robin.

First, the index table will be described. An index table is shown in FIG. The index table shown in FIG. 9 is an index table corresponding to the synonym dictionary shown in FIG. 5 incorporating a new synonym group.
The “index table” (see FIG. 9) has the corresponding word in the Z column as the “entry word”, and the row numbers of all the rows including “the relevant word” in the “synonym dictionary” (FIG. 5) in the row direction. It is a lined table.
In the index table of FIG. 9, the H column is the word number, the Z column is the word, and the J column is the usage frequency. The “display” of word number 1 in the index table of FIG. 9 will be described. All row numbers including “display” of “synonym dictionary” in FIG. As for other relevant words, all line numbers including the word are written. This is an “index table”. This index table is shown in order of use frequency shown in the J line.
The index table may be created once for each integration process, and this is the reason why the processing cost can be reduced by using the index table.

A procedure for determining a pair of a main row and a sub row using an index table in the processing of the flowchart of the integration processing shown in FIG. 3 will be described below. First, in the flowchart of FIG. 3, after taking in a new synonym group, an index table is also created.
During the process of setting the main row (FIG. 3: S150, S190), the index table (FIG. 9) is used to create a table (see FIG. 11) that shows the range of sub-rows for that main row. In the table of the sub-row range, select the sub-row to perform “row integration processing”.

An example of a process of creating a sub-row range table (see FIG. 11) when the sixth line of the synonym dictionary shown in FIG. 5 is the main line (during the process of S190 in FIG. 3) will be described.
The sixth line of the synonym dictionary of FIG. 5 is composed of nine words, but when the number of main comparison words is six (mg = 6), the six words from the first word are used for determining whether or not to integrate. A table (see FIG. 10) containing the main row comparison word is first created from the index table (see FIG. 9).
Since the table in FIG. 10 is a table of rows including main row comparison words, the first six words of the first row are stored in the Z column in the word order in the Z column, and the word numbers (Fig. 9 word number). After the column K, the row numbers of the rows including the main row comparison words are obtained from the index table of FIG. 9 and stored in the table shown in FIG. Since words not included in the main comparison word are ignored, only 6 words are created.

Next, using the table shown in FIG. 10, a sub-row table including N main row comparison words (a table showing the range of sub-rows: see FIG. 11) is created. The table showing the range of sub-rows shown in FIG. 11 includes two words (N = 2) from the main row comparison words (FIG. 10: six words “sightseeing” to “famous place”) in the sixth row of FIG. 5A. Indicates the line number of the line.
In the flowchart of FIG. 3, since the minimum value of the number of matching words in in the determination condition example for integration in the integration possibility determination (S160) is two words, even when the sub-line range is determined, two-word matching (N = 2) ). The value of N needs to be the smallest value of the number of matching words in used in determining whether or not integration is possible. Thus, it is possible to reduce the number of sub-lines that should be determined for integration.
In the sub-line range, the line below the main line (in this case, the sixth line) is always the sub-line, and the main line itself (in this case, the sixth line) is not entered.

In the processing (S150, S180, S190) in which the sub-rows in FIG. 3 are updated, the sub-rows can be sequentially selected from the row numbers in this table, and the integration processing with the main row can be performed.
The table showing the range of the sub-rows shown in FIG. 11 is regenerated whenever the word type of the main row comparison word is changed, for example, when the rows are merged in the row integration process and the words are rearranged. . Further, when the sub-row is integrated with the main row, the row is deleted from the table showing the range of the sub-row in FIG.
In this way, by narrowing down the selection range of sub-rows in advance using the index table, it is possible to limit the sub-rows on which the integration possibility determination (S160) should be performed, and it is not necessary to perform round-robin processing.

<Delete merged sub-row>
When performing row integration processing (FIG. 3: S170), processing cost reduction can also be realized by deleting the integrated sub-row. Unnecessary words can also be deleted by line integration processing.
When this processing is performed, the sub-row integrated in the row integration processing (S170) is deleted in the flowchart of FIG. 3, so that the sub-row is updated (S180) or the main row is updated (S190). In addition, it is not necessary to reduce to the line where the word frequency of the first word is not 0, and it may be the next line. Further, it is not necessary to delete unnecessary lines and words after integration to the last line (see S200).
6A and 6B showing the synonym dictionaries when the integrated sub-rows are deleted by the line integration processing and the final lines are integrated (before S210) are shown in FIGS. 12A and 12B.

<Processing of synonyms>
Synonyms are a special case included in synonyms. A synonym is a set of words with similar concepts, but synonyms are almost 100% identical in concept. Although there are different notations in word groups that differ only in character strings, words that have almost the same concept, including different notations, are called synonyms in the present invention. Depending on the word, it may not be possible to determine whether it is a synonym or a synonym. Therefore, in the present invention, a word group registered in the synonym dictionary is treated as a synonym, and a word group not registered in the synonym dictionary is treated as a synonym.

The next word group is synonymous. For example,
Display, Display, Display, Display, Display, Display --- (21)
It is. These six words are used in exactly the same concept. A line in the synonym dictionary
Display, Display, Display, Display, Display, Display, Display, Posting, (22)
In this case, even if the main comparison word count is set to mg = 6 assuming that the main bank is the main bank, these 6 words are occupied by synonyms and are equivalent to mg = 1. There is a case. Such a line is also displayed when using the synonym dictionary. Words with slightly different character strings using the same concept occupy the same number of synonyms, making the display screen difficult to see and searching for the necessary words. Therefore, many cells must be watched, which increases the searcher's fatigue.

Further, consider the case of the following two lines. However, the frequency in the brackets is the frequency of the word.
Display [1], Display [1], Display [1], Display [1], Posting [1]
--- (23)
Display [1], Display [1], Display [1], Display [1], Guidance [1]
-(24)
In this case, it is determined that the integration is impossible if the determination criterion described in S160 of FIG. 3 is used. These two lines are determined to be unintegrated even though they should be integrated. Since there is such a problem with synonyms, it is necessary to process the synonyms and remove the obstacles.

As for synonyms, a synonym dictionary (word part) is prepared as shown in FIG. 13A and a synonym dictionary (word relative frequency part) as shown in FIG. 13B. A synonym representative word (first word) is defined in each line of the synonym dictionary. What is appropriate as a synonym representative word (first word) is a word that is most frequently used in each row, and is written in the leftmost cell of the row.
The “synonym processing” uses the synonym dictionary, and the word described in the synonym dictionary in the synonym dictionary prior to the process (S110) of importing the synonym dictionary before integration in the flowchart of FIG. Replace with the first word. As a result, when there are a plurality of the same synonym head words on the same line, they can be combined into one synonym head word.
As shown below, the word frequency can be selected according to need between the case of summing up the word frequencies of synonyms and the case of not summing up.
The word frequencies in the above synonym group examples (23) and (24) are as follows by this synonym processing.
(When adding word frequencies after synonym processing)
Display [3], Display [1], Posting [1]-(25)
Display [3], Display [1], Guidance [1]-(26)
These two lines are easy to see when displayed, and even in the integration process, it is possible to determine whether integration is possible in light of the determination criterion 2.
(After synonym processing, word frequency is not added)
Display [1], Display [1], Posting [1] --- (27)
Display [1], Display [1], Guidance [1]-(28)

  The process of replacing the synonym in the synonym dictionary with the synonym head word using the synonym dictionary is not only useful for organizing and synonym of synonym dictionaries, but also useful for general synonym dictionaries including synonyms. A synonym dictionary in which synonyms are replaced with the above synonyms in a synonym dictionary that has been integrated or that has not been integrated, and each synonym is replaced with a synonym representative (first word) Sometimes it is expanded into multiple synonyms using a synonym dictionary. The synonyms developed at this time are not necessarily the same as the original word group before replacement. An example of developing synonym representative words (first words) is shown below.

1) When a word existing in the synonym dictionary is used at the time of creating the search logical expression, it is automatically expanded on the logical expression with reference to the synonym dictionary. Search formula is display + display + posting ---- (29)
At this time, the automatically expanded search formula is as follows when the synonym dictionary is FIG.
Display + Display + Display + Display + Display + Display + Display + Display ---- (30)
It becomes. For example, set a limit on the number of words to be expanded to the top three words,
Display + Display + Display + Display + Posting ---- (31)
It can also be. As a restriction on the number of words to expand,
a) a predetermined number of words from the highest frequency,
b) a word whose frequency is a relative frequency with respect to a synonym representative word (first word) or more,
c) Others can be set as necessary.

  To show the frequency of each word, the word and frequency can be displayed side by side. Words and frequencies may be arranged horizontally or vertically. The searcher copies a necessary word from the displayed synonyms into the “search formula creation area” to construct a material search formula. After the search formula is completed, the database is searched.

2) When a word existing in the synonym dictionary is used at the time of creating the search logical expression, the corresponding line of the synonym dictionary is expanded and displayed along with the frequency on the display device 25 of the search terminal in FIG. To select the necessary synonyms.
Moreover, you may display a synonym dictionary simultaneously with the display of synonyms as shown in FIG.
FIG. 14 shows the synonym of the word “display” and its frequency, which are automatically or manually displayed when a search expression including the word “display” such as the expression (29) is input. It is an example of a screen to display. In FIG. 14, the frequency used for the synonym representative word (first word) is shown, and the relative frequency normalized with the frequency of the synonym representative word (first word) as 1 is shown for the other words. The searcher looks at the display of FIG. 14 and determines how many synonyms are used. Further, even when a synonym is automatically added, displaying a table as shown in FIG. 14 for confirmation gives the search formula creator a sense of security.

It is a figure which shows the structure of the conventional text search system. It is a figure which shows the system configuration example for creating a synonym dictionary. It is a figure which shows the flowchart of an integration process. It is a figure which shows a new synonym group. It is a figure which shows a synonym dictionary (word part). It is a figure which shows a synonym dictionary (word frequency part). Synonym dictionary (word part) after integration to the last line (S190) Synonym dictionary (word frequency part) after integration to the last line (S190) Synonym dictionary (word part) after the integration process is completed (S210) Synonym dictionary (word frequency section) after the integration process is completed (S210) It is a figure explaining the usage example of a synonym dictionary. It is a figure which shows an index table. It is a figure which shows the line | wire where six words exist from the head of the main line (6th line | wire of FIG. 5). It is a figure which shows the range of the sub line with respect to the main line (6th line of FIG. 5). Synonym dictionary (word part) after integration to the last line in the case of deleting the integrated subline (S190) Synonym dictionary (word frequency part) after integration to the last line when deleting the integrated sub-row (S190) It is a figure which shows a synonym dictionary (word part). It is a figure which shows a synonym dictionary (word relative frequency part). It is a figure which shows the screen display of a synonym dictionary.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Storage devices, such as a disk 11 New synonym group 12 Integrated synonym dictionary 14 Other data 15 Processing device 16 Processing program 17 Display device 18 Input device 20 Text database system 23 Search terminal 25 Display device 26 Input device 27 Telecommunication Line 30 Search screen 31 Search term input area 32 Word search button 33 Synonym search result display area 34 Search formula creation area 35 Material search button 36 First word frequency display area

Claims (9)

A synonym integration system for creating a synonym dictionary from a synonym group of a set of OR-linked words of a search expression inputted at the time of search,
Synonym group importing means for setting the word frequency corresponding to each word of the synonym group line to 1, storing the line and importing it into the synonym dictionary;
A main line sub-line determination means having one line of the synonym dictionary as a main line and the other line as a sub-line;
A row comparison means for comparing the main row and the sub row, checking whether a predetermined condition is satisfied, and determining whether to perform integration processing;
As a result of the comparison by the row comparison means, if a predetermined condition is satisfied, a word that is the same as the main row in the sub-row is added to the word frequency of the corresponding word in the main row, and the sub-row A synonym integration system comprising: a line integration unit that adds words that are not in the main line to the main line, integrates the sub lines into the main line, and rearranges the words in the main line in order of word frequency. The synonym integration system according to claim 1,
Further, the synonym integration system further comprising: a line rearrangement unit that rearranges the synonym dictionary in a certain order using the first word of the line in the synonym dictionary after the synonym dictionary line integration processing ends. The synonym integration system according to claim 1 or 2,
If there is an integrated synonym part in the synonym dictionary,
The synonym group capturing means distinguishes the imported synonym part from the integrated synonym part,
The synonym synthesizing system, wherein the main sub-determining means sets a main line from the integrated synonym part, and the sub line is a part of a newly imported synonym group. In the synonym integrated system in any one of Claims 1-3,
The synonym integration system characterized in that the predetermined condition of the row comparison means limits the number of words to be compared in the main row, and the number of words matched with the sub-row is not less than a predetermined number. In the synonym integration system in any one of Claims 1-4,
The synonym integration system, wherein the line integration unit further deletes the integrated sub-line. In the synonym integration system in any one of Claims 1-4,
The synonym integration system, wherein the main row / sub row determination means creates a sub row list including at least two words to be compared in the main row in advance and selects a sub row from the list. In the synonym integration system in any one of Claims 1-6,
In addition, it has a synonym dictionary,
A synonym integration system comprising synonym processing means for performing synonym processing for grouping synonyms into the same word in the synonym dictionary before working on the synonym dictionary by the synonym group capturing means. The program for making a computer system implement | achieve each function of the synonym integrated system in any one of Claims 1-7. The recording medium which recorded the program for making each computer system implement | achieve each function of the synonym integrated system in any one of Claims 1-7.

Images