JPH1115854A - Document processing method, its device and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document processor capable of changing knowledge data following a change in user's interesting object and executing precise judgement. SOLUTION: A keyword extraction part 12 extracts a keyword and its importance from an applied document and stores the extraced contents in a storage part 13 together with the preparation time of the document. A document information/keyword set extraction part 15 extracts a keyword and its importance in a document existing in an object section specified by optional time on a time base applied by a time key input means 14 front the storage part 13, generates knowledge data in the object section based on the extracted keyword and importance and stores the generated data in a document information storing part 16 in the object section. When an unknown document is applied, an evaluation part 17 compares a keyword and importance extracted from the unknown document with knowledge data stored in the storage part 16 to evaluate the unknown document.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document processing method and apparatus, and a recording medium.
A keyword is extracted from a given document, and the characteristics of the document are obtained to generate knowledge data serving as an evaluation criterion, or an unknown document is evaluated based on the generated knowledge data.
The present invention relates to a document processing apparatus capable of estimating a content of a user by determining a degree of coincidence with knowledge data or the like and estimating the content.

[0002]

2. Description of the Related Art Conventionally, document processing using document features has been generalized. For example, in the field of document retrieval, a document of a query keyword is used as a scale for calculating the degree of matching between a query formula and a document. The appearance frequency (indicating the importance of the keyword) and the number of documents in which the keyword appears (indicating the keyword discrimination ability) are used. This makes it possible to extract a document in which desired contents are described from a large number of documents, and to roughly determine whether or not a new unknown document belongs to a certain category.

[0003]

By the way, whether or not the content of a new document is a document of interest to a user is determined, for example, when a large number of mails arrive in an electronic mail document or the like. If the contents of the e-mail can be analyzed on the computer side in advance to determine whether or not the document is of interest (view) to the user, it becomes an index for determining whether or not to view the e-mail. This is preferable because mail can be processed well.

Therefore, it is necessary to judge with high accuracy.
The accuracy of the above-mentioned judgment is largely determined by the quality of the knowledge data serving as the comparison standard. And, the documents that the user needs to view / interest usually change over time. Therefore, knowledge that can be effectively determined at a certain time may not be able to be effectively determined over time. Therefore, it is necessary to maintain knowledge data so that it is always suitable for the user, but since individual users have different interests, it is cumbersome to actually perform maintenance for each user, and it is not practical. I can't offer it.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described background, and has as its object to solve the above-described problems and to provide a user with a time element in addition to the conventional document feature amount. Even if the interests of the user change, the knowledge data can be changed to follow the changes and accurate judgments can be made, and how the interests of the user and vice versa have changed over time. An object of the present invention is to provide a document processing method and apparatus and a recording medium that can be detected.

[0006]

In order to achieve the above object, in a document processing method according to the present invention, a document given as a first invention (implemented in the first embodiment) is provided. A keyword is extracted from the keyword, the importance of the keyword in the document is calculated, and data including the extracted keyword and importance and the creation time information of the given document is stored in the first storage unit. I do. The steps up to this point are realized by the flow shown in FIG. 3 in the embodiment. Next, the first storage unit is accessed based on a target section set by inputting an arbitrary time on a time axis, and the time of creation of the document is set in the data stored in the first storage unit. While extracting the keyword and importance for the document existing in the target section,
Knowledge data in the target section can be generated based on the extracted keywords and importance, and the generated knowledge data in the target section can be stored in the second storage means (claim 1). The steps so far are realized by the flow shown in FIG. 4 in the embodiment.

On the premise of the above method, when an unknown document is given, keywords and importance are extracted from the unknown document, and the keywords and importance of the extracted unknown document are stored in the second storage means. The document may be evaluated by comparing it with the knowledge data stored in the document (claim 2). This step is realized by the embodiment in FIG.

In providing the document, a category is input in association with a category, data to be stored in the first storage unit is divided and stored for each category, and the knowledge data is generated for each category. The information may be stored in the second storage means for each category (claim 3).

When an unknown document is given, keywords and importance are extracted from the unknown document, and the extracted keywords and importance and the knowledge data for each category stored in the second storage unit are extracted. May be compared to determine the degree of coincidence for each category, and unknown documents may be classified based on the degree of coincidence for each category determined (claim 4).

[0010] A first invention suitable for carrying out each of the above-mentioned methods is, for example, a method of extracting a keyword from a given document and calculating the importance of the keyword in the document. Extraction means;
First storage means for storing data including the keyword extracted by the keyword extraction means and the creation time information of the given document (corresponding to “storage section 13 of document information and keyword set” in the embodiment) A time key input unit for inputting an arbitrary time on a time axis, and a time specified by the time key input unit from the data stored in the first storage unit. Generating means for extracting a keyword and importance of a document existing in the target section and generating knowledge data in the target section based on the extracted keyword and importance; and a generating means (in the embodiment, ,
The second storage means (in the embodiment, the "document information storage unit 16 in the target section") that stores the knowledge data in the target section generated by the "document information and keyword set extraction unit 15" (Corresponding to (1)) (claim 8).

Further, when an unknown document is given, a keyword and importance extracted from the unknown document,
Comparing with the knowledge data stored in the second storage means,
It may be configured to further include an evaluation means for evaluating the document (claim 9).

When the document is given, a category is input in association with the document, the data stored in the first storage unit is stored separately for each category, and the generation unit stores the data for each category. The knowledge data may be generated and stored in the second storage means for each category.
0).

Furthermore, when an unknown document is given, the keyword and importance extracted from the unknown document are compared with the knowledge data for each category stored in the second storage means. Classify documents evaluated based on the degree of matching to the category (in the embodiment,
(Implemented by the “evaluation unit 17”).

A second invention for achieving the above object (implemented as a third embodiment) includes:
Based on the above-mentioned claims 1-4, a plurality of different target sections are designated, knowledge data for each of the plurality of designated target sections is generated, and a characteristic of each knowledge data is compared to change a user's interest. (Claim 5).

[0015] As an apparatus for carrying out such a method, the apparatus according to claims 8 to 11 is premised. By designating a plurality of target sections, a plurality of pieces of knowledge data are generated. The feature of the present invention is to further include a function (corresponding to “the user's interest transition calculating unit 19” in the embodiment) for calculating the transition of the user's interest by comparing the features (claim 12).

As a third invention (implemented as a second embodiment) for achieving the above-mentioned object, a keyword is extracted from a given document, and a structure level and a keyword in the document are extracted. The importance of the keyword is calculated according to the frequency of the keyword, and the extracted keyword and the importance are added to the keyword and the importance generated based on the document given in the past by the cumulative calculation to generate the knowledge data. Then, the generated knowledge data is stored in the storage means, and the cumulative operation is to increase the influence of the keyword and importance extracted from the given document (claim 6). ).

On the premise of the above method, keywords and importance are extracted from a given unknown document, and the extracted keywords and importance are compared with the accumulated knowledge data stored in the storage means. The document may be evaluated (claim 7).

An apparatus for performing the method includes a keyword extracting unit that extracts a keyword from a given document and calculates an importance of the keyword in the document, and a keyword extracting unit. Generating means for generating knowledge data by adding the keywords and importances extracted in the above to keywords and importances generated based on documents given in the past by cumulative operation, and knowledge generated by the generating means Storage means for storing data, and wherein the cumulative operation in the generation means is configured to increase the influence of the keyword extracted by the keyword extraction means and the degree of importance (claim 13).

Then, on the premise of the above-mentioned apparatus, the keyword and importance extracted from a given unknown document are compared with the accumulated knowledge data stored in the storage means, and the document is evaluated. It is preferable that an evaluation means is further provided.

Further, as defined in claims 15 to 21, the recording medium may store a program for executing the above-described processes.

[0021] In addition to the document features in the related art, the temporal and semantic properties of the keywords are taken into account, and the keyword set is captured in a time-series manner. It can indicate whether the document is currently the main topic or a rare topic document.

Further, claims 1, 3, 6, 8, 10, 1
According to the inventions of 3, 15, 17, and 20, it is possible to generate a feature amount (knowledge data) of a document that follows a change in user's interest based on a time transition. As a result, claims 2, 4, 7,
According to the inventions of 9, 11, 14, 16, 18, and 21, it is possible to accurately evaluate an unknown document based on the above-described knowledge data, and easily determine whether the document is necessary for the user. It can be carried out.

Further, claims 6, 7, 13, 14, 2
By setting 0 and 21, the storage capacity of the knowledge data can be reduced. Further, by comparing the current keyword set with the keyword set at a certain point in the past, it is possible to express the interest of the user and the transition of the topic. Furthermore,
It is also possible to predict the interests and topics of users in the near future (claims 5, 12, 19).

* Definition of terms The "target section" defined in claims 1 and 8 is, for example, 2
One time may be specified by specifying one time, or one time may be specified, such as from a certain time to the most recent time (from the beginning to a certain time). )), It is a concept that also includes "time".

The keyword extracting means may realize the function of extracting the keyword and the function of determining the importance in separate processing units. In such a case, the plurality of processing units are referred to in the present invention as a whole. Keyword extraction means.

[0026]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Before describing a specific configuration of an embodiment of the present invention, a conceptual embodiment will be described. Generally, when automatically determining whether or not a document is a document of interest to a user, a keyword in the document to be determined is extracted, and the keyword is compared with a dictionary (keyword) created in advance. Thus, the degree of coincidence can be obtained, and the judgment can be made based on the magnitude of the degree of coincidence.

At this time, when extracting a keyword appearing in a document, a time axis is set, and the time at that time and the extracted keyword set are stored (see FIG. 1). Then, by designating a certain time point A and a certain time point B and accumulating the importance of the keyword in the section, it is possible to represent a document of interest to the user or a document that is the opposite. Also, when a new document is read, it is possible to indicate from a set of past keywords whether the document is a document of interest to the user or vice versa. In the embodiment described below, the document received by the user is an e-mail document. However, the same processing can be performed for other documents.

In this embodiment, the received e-mail document is
This is a device for determining whether or not the document is meaningful to the user (a document of interest). FIG. 2 shows a first preferred embodiment of the present invention. As shown in the figure, first, a document information input unit 11 is provided on the input side, and the e-mail document information received by the document information input unit 11 is converted into a keyword extraction unit 12 and a To the storage unit 13. The document information to be given here may be only the e-mail document itself. In addition to this, the input time of the document information and information from the user (category (whether or not interested), etc.) for the document should also be input. You may. The input time may be manually input by the user. However, in the present invention, the time is an important factor. Therefore, when the document information is input, a built-in clock is accessed to set the time. You may make it acquire and give it to the memory | storage part 13 in association. When the document is an e-mail, the input time is most preferably the time when the e-mail is received. However, under the simplification of the system and under normal use, the reception time of the e-mail and the Since there is little difference between the information input time and the information input time, if any, the input time to the apparatus is stored.

The keyword extracting section 12 extracts the keywords of the given electronic mail document information (documents constituting the electronic mail), obtains the importance of the extracted keywords, associates the obtained keywords with the importance, and associates the obtained keywords with the documents. The information and the keyword set are sent to the storage unit 13. Here, in the keyword extraction, for example, after the document is separated and written, the part of speech is analyzed to extract only the noun, which can be used as a keyword. The importance can be determined based on the frequency of appearance of the extracted keywords (nouns) (the higher the frequency of appearance, the higher the importance). It should be noted that the keywords to be finally registered and held may be all extracted ones, or various methods such as those having a determined importance of a certain value or more or only the top n keywords may be used as final keywords. Can be taken. Note that, for such keyword extraction processing and the like, various automatic keyword extraction methods used when making a document into a database can be applied.

Document Information and Keyword Set Storage Unit 13
Stores the document information including the input time of the document information and the keyword (with importance) extracted by the keyword extracting unit 12 in association with each other. When a category indicating the user's interest or vice versa is specified by the document information input unit 11, the storage unit 13 stores predetermined information classified into each category.

On the other hand, a time key input unit 14 for document information is provided as input means.
Reference numeral 4 denotes input of time information as a target period when generating knowledge data for evaluating a new document. The time information is given to the document information and keyword set extraction unit 15 at the next stage. Can be The document information / keyword set extraction unit 15 accesses the document information / keyword set storage unit 13 based on the given time information, and extracts information on the document input within the target period set by the time information. Extract and generate the knowledge data. Then, the generated knowledge data (total of keywords) is stored in the document information storage unit 16 in the target section.
Note that this target section is set as a default value for all sections from the start to the current time.

Further, the present embodiment has an evaluation unit 17 in which a newly received e-mail document is evaluated based on the keyword set and the cumulative importance stored in the document information storage unit 16 within the target section. I do. That is, the evaluation unit 17 compares the “keyword (with importance) of the e-mail document to be evaluated” with the “total of the keyword set and importance stored in the document information storage unit 16” By judging the degree, it is determined whether or not the mail is an object of interest to the user and needs to be viewed. Therefore, in the present embodiment, the e-mail document to be evaluated is also stored in the document information input unit 11.
After the keyword is extracted by the keyword extraction unit 12, the storage unit 1 of the document information and the keyword set
3 is stored. Then, the evaluation unit 17
Keywords (with importance) of the e-mail document to be evaluated stored in both storage units 13 and 16, and knowledge data generated based on document data accumulated up to that time (total of keyword set and importance) Are read and compared to determine the degree of coincidence.

In other words, if the knowledge data is generated based on a given electronic mail document without being categorized in particular, a document that matches a keyword belonging to the keyword set is a document that is likely to be of interest to the user. (Assuming that the user registers a document of interest and interest). Then, it can be said that a keyword having a larger cumulative importance among keywords is more likely to be present in a document of interest.

Therefore, for example, when a keyword of high importance extracted from a document to be evaluated is also extracted as a keyword in the knowledge data and the cumulative importance is large (high coincidence), the user is interested. You can judge it as a certain document. Conversely, if a keyword of high importance extracted from the document to be evaluated is not present in the knowledge data, or if the cumulative importance is large even if it is present, it can be determined that the document does not interest the user (knowledge data Are more likely to be uninterested (lower match). Furthermore, even if the extracted keywords and the keywords stored in the knowledge data match, for example, a keyword having a high importance in the document to be evaluated has a small cumulative sum of importance in the knowledge data, and an important keyword in the document to be evaluated. A sentence in which the magnitude of the importance is reversed, such as a keyword with a low degree of importance being large in the knowledge data, also has a low degree of matching.
According to the system design and actual user usage,
In such a case, the degree of coincidence may be increased.

The specific degree of coincidence may be calculated in view of the above-mentioned matters by comprehensively judging the keyword and the degree of importance (cumulative), and may be obtained by fuzzy inference or another evaluation method, or simply the same. Importance of keywords (evaluation documents) and cumulative importance (knowledge data based on past documents)
, And the sum of the obtained values is obtained, and the larger the sum is, the higher the matching degree is.
When there is only one keyword, various methods such as setting the value to 0 or making the value a subtraction target can be adopted. The processing of the evaluation unit in the present invention is not limited to these arithmetic processing.

The evaluation result obtained by the evaluation unit 17 and the e-mail document and the document information are output to an output unit 18.
Is output via. The output unit 18 is a display (display device) such as a CRT or a printing device such as a printer.

Next, each part will be described in detail while explaining the method of use and operation of the apparatus having the above configuration.

Document Information Collection Step As a preparation for generating knowledge necessary for evaluating an unknown document, document information on a received electronic mail document is collected. Specifically, as shown in FIG. 3, first, the document information (document + input time, etc.) of the received document mail is input using the document information input unit 11 (ST1). In this input process, under normal usage conditions, every time an e-mail is received or an e-mail is downloaded,
Various methods can be adopted, such as inputting all at once for a certain period. By doing so, information is accumulated with the passage of time, and accurate evaluation and determination can be performed.

Next, the keyword extraction unit 12 calculates the importance of each keyword based on the frequency of appearance of the keyword in the document information provided, and generates a keyword set for the time at the time of inputting the document. Then, the generated document information including the keyword set and the time is stored in the storage unit 13 of the stored document information and the keyword set (ST3). Thereby, the information collecting step is completed. If a category indicating the user's interest or the reverse is specified, a document information storage is generated for each category. Steps 1 to 3 described above are repeated each time a new document is input, and are stored in the storage unit 13 of the document information and the keyword set in chronological order.

* Knowledge data generation step This step is to generate, from the document information stored through the above-mentioned document information collection step, knowledge data which is a reference for evaluating an unknown electronic mail. In particular,
As shown in FIG. 4, a time key input unit 1 for document information
4, a certain point on the time axis or a certain section is designated (ST5).

Using the designated section as a key, the document information and the information stored in the keyword set storage unit 13 are searched, and the document information and keywords stored in the corresponding section are extracted (ST6).

Next, the total of the importance assigned to all the extracted keywords is obtained, and the keyword in the section and the obtained total are associated with each other and registered in the document information storage section 16 in the target section (ST7). . This sum may be obtained by simply calculating the average value, for example, by calculating the sum of the degrees of importance assigned to the same keyword and dividing the total by the total number of extracted documents. Can be used. The calculation of the total is performed for all the keywords.

An example of this processing is as follows. First, as shown in FIG. 5, the data structure inside the storage unit 13 of the document information and the keyword set includes a document name for specifying the document, a date and time when the e-mail document is input, and a keyword (with importance). It is assumed that they are stored in association with each other. The specific contents of the document may be stored together, or may be stored in another storage area, and the document name or the like may be linked to the key. Although three documents “A1, D1, D13” are shown in the figure, other documents are also stored.

In such a case, in step 5, [1997/0
5/12 01:10 to 1997/05/30 00:
00] as a time key,
The “time” column in FIG. 5 is accessed, and the one existing in the section specified by the time key is extracted. This allows
As shown in FIG. 6, a target section set by a time key is associated with a keyword (total importance) and stored.

* Evaluation Step of New E-mail Document When a new e-mail is received, as shown in FIG. 7, the new e-mail document is input using the document information input section 14 (ST10). Next, the keyword extracting unit 12 operates, extracts keywords from the document, and generates a keyword set (ST11). The steps so far are the same as steps 1 and 2 in the above (1).

Next, the evaluation unit 17 accesses the document information storage unit 16 within the target section, and compares the keyword sets generated in step 11 (keyword sets extracted from the evaluation document (with importance)) in keyword units. Then, a degree of coincidence is obtained as an index indicating whether the document is a document of interest to the user or a document opposite thereto (S
T12).

Then, the document is classified based on the obtained degree of coincidence. That is, for example, the e-mail document to be evaluated is classified into a document that the user needs to be interested in, or a document that is not interested and need not be viewed. Then, the output unit 8 matches the degree of coincidence and the classification.
(ST13). Therefore, since it is possible to know before reading the contents whether the document is the object of interest of the user or vice versa, it is possible to easily determine whether or not to read the mail.

If a category is input together with a document when generating the knowledge data, the knowledge data is also separated and formed for each category, so that the degree of coincidence is calculated for each category. Then, it is determined that there is a high possibility that the document belongs to the category with the highest matching degree, and the document can be classified as a document belonging to the category.

In such a case, if the category is, for example, work, hobby, private, or the like, an approximate idea of the contents of the e-mail can be obtained, and whether the contents should be viewed immediately or at a later time is acceptable. Makes judgment easier. If the category classification is interesting or not, it is possible to directly determine whether the e-mail document to be evaluated is of interest based on which category has a high degree of matching.
In addition, when the degree of coincidence with either category is low, it can be determined that there is a high possibility that the document is a document that is currently the center of a topic or a document with an unusual topic.

When there are many matching e-mails, they can be classified and displayed by outputting and displaying each attribute. Furthermore, when a search function is added to this embodiment, by adding the current degree of interest of the user to the weight of the query keyword, a search in accordance with the temporal transition of the interest of each user can be performed. It becomes possible.
Furthermore, also in the case where a summary function is added to this embodiment, it is possible to summarize according to the temporal interest transition of each user in the same manner as described above.

FIG. 8 shows a second embodiment of the present invention. The set of keywords in the email document received by the user changes over time. Therefore, in the above-described first embodiment, all the keyword sets for each document are stored, and a keyword in a time section in which a time key is input as necessary (with the cumulative importance) is stored.
To form knowledge. Therefore, if the elapsed time from the start is long, the number of documents to be stored is also large, so that a large storage capacity of the document information and keyword set storage unit 13 is required.

Therefore, in the present embodiment, the keyword sets in the document received by the user are accumulated as needed, so that the user is not required to simply add new information.
The storage capacity can be reduced. For keywords that appear frequently in these keyword sets,
It can be considered that it is well discussed. In addition, a keyword that gradually disappears over time can be given a forgetting function by reducing its importance.

Before describing a specific configuration, the concept (image) will be described. First, each time a new e-mail document is input, the following features (1) and (2) are captured, the e-mail document is evaluated to which it belongs, and classified and output. (1) Documents that are necessary or interesting for the user (2) Documents that are not necessary or not interesting Also, at any time, a keyword set as document information and its importance are accumulated to store keywords for each document. Therefore, the storage capacity can be reduced, and the characteristics of the new document can be reflected in the evaluation of the next document.

Next, a specific configuration will be described. As shown in FIG. 8, a new e-mail document is provided to a keyword extracting unit 22 via a document information input unit 21, where the keywords are extracted and the importance of each keyword is determined using an appearance frequency or the like. Ask. The input unit 21 and the keyword extraction unit 22 can be basically the same as the input unit 11 and the keyword extraction unit 12 in the first embodiment.

The keywords (with importance) extracted by the keyword extracting unit 22 are provided to the evaluation unit 23 and the knowledge updating unit 25. The evaluation unit 23 evaluates a new e-mail document to be evaluated based on the knowledge data generated from the document information provided so far stored in the document information storage unit 24 (for example, whether or not the document is of interest to the user). )I do. In addition, the knowledge updating unit 25 generates new knowledge data by accumulating the document information including the keyword set, the importance, and the time obtained for the new document in preparation for the next evaluation to the document information stored in the past, The document information storage unit 24 is updated.

Further, the evaluation unit 23 uses the output unit 25 to output the evaluation result obtained for the new document, that is, whether or not the document is of interest to the user.

Next, the function of the knowledge updating unit 26, which is a main part of the present embodiment, will be described. A new electronic mail document is input, and the importance of keyword A extracted from the document is input to the following equation, thereby obtaining the importance of keyword A stored in the document information storage unit.

[0058]

(Equation 1) As a result, for keywords that are not extracted in the new e-mail document, the second term of the denominator on the right side of the above equation becomes 0, so that almost half of the importance in the document information storage unit is newly updated. It becomes the importance of the keyword. Therefore, it becomes possible to realize a function of decreasing the importance of a keyword whose use frequency has decreased or is no longer used, that is, a forgetting function. Even if the keyword is extracted as a keyword, the importance of the keyword when newly updated is increased or decreased according to the increase or decrease in the extracted importance. As a result, frequently used keywords that have a tendency to increase in importance in recently received e-mail documents can be determined to have recently increased interest, and keywords that are stored as knowledge data can be determined. It works to increase the importance.

As a specific example, first, the data stored in the document information storage unit 24 immediately before the input of a new document is:
It is assumed that the state is as shown in FIG. Next, it is assumed that a new document is input and the following keyword set is generated by the keyword extracting unit 22. In each case, the value in parentheses is the importance of the keyword in the new document.

[0060]

[Table 1] Then, the knowledge update unit 25 receives the keyword set (with importance) output from the keyword extraction unit 22 and accesses the document information storage unit 24 to acquire the current knowledge data stored therein. A cumulative operation is performed on each keyword to obtain a new importance. As an example of the cumulative calculation, in the case of the keyword "Peru",

[0061]

(Equation 2) Becomes

Similarly, in the case of “Fujimori” which is not extracted as a keyword in a new document,

[0063]

(Equation 3) Becomes

By calculating the other keywords in the same manner, the importance of each keyword is obtained, and the obtained new keyword set (with importance) is stored in the document information storage unit 24. As a result, it is updated as shown in FIG. Here, the importance of the keyword “hostage” decreases, and the importance of the keyword “Aoki”, which is the center of the topic, increases. In the above description, the evaluation unit 23 and the knowledge update unit 25 are configured separately, but the information (data) used is the same, so that the processing may be performed by one processing unit. .

Next, an embodiment of the method according to the present invention, which is an operation using the above-described apparatus, will be described. It is assumed that document information has already been input in the past, and a keyword set (with importance) has been registered in the document information storage unit 24. As in the flowchart shown in FIG. 11, first, new document information is input (ST21).
In response to this, the keyword extraction unit 22 calculates the importance for each keyword based on the appearance frequency or the like, and sends it to the evaluation unit 23 (ST22). The evaluation unit 23 uses the keyword (with importance) for the given new document and the knowledge data generated based on the document information given in the past,
By making a comparison on a keyword-by-keyword basis, it is evaluated whether the new document is a document of interest to the user or a document opposite thereto (ST23). In addition, the knowledge updating unit 25 generates and adds new knowledge data by adding and accumulating a keyword set (with importance) based on the new document,
It is added to and updated in the document information storage unit 24 (ST24).
On the other hand, based on the above-mentioned evaluation result, classification output is performed for each user's interest (ST25).

The above-described processing flow is repeated as needed each time a new document is input, and updates the knowledge data in the document information storage unit.

With the above configuration, it is possible to indicate, at any time, a document that is the object of interest of the user or the opposite, a document that is currently the main topic, or a document with a rare topic. By accumulating new information in the document information storage unit, it is possible to prevent the storage capacity from becoming excessively large as compared with simply adding new information. Thus, the load on the access processing speed to the document information storage unit can be reduced.

FIG. 11 shows a third embodiment of the present invention. In the present embodiment, two or more target sections are compared with each other by using the knowledge data stored in the document information storage unit 16 generated in the first embodiment, so that the user's document information (user's A function to show the transition of interest) is added. It is the user's interest transition calculating unit 19 that calculates the transition, and outputs the calculated result to the output unit 18.

In order to make it easier to understand the intent of this embodiment, a specific example will be described. First, it is assumed that the keywords indicating the interests of the user last month and this month are as follows.

[0070]

[Table 2] The section is, for example, a time key input unit 1 for document information.
4 can be performed by designating two sections where the user wants to see the transition. Alternatively, the input may be automatically input by specifying in advance a section whose transition is to be viewed, such as last month and two months before.

The transition calculating unit 19 acquires a keyword set (with importance) of two sections to be compared, and obtains a keyword vector for each section by a vector space method used in the field of document search technology. As a result, graphs as shown in (1) and (2) in FIG. 12 are obtained, and (2) is obtained from the keyword vector of (1).
It turns out that the interest of the user has shifted to the keyword vector of. Then, for example, such a graph can be output to the output unit 18.

Further, as described above, a keyword aggregate is obtained not only in two sections but also in three or more sections, and a keyword vector in each section is obtained.
It is clearer how the interest of the user has changed over time, and by predicting future changes, for example, we will predict the presence of the keyword vector next month, The predicted keyword set (with importance) is stored in the document information storage unit 16, and the evaluation in the evaluation unit 17 can be performed based on the predicted knowledge data (keyword set).

As described above, by comparing the keyword sets in the target sections, it is possible to extract the user's interest or the reverse temporal transition, and to extract the user's topic temporal transition. Users can extract common topics of users, extract novel keywords that did not appear from a certain point in time, and extract new documents based on them. Can be represented.

Each of the above-described embodiments has been described on the premise that each of the above-described embodiments is incorporated in a device. However, a program for causing a computer to execute each of the above-described processes is provided.
Such a program may be recorded on a predetermined recording medium and provided. For example, the program may be a program for executing each processing step of the flowcharts shown in FIGS. 3, 4, 7, and 11, or may be a program for calculating a user's interest transition. The program to be recorded on the recording medium described above may not be a program for executing all the flowcharts, but may be a part such as, for example, only a program for executing the flowchart of FIG.

As shown in FIGS. 13 and 14, the recording medium is a floppy disk (FD) 30 or a CD-RO.
M31, etc., and the programs stored in the recording media 30, 31 are installed in an HD unit 35 connected (built-in) to a computer 34 via an FD drive 32 or a CD-ROM drive 33. The computer 34 constitutes the device described in each of the above embodiments. Specifically, the storage units 13, 16, 24 and the like in each embodiment are realized by the HD unit 35, and the extraction units 12, 15, 22
And the evaluation units 17 and 23, the knowledge updating unit 25, and the user's interest transition calculating unit 19, for example, each time the HD unit 35 is used.
Is transferred to the internal memory 36, where various processes can be performed at high speed. A document information input unit 1
1 and 21 are realized by a keyboard 37 and a modem 38, and the time key input unit 14 is realized by the keyboard 37 and the like. Further, the output units 18 and 26 correspond to a CRT display (display device) 39 in the illustrated example, but a printer or other various output devices can be used.

[0076]

As described above, in the document processing method and apparatus and recording medium according to the present invention, the user's interests and the like are changed by taking into account the time factor in addition to the conventional document feature amount. Following this, the knowledge data can be changed, accurate judgment can be made, and how the object of interest of the user and vice versa change over time can be detected.

[Brief description of the drawings]

FIG. 1 is a diagram for explaining the concept of a first embodiment of the present invention.

FIG. 2 is a block diagram showing a first embodiment of the present invention.

FIG. 3 is a flowchart illustrating a part of the operation of the apparatus according to the first embodiment.

FIG. 4 is a flowchart illustrating a part of the operation of the apparatus according to the first embodiment.

FIG. 5 is a diagram illustrating an example of a data structure of document information and a keyword set storage unit 13 and registered data.

FIG. 6 is a diagram illustrating an example of a data structure of a document information storage unit 16 and registered data in a target section.

FIG. 7 is a flowchart illustrating a part of the operation of the apparatus according to the first embodiment.

FIG. 8 is a block diagram showing a first embodiment of the present invention.

9A and 9B are diagrams illustrating an example of a data structure of a document information storage unit and registered data, wherein FIG. 9A illustrates data before update and FIG. 9B illustrates data after update, respectively.

FIG. 10 is a flowchart illustrating an operation of the device according to the second embodiment.

FIG. 11 is a block diagram showing a third embodiment of the present invention.

FIG. 12 is a diagram illustrating a function of a user interest transition calculating unit according to the third embodiment.

FIG. 13 is a diagram showing a system configuration for implementing a recording medium according to the present invention.

FIG. 14 is a diagram showing a system configuration for implementing a recording medium according to the present invention.

[Explanation of symbols]

 11 Document Information Input Unit 12 Keyword Extraction Unit 13 Document Information and Keyword Set Storage Unit 14 Time Key Input Unit for Document Information 15 Document Information and Keyword Set Extraction Unit 16 Document Information Storage Unit in Target Section 17 Evaluation Unit 18 Output unit 19 User interest transition calculation unit 21 Document information input unit 22 Keyword extraction unit 23 Evaluation unit 24 Document information storage unit 25 Knowledge update unit 26 Output unit 30 FD (recording medium) 31 CD-ROM (recording medium)

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Kure A-Bridge Omron Co., Ltd. 10 Hanazono Todocho, Ukyo-ku, Kyoto-shi, Kyoto

Claims (21)

[Claims] 1. Extracting a keyword from a given document, calculating the importance of the keyword in the document, and calculating the extracted keyword and importance and the creation time information of the given document. The stored data is stored in the first storage means, the first storage means is accessed based on a target section set by inputting an arbitrary time on the time axis, and the data stored in the first storage means is stored. And extracting a keyword and importance for a document whose creation time of the document is within the target section, and generating knowledge data in the target section based on the extracted keyword and importance. Storing the obtained knowledge data in the target section in a second storage means. 2. When an unknown document is given, a keyword and importance are extracted from the unknown document, and the keyword and importance of the extracted unknown document are stored in the knowledge stored in the second storage unit. 2. The document processing method according to claim 1, wherein said document is evaluated in comparison with data. 3. When the document is given, a category is input in association with the category, data to be stored in the first storage unit is stored separately for each category, and the knowledge data is generated for each category. 2. The document processing method according to claim 1, wherein the information is stored in the second storage unit for each category. 4. When an unknown document is given, a keyword and importance are extracted from the unknown document, and the extracted keyword and importance and knowledge data for each category stored in the second storage unit are extracted. 4. The document processing method according to claim 3, wherein the degree of coincidence for each category is determined by comparing the unknown document with the category based on the determined degree of coincidence for each category. 5. A method of designating a plurality of different target sections, generating knowledge data for each of the plurality of specified target sections, and comparing a characteristic of each knowledge data to obtain a transition of a user's interest. Claims 1-4
The document processing method according to any one of the above. 6. A keyword is extracted from a given document, the importance of the keyword is calculated according to the structure level and the frequency of the keyword in the document, and the extracted keyword and importance are calculated by cumulative operation. The knowledge data is generated by adding the keyword and the degree of importance generated based on the document given in the past, and the generated knowledge data is stored in the storage means. A document processing method, wherein the influence of a keyword and importance extracted from a document is increased. 7. Extracting a keyword and importance from a given unknown document, comparing the extracted keyword and importance with accumulated knowledge data stored in the storage unit, and evaluating the document. 7. The document processing method according to claim 6, wherein 8. A keyword extracting unit for extracting a keyword from a given document and calculating the importance of the keyword in the document; a keyword extracted by the keyword extracting unit; First storage means for storing data including document creation time information; time key input means for inputting an arbitrary time on a time axis; and creation of the document among the data stored in the first storage means. A keyword is extracted for a document existing in the target section specified by the time given by the time key input means, and the importance is extracted. Based on the extracted keyword and importance, the knowledge data in the target section is extracted. Generating means for generating knowledge data; second storage means for storing knowledge data in the target section generated by the generating means; Document processing apparatus characterized by comprising. 9. When an unknown document is given, a keyword and importance extracted from the unknown document are compared with knowledge data stored in the second storage means, and the evaluation of the document is evaluated. 9. The document processing apparatus according to claim 8, further comprising an evaluation unit for performing the evaluation. 10. A function of associating and inputting a category when providing the document, wherein data to be stored in the first storage means is stored separately for each of the categories. 9. The document processing apparatus according to claim 8, wherein the knowledge data is generated for each category and stored in the second storage unit for each category. 11. When an unknown document is given, a keyword and importance extracted from the unknown document are compared with knowledge data for each category stored in the second storage means, and each category is compared. 11. The document processing apparatus according to claim 10, wherein the documents evaluated based on the degree of coincidence with are classified. 12. By specifying a plurality of target sections,
9. A function for generating a plurality of pieces of knowledge data and comparing the characteristics of each piece of knowledge data to obtain a transition of a user's interest.
12. The document processing device according to any one of items 11 to 11. 13. A keyword extracting means for extracting a keyword from a given document and calculating the importance of the keyword in the document, and accumulating the keywords and the importance extracted by the keyword extracting means. Generating means for generating knowledge data by adding a keyword and importance generated based on a document given in the past, and storage means for storing the knowledge data generated by the generating means; and The document processing apparatus according to claim 1, wherein the cumulative operation performed by said generating means has a greater effect on the keywords and importance extracted by the keyword extracting means. 14. When an unknown document is given, a keyword and importance extracted from the unknown document are compared with accumulated knowledge data stored in the storage unit, and the evaluation of the document is performed. 14. The document processing apparatus according to claim 13, further comprising an evaluation unit that performs the following. 15. A process for extracting a keyword from a given document, a process for calculating the importance of the keyword in the document, the extracted keyword and importance, and a creation time of the given document A process of storing data including information in a first storage unit in a computer; accessing the first storage unit based on a target section set by inputting an arbitrary time on a time axis; Extracting, from the data stored in the means, a keyword and importance of a document existing in the target section, and generating knowledge data in the target section based on the extracted keyword and importance; Storing the generated knowledge data in the target section in the second storage means in the computer. Recording medium capable computer read storing a document processing program including. 16. A process of extracting a keyword and a degree of importance from a given unknown document, comparing the keyword and the degree of importance of the extracted unknown document with knowledge data stored in the second storage unit, 16. The recording medium according to claim 15, wherein a program further including a process of evaluating a document is stored. 17. A process of storing data stored in the first storage unit by dividing the data into each category, a process of generating the knowledge data for each category, and storing the knowledge data in the second storage unit for each category. 16. The recording medium according to claim 15, further comprising a program stored therein. 18. When an unknown document is given, a process of extracting a keyword and an importance from the unknown document, the extracted keyword and the importance, and a process for each category stored in the second storage means. 18. The recording medium according to claim 17, further comprising a program for comparing the knowledge data with each other to obtain a degree of coincidence for each category, and a process for classifying an unknown document based on the degree of coincidence for each category obtained. 19. The program according to claim 15, further comprising a process of generating knowledge data for each of a plurality of designated target sections, and a process of obtaining a transition of a user's interest by comparing characteristics of each knowledge data. The recording medium according to the above. 20. A process for extracting a keyword from a given document, a process for calculating the importance of the keyword according to a structure level and a frequency of the keyword in the document, and an influence of the extracted keyword and the importance. A process of generating knowledge data by adding the keywords and importance levels generated based on the documents given in the past by the cumulative operation so that is larger, a process of storing the generated knowledge data in storage means, Computer-readable recording medium storing a document processing program including instructions for causing a computer to execute the program. 21. A process for extracting a keyword and importance from a given unknown document, comparing the extracted keyword and importance with accumulated knowledge data stored in the storage means, and 21. The recording medium according to claim 20, storing a program further including a process of performing an evaluation.