JP6593873B2 - Information analysis apparatus and information analysis method - Google Patents

Description

  The present invention relates to an information analysis technique for acquiring information related to a user's interest.

  With the rapid progress of information and communication technology, the problem of digital divide has been pointed out in recent years. This problem is caused not only by unfamiliarity with IT (Information Technology) equipment but also by the huge amount of information that can be obtained. Even those who can use IT equipment to a certain extent find it very difficult to find useful information from a vast amount of information.

  Patent Document 1 below describes a method for recommending a program according to the user's most recent interest. In this method, life log information including a keyword for identifying an event experienced by the user (visited place, watched TV program, file browsing, music playback, etc.) is acquired, and the keyword in the acquired life log information is accumulated. Then, for each of the accumulated keywords, a degree of memory representing the degree of interest of the user is calculated. The degree of storage is calculated so that the greater the difference between the last date and time when the event occurred and the reference date and time (current date and time), the greater the attenuation, and the smaller the acquisition frequency, the greater the degree of storage. Patent Document 2 below describes a method for extracting event date / time information from homepage information and email information, and notifying event information and related information at an appropriate time before the event date / time or due date. Yes.

JP 2010-266959 A JP 2000-067082 A

  In the above-mentioned patent documents 1 and 2, the viewpoint of managing the user's interest and the degree of interest with respect to each word with high accuracy is not mentioned. Patent Document 2 merely describes the optimization of the notification timing of certain information. In Patent Document 1, the longer the elapsed time from the date and time when the user experienced an event, the greater the attenuation of the storage degree of the keyword specifying the event, and the lower the acquisition frequency, the lower the attenuation of the keyword storage degree. It is only described that it grows. However, the user's interest in each word and the degree of interest are not determined only by such causes, and there is room for improvement in this method.

  This invention is made | formed in view of such a situation, and provides the technique of managing a user's interest with respect to each word and the degree of interest with high precision. In the present specification, “interest” and “interest” are not particularly distinguished.

  Each aspect of the present invention employs the following configurations in order to solve the above-described problems.

  The first aspect relates to an information analysis apparatus. An information analysis apparatus according to a first aspect includes an acquisition unit that acquires word data generated due to a user's action, and stores the user word by associating the word data with a generation date and a scheduled date and time related to the word data. Storage processing means stored in the section, and update means for updating the weight value of the word data based on the relationship between the generation date / time and the scheduled date / time and the current date / time.

  The second aspect relates to an information analysis method executed by at least one computer. The information analysis method according to the second aspect acquires word data generated due to a user's action, and stores the word data in a user word storage unit in association with a generation date / time and a scheduled date / time related to the word data. And updating the weight value of the word data based on the generation date / time and the relationship between the scheduled date / time and the current date / time.

  Another aspect of the present invention is a program that causes at least one computer to execute the method according to the second aspect, or a computer-readable recording medium that records such a program. This recording medium includes a non-transitory tangible medium.

  According to each said aspect, the technique which manages a user's interest with respect to each word and the degree of interest with high precision can be provided.

It is a figure which shows notionally the hardware structural example of the information analysis apparatus (analysis apparatus) in 1st embodiment. It is a figure which shows notionally the process structural example of the information analyzer (analyzer) in 1st embodiment. It is a figure which shows the 1st example of the time change model of the weight value in 1st embodiment. It is a figure which shows the 2nd example of the time change model of the weight value in 1st embodiment. It is a figure which shows the 3rd example of the time change model of the weight value in 1st embodiment. It is a figure which shows the example of the provision destination relation table of keyword DB. It is a figure which shows the example of the query category table of keyword DB. It is a figure which shows the example of the query word table of keyword DB. It is a flowchart which shows the operation example of the information analysis apparatus (analysis apparatus) in 1st embodiment. It is a figure which shows the example of the flow of one operation example of the information analysis apparatus (analysis apparatus) in 1st embodiment, and the processing example of data. It is a figure which shows notionally the process structural example of the information analyzer (analyzer) in 2nd embodiment. It is a figure which shows the example of the time change model of the weight value in 2nd embodiment. It is a flowchart which shows the operation example of the information analysis apparatus (analysis apparatus) in 2nd embodiment. It is a figure which shows the flow of one operation example of the information analysis apparatus (analysis apparatus) in 2nd embodiment, and the example of a data processing. It is a figure which shows notionally the process structural example of the information analyzer in 3rd embodiment.

  Embodiments of the present invention will be described below. In addition, each embodiment given below is an illustration, respectively, and this invention is not limited to the structure of the following embodiment.

[First embodiment]
Hereinafter, an information analysis apparatus and an information analysis method in the first embodiment will be described.
〔Device configuration〕
FIG. 1 is a diagram conceptually illustrating a hardware configuration example of an information analysis apparatus (hereinafter sometimes abbreviated as an analysis apparatus) 10 in the first embodiment. The analysis device 10 is communicably connected to the client device 8 and the information providing site 7 used by the user via the communication network 5. The communication network 5 is a public network such as the Internet, a WAN (Wide Area Network), a LAN (Local Area Network), a wireless communication network, or the like. In the present embodiment, the communication mode among the analysis device 10, the information providing site 7, and the client device 8 is not limited.

  The analysis apparatus 10 is a general computer and has hardware elements as shown in FIG. Specifically, the analysis apparatus 10 includes a CPU (Central Processing Unit) 1, a memory 2, an input / output interface (I / F) 3, a communication unit 4, and the like as a hardware configuration. Each of these hardware elements is connected by a bus.

In addition to a general CPU, the CPU 1 includes an application specific integrated circuit (ASIC), a DSP (Digital Signal Processor), a GPU (Graphics Processing Unit), and the like.
The memory 2 is a RAM (Random Access Memory), a ROM (Read Only Memory), or an auxiliary storage device (such as a hard disk).

  The input / output I / F 3 can be connected to a user interface device such as a display device (not shown) or an input device (not shown). The display device is a device that displays a screen corresponding to the drawing data processed by the CPU 1, such as an LCD (Liquid Crystal Display) or a CRT (Cathode Ray Tube) display. The input device is a device that receives an input of a user operation such as a keyboard and a mouse. The display device and the input device may be integrated and realized as a touch panel.

  The communication unit 4 communicates with other computers via the communication network 5 and exchanges signals with other devices. A portable recording medium or the like can be connected to the communication unit 4. The analysis apparatus 10 may include hardware elements not shown in FIG. 1, and the number of hardware elements is not limited to the number shown in FIG. For example, the analysis device 10 includes a plurality of CPUs 1 and a plurality of memories 2. The hardware configuration of the analyzer 10 is not limited.

  The information providing site 7 is a general computer, and is a server device that provides various types of information such as a WEB server and a database server. The present embodiment does not limit the hardware configuration and software configuration of the information providing site 7. Hereinafter, the information providing site 7 may be abbreviated as the site 7.

  The analysis device 10 is connected to one or more sites 7 so as to be communicable. In the present embodiment, the number of sites 7 accessed by the analysis apparatus 10 is not limited. The analysis device 10 picks up information that each user is supposed to want from the large amount of information provided by the site 7 and provides the user (client device 8) with desired information.

  The client device 8 is a general user terminal such as a PC (Personal Computer), a tablet terminal, a smartphone, a mobile phone, or the like. The client device 8 may be a communication robot having a user interface. The hardware configuration and the software configuration of the client device 8 are not limited as long as the client device 8 can output information provided by the analysis device 10 to the user and can receive input from the user. Information can be output by a display device, a printer device, an audio output device, or the like. In addition, an input from a user is an input device (not shown) that accepts an input of a user operation such as a keyboard and a mouse, a microphone (not shown) that acquires a user's speech, and a camera that captures a user's gesture. Etc. can be received. Hereinafter, the client device 8 may be abbreviated as a client 8.

  The client 8 sends the character string data acquired due to the user's operation to the analysis device 10. Specifically, the client 8 acquires character string data input by a user operation on the input device. For example, any character string data such as a search character string input to the browser, a schedule character string input to the calendar application, and a memo can be acquired.

  The client 8 can also record the user's voice collected by the microphone and send the recorded voice data to the analyzer 10. The client 8 may acquire character string data from the voice data using a known voice recognition technique and send the character string data to the analysis device 10.

  In the present embodiment, as long as the character string data or the voice data is caused by the user's action, the acquisition method of the character string data or the voice data by the client 8 is not limited. For example, an application (including a WEB browser) installed in the client 8 acquires the character string data or the voice data using a known technique, and sends the acquired data to the analysis apparatus 10. Further, the client 8 can extract word data by decomposing the acquired character string data into parts of speech, and can transmit the extracted word data to the analysis device 10.

[Processing configuration]
FIG. 2 is a diagram conceptually illustrating a processing configuration example of the analysis apparatus 10 in the first embodiment. The analyzer 10 has software elements as shown in FIG. Specifically, the analysis apparatus 10 includes an acquisition unit 11, a storage processing unit 12, a user word DB (DataBase) 13, an update unit 14, an extraction unit 15, a keyword DB (DataBase) 16, a determination unit 17, and a search control unit 18. And an output processing unit 19 and the like. Each of these software elements is realized, for example, by executing a program stored in the memory 2 by the CPU 1. In addition, the program is installed from a portable recording medium such as a CD (Compact Disc) or a memory card or another computer on the network via the input / output I / F 3 or the communication unit 4 and is stored in the memory 2. It may be stored.

  When the acquisition unit 11 receives character string data or voice data from the client 8 or another device, the acquisition unit 11 extracts one or more word data from the received character string data or voice data. Specifically, the acquisition unit 11 extracts one or more word data by performing part-of-speech decomposition (morphological analysis) on the received character string data. Various known methods such as MeCab can be used for morphological analysis. For example, the acquisition unit 11 may acquire only word data of specific parts of speech such as nouns and verbs from words of various parts of speech. Furthermore, the acquisition unit 11 can also acquire only word data of specific nouns such as general nouns, proper nouns, and sagittal connection nouns among nouns, and can also normalize the acquired verbs to basic forms.

  The acquisition unit 11 can acquire character string data in which a user's voice is converted into text by applying a known voice recognition technique to the voice data, and can also decompose the part of speech of the character string data. When the part-of-speech decomposition or speech recognition as described above is executed on the client 8, the acquisition unit 11 may receive the plurality of word data extracted by the client 8 from the client 8.

  The acquisition unit 11 further acquires the generation date / time or both the generation date / time and the scheduled date / time regarding the acquired word data.

  “Generation date and time related to word data” means the date and time when the word data was generated or the date and time when the character string data or the voice data that is the source of the word data was generated. The “date and time when word data was generated” includes the date and time when the word data was acquired by the client 8 or the analysis device 10. The “date and time when the character string data or voice data that is the source of the word data is generated” includes the date and time when the client 8 or the analysis device 10 acquires the character string data or the voice data. For example, when the user browses certain information on the client 8, the acquisition unit 11 acquires the viewing date and time as a generation date and time related to word data obtained from the browsed information. When the user operates the client 8 and inputs a character string, the acquisition unit 11 acquires the input date and time as a generation date and time related to word data obtained from the input character string. The “generation date and time related to word data” may be determined by the acquisition unit 11 or may be acquired from the client 8 by the acquisition unit 11.

  The “scheduled date and time related to word data” means the scheduled date and time of the plan represented by the word data or character string data or voice data that is the source of the word data. For example, when the user inputs a certain schedule (plan) on a certain date on the calendar screen, the acquisition unit 11 sets the date as the scheduled date and time related to word data extracted from the character string data input as the schedule. get. Further, when the user browses information on an event on the client 8, the acquisition unit 11 acquires the scheduled date and time of the event indicated by the information as the scheduled date and time related to word data obtained from the browsed information. . The scheduled date and time of the event may be extracted from character string data included in the browsed information using an existing character recognition technique. When the acquisition unit 11 acquires the character string data and the generation date / time of the character string data from the client 8, the acquisition unit 11 sets the generation date / time related to the word data extracted from the character string data to the generation date / time of the character string data. The date / time information extracted from the character string data may be set as the scheduled date / time related to the word data. The “scheduled date and time related to word data” may be determined by the acquisition unit 11 or may be acquired from the client 8 by the acquisition unit 11. In addition, “scheduled date and time related to word data” may not be acquired.

  In this specification, the time granularity represented by the word “date and time” is arbitrary. The wording “date and time” may represent only the date, or may represent the date and time. The time may be represented by an arbitrary time granularity such as a second unit or a millisecond unit.

  The storage processing unit 12 stores the word data acquired by the acquisition unit 11 for each transmission source user or transmission source client 8 in the user word DB 13 in association with the generation date and the scheduled date and time related to the word data. When the scheduled date / time is not acquired, the storage processing unit 12 stores the word data in the user word DB 13 in association with the generation date / time without associating with the scheduled date / time. “Do not associate with the scheduled date and time” includes setting the scheduled date and time field provided in the user word DB 13 to an unset state. Moreover, when the same word data is acquired by the acquisition unit 11 in duplicate with different generation dates and times, the storage processing unit 12 stores the word data in the user word DB 13 in association with a plurality of different generation dates and times.

  The user word DB 13 stores word data generated for each user or for each client 8 due to user actions. In the user word DB 13, separate tables may be provided for each user or for each client 8. The word data may be stored in the user word DB 13 in association with the user identification information or the client 8 identification information. When the user identification information is used, the user identification information may be received from the client 8 by the acquisition unit 11 together with the word data or the data that is the basis of the word data.

  The user word DB 13 stores each word data in association with only the generation date / time and the scheduled date / time acquired with respect to the word data or only the generation date / time. Thus, the word data stored in the user word DB 13 may include word data that is not associated with the scheduled date and time. The state where the scheduled date and time is not associated with the word data is also expressed as no scheduled date and time is set for the word data.

  In the present embodiment, the user word DB 13 stores the same word data having different generation dates and times with respect to one user. For example, when the acquisition unit 11 repeatedly acquires the same word data “fireworks” related to the same user together with different generation dates and times, the storage processing unit 12 acquires the same word data “fireworks” related to the same user as user words as follows: Store in DB13. For example, the user word DB 13 stores word data “fireworks”, generation date and time “August 10 (Monday) 10:25” and scheduled date and time “none” for the user, and further includes word data “fireworks”, The generation date and time “August 10 (Monday) 10:30” and the scheduled date and time “August 28 (Friday) 19:00” are stored.

  The updating unit 14 calculates a weight value for each word data stored in the user word DB 13 at an arbitrary timing. Specifically, the update unit 14 updates the weight value of each word data based on the relationship between the generation date / time and the scheduled date / time stored in the user word DB 13 and the date / time (current date / time) at that time. The updating unit 14 calculates the weight value of each word data so that the weight value indicates the user's interest or degree of interest in the word.

A study in the field of psychology is called the Ebbinghaus forgetting curve. It is a schematic representation of temporal changes in human memory. As shown by this forgetting curve, it is known that human memory fades in inverse proportion to elapsed time. The degree of interest and the degree of interest are also considered to change in the same manner as the memory, and are considered to decrease with time.
In addition, it is considered that the interest and the degree of interest regarding the future schedule will decrease in inverse proportion to the elapsed time from the time when the schedule was created, but will increase again as the date and time of the schedule approaches.

The update unit 14 can approximate the change in the weight value to the change in the interest and the degree of interest by updating the weight value by a method that reflects such psychological characteristics of the human.
Specifically, the update unit 14 uses a time change model that decreases according to each elapsed time from the generation date / time and the scheduled date / time and increases according to a decrease in the remaining time until the scheduled date / time. The weight value of is calculated. The update unit 14 reduces the weight value of the word data for which the scheduled date and time is not set according to the elapsed time from the generation date and time.

  FIG. 3 is a diagram illustrating a first example of a time value model of weight values in the first embodiment. In this first example, a time change model of weight values related to word data for which a scheduled date and time is set is shown. That is, the weight value at the time of generation date and time (0-day elapsed) gradually decreases according to the elapsed time from the generation date and time. The weight value rises as it approaches the scheduled date and time (2 days have passed), and returns to the same value as the generated date and time at the scheduled date and time. Thereafter, the weight value gradually decreases in accordance with the elapsed time from the generation date and time and the scheduled date and time. This time change model approximates a person's interest and a changing tendency of the degree of interest regarding a future schedule.

  In the first embodiment, as described above, the same word data is duplicated and registered in the user word DB 13. That is, the number of generations for each word data is managed. The number of generations for each word data corresponds to the number of user actions that generate each word. Humans are expected to take a lot of actions on information that is of interest and interest. For example, a person searches for, inputs, speaks, and schedules information that is of high interest. That is, a word that is frequently touched by a user is considered a word that is highly interested and interesting for the user.

The update unit 14 can update the weight value by a method that further reflects such psychological characteristics of the human, thereby further approximating the change in the weight value to the change in the interest and the degree of interest of the person. .
Specifically, the update unit 14 calculates the weight value of the word data so as to increase as the number of generation times of word data increases.

  FIG. 4 is a diagram illustrating a second example of a time value model of weight values in the first embodiment. In this second example, word data stored twice in the user word DB 13 at different generation dates is illustrated. The weight value of the word data gradually decreases according to the elapsed time from the first generation date and time, and is set to a value larger than the weight value at the first generation date and time at the second generation date and time. Thereafter, the weight value gradually decreases according to the elapsed time from the first and second generation dates. This time change model approximates the change tendency of interest and the degree of interest of a user who has taken action for the same word a plurality of times.

  Thus, the update unit 14 calculates one weight value for the same word data stored redundantly in the user word DB 13. For example, the update unit 14 calculates a weight value for each generation date and time by applying the generation date and time or the pair of the generation date and the scheduled date and time to the time change model. The updating unit 14 calculates one weight value for the word data stored redundantly by adding the calculated weight values.

Such a calculation method by the update unit 14 can be implemented by the following calculation formula, for example. However, the weight value calculation method used by the updating unit 14 in the present embodiment is not limited to the following (Expression 1), (Expression 2), and (Expression 3). In this example, the weight value calculated for each word data is stored in the weight table in association with the word data. In each of the following expressions, s _i indicates the i-th (i-th row = i-th record) word data stored in the user word DB 13, and s ′ _j is the j-th (j-th row) stored in the weight table. Word = jth record). w _i represents a weight value of the word data s _i , and w ′ _j represents a weight value calculated for the word data s ′ _j . Additionally, date and time of the calculation time (current date) is indicated by d _0, generation date and time are stored in association with the word data s _i to a user word DB13 is indicated by d _i, the date and time is represented by d _'i It is.

That is, in (Formula 1), the weight value w ′ _j of the word data s ′ _j is calculated by the sum of the weight values w _i satisfying s _i = s ′ _j . The weight value w _i of the word data s _i for which the scheduled date / time is not set is calculated by (Equation 2), and the weight value w _i of the word data s _i associated with the scheduled date / time and the generation date / time is (Equation 3) ). (Equation 2) is a mathematical model in which the weight value decreases to 1/3 when one or two days have elapsed and the weight value has decreased to one third when one day has elapsed. (Formula 3) is a mathematical model that becomes maximum when the current date and time coincides with the generation date and time or the scheduled date and time, and decreases as the date and time depart from the generation date and time.

FIG. 5 is a diagram illustrating a third example of a time change model of weight values in the first embodiment. In this third example, the target word data is stored twice in the user word DB 13. In the first time, the word data is stored in association with the generation date / time and the scheduled date / time. In the second time, the scheduled date / time is not set, and the word data is stored in association with the generation date / time different from the first time. . Here, this example is applied to the above equation. The word data registered for the first time is indicated by s ₃ (i = 3), the word data registered for the second time is indicated by s ₇ (i = 7). In the weight table, the word data is 1 Suppose that it is registered in the second. In this case, the weight value calculated when 8 days have elapsed can be expressed by the following equation.
w ′ ₁ = (1/9 + 1/7) + 1 / 3≈0.587

  The update unit 14 does not have to add the weight values calculated for each generation date and time when calculating the weight values of the same word data stored redundantly in the user word DB 13. For example, the update unit 14 applies only the latest generation date / time or only the latest generation date / time / scheduled date / time pair to the time change model, and corresponds to the obtained weight value to the number of generations (stored number) of the word data. The weight value of the word data can also be calculated by adding or multiplying the coefficients.

  Based on the weight value of each word data calculated by the update unit 14, the extraction unit 15 extracts one or more word data from the plurality of word data stored in the user word DB 13 for each user. When a weight table is provided as in the above example, the extraction unit 15 may extract word data from the weight table. The extraction unit 15 extracts word data of words that are highly interested in the user. The extraction unit 15 may extract word data having a weight value equal to or greater than a predetermined threshold, or may extract a predetermined number of word data in descending order of the weight value for each user. The extraction timing by the extraction unit 15 may be an arbitrary timing, and includes a random timing and a predetermined cycle. The extraction unit 15 may extract word data at a timing immediately after the updating unit 14 updates the weight value. The extraction unit 15 may extract word data for all users whose word data is stored in the user word DB 13, or from the user word DB 13 for only the logged-in user based on the login status. Word data may be extracted.

  The determination unit 17 determines an information providing destination and a search keyword based on one or more word data extracted by the extraction unit 15. In the present embodiment, as long as word data extracted based on the weight value is used, the method for determining the information providing destination and the search keyword is not limited. For example, the determination unit 17 may specify a predetermined information providing destination and determine all the word data extracted by the extraction unit 15 as search keywords. The determination unit 17 may determine word data of a specific part of speech such as a proper noun in the word data extracted by the extraction unit 15 as a search keyword.

Hereinafter, a specific example of an information providing destination and a search keyword determination method will be illustrated. However, the determination unit 17 in the present embodiment may determine the information providing destination and the search keyword by a method other than the following method.
The determination unit 17 specifies the site 7 corresponding to the extracted word data as an information providing destination by referring to a keyword DB 16 described later. The determination unit 17 selects word data to be used as a search keyword from the extracted word data by referring to a keyword DB 16 described later.

  The keyword DB 16 includes a provider relationship table, a query category table, and a query word table.

  The providing destination relationship table stores providing destination relationship information indicating the relationship between the providing destination category and the providing destination keyword. Specifically, the provision destination relation table stores identification information of the provision destination category and keyword data of the provision destination keyword in association with each other. Each provision destination category indicates a category of information provided by each site 7 and is associated with the site 7 that provides information of the corresponding category. The identification information of the provider category is associated with the address information of the site 7 separately. The address information of the site 7 is information for acquiring information provided by the site 7 such as a URL (Uniform Resource Locator). The provision destination keyword is a word having some relationship with the provision destination category. For example, a word having a meaning belonging to the provision destination category is stored as the provision destination keyword. In addition, a word indirectly related to the provision destination category may be stored as the provision destination keyword.

  FIG. 6 is a diagram illustrating an example of the provision destination relation table of the keyword DB 16. In FIG. 6, for the sake of easy understanding, the identification information of the provision category is illustrated by a word (noun) indicating the category. However, the identification information of the provider category stored in the provider relationship table may be a numerical value instead of the word illustrated in FIG. In the example of FIG. 6, “go” is associated with the provision destination category “gourmet” as the provision destination keyword. Since the word “go” means going out, it is reminiscent of “meal when going out”. Thus, although the word “go” does not directly belong to the category “gourmet”, it is related to “gourmet” from an indirect relationship (for example, association of “meal when going out”).

  As described above, since the provision destination category is associated with the site 7, the provision destination relationship table indicates the relationship between the word and the site 7 that provides information directly and indirectly related to the word. it can. Therefore, the providing destination relationship table may store the identification information (may be address information) of the site 7 as the providing category identification information.

  The query category table stores information indicating the relationship between the provider category and the query category. Specifically, the query category table stores the identification information of the provider category and the identification information of the query category in association with each other. The provider category stored in the query category table is the same as the provider category stored in the provider relationship table described above. The query category indicates an information category included in information provided from the site 7 corresponding to the provision destination category. In other words, a query category that is not associated with a certain provider category in the query category table indicates a category that is not used as a search keyword for information provided from the site 7 corresponding to the provider category.

  FIG. 7 is a diagram illustrating an example of a query category table in the keyword DB 16. In FIG. 7, for the sake of easy understanding, each identification information of the provider category and the query category is illustrated by a word (noun) indicating the category. However, each identification information of the provision destination category and the query category stored in the query category table may be a numerical value instead of the word illustrated in FIG. In the example of FIG. 7, “dishes name” and “place name” are stored as query categories in the provision destination category “gourmet”. Therefore, it is shown that the words of the categories “dishes names” and “place names” are used as search keywords for the information provided from the site 7 corresponding to “gourmet”. On the other hand, the “food name” is not associated with the provision category “gourmet”. Therefore, it is shown that the word of the category of “food name” is not used as a search keyword for information provided from the site 7 associated with “gourmet”.

  The query word table stores information indicating the relationship between the query category and the query keyword. Specifically, the query word table stores query category identification information and query keyword keyword data in association with each other. The query category stored in the query word table is the same as the query category stored in the above-described query category table. The query keyword is a word having a meaning belonging to the query category.

  FIG. 8 is a diagram illustrating an example of a query word table of the keyword DB 16. In FIG. 8, for easy understanding, the identification information of the query category is illustrated by a word (noun) indicating the category. However, the query category identification information stored in the query word table may be numeric values instead of the words illustrated in FIG. In the example of FIG. 8, “Tomorrow”, “Tomorrow”, and “June 8” are associated with the query category “date name” as query keywords.

  As described above, the query category table and the query word table are associated with each other by the query category. Therefore, the query category table and the query word table can also be expressed as query relation information. In addition, the query category table and the query word table can be integrated using the query category. In this case, the integrated table shows the relationship between the provision destination category and the query keyword. By providing the query category table and the query word table separately as in this embodiment, it is not necessary to store duplicate query keywords in the table, leading to a reduction in memory resources. Moreover, the merit that it is easy to manage the data of each table can also be enjoyed.

  The user word DB 13 and the keyword DB 16 may be relational databases or files written in a predetermined description language such as JSON (JavaScript (registered trademark) Object Notation) or XML (Extensible Markup Language). Also good.

  The deciding unit 17 refers to the above-mentioned provision destination relation table, and specifies a provision destination category corresponding to one or more word data extracted by the extraction unit 15. For example, the determination unit 17 determines whether the word data extracted by the extraction unit 15 is stored as a provision destination keyword in the provision destination relation table of the keyword DB 16. The determination unit 17 specifies a provision destination category in which one or more word data is stored as a provision destination keyword. The determination unit 17 may specify a plurality of provision destination categories. Further, when there are a plurality of provision category categories in which one or more word data are stored as the provision destination keywords, the determination unit 17 specifies one provision destination category in which the most word data is stored as the provision destination keywords. May be.

  Further, the determination unit 17 may calculate a similarity between the feature amount of the provision destination keyword and the feature amount of each word data extracted by the extraction unit 15, and specify the provision destination category using the similarity. it can. Various known methods can be used as a method for expressing the feature amount of word data and a method for calculating similarity between words. The determination unit 17 converts each word data extracted by the provision destination keyword data and the extraction unit 15 into feature amount data, and calculates a similarity between the converted feature amount data. The feature amount data may have a vector structure. In the present embodiment, the method for calculating the similarity between words is not limited. However, it is desirable to use a method in which the closeness of meaning and concept is calculated with high accuracy as the similarity.

  The determination unit 17 can also specify the provision destination category based on the distance of the similarity calculated as described above. As another example, the determination unit 17 may specify a provision destination category associated with a provision destination keyword that is equal to or higher than a predetermined threshold and shows the highest similarity. Further, the determination unit 17 may specify a provision destination category having the largest number of provision destination keywords that indicate a degree of similarity equal to or greater than a predetermined threshold. Moreover, the determination part 17 may specify the provision destination category with the highest average similarity of the provision destination keyword for every provision destination category. The determination unit 17 may specify a provider category whose average similarity is equal to or greater than a predetermined threshold.

  The determination unit 17 may hold feature amount data calculated in advance for a large number of words using the above-described known technique. Such feature amount data for each word may be generated by a device other than the analysis device 10. In this case, the determination unit 17 replaces each word data extracted by the extraction unit 15 with feature amount data using a table of feature amount data for each word. In the keyword DB 16, feature amount data corresponding to each of the providing destination keyword and the query keyword may be stored in advance. The provision destination relation table may store average feature amount data of a plurality of provision destination keywords for each provision destination category. In this case, the determination unit 17 specifies the provision category by calculating the similarity between the feature amount data of each word data extracted by the extraction unit 15 and the average feature amount data for each provision destination category.

  The determination unit 17 selects word data to be used as a search keyword from one or more word data extracted by the extraction unit 15. Specifically, the determination unit 17 is a query in which each word data extracted by the extraction unit 15 is associated with the provision destination category specified by the determination unit 17 based on the query category table and the query word table of the keyword DB 16. Whether it belongs to a category is determined. Based on this determination, the determination unit 17 selects word data to be used as a search keyword from one or more word data extracted by the extraction unit 15. Hereinafter, the query category associated with the provider category specified by the determination unit 17 may be referred to as “target query category”. When a plurality of supply destination categories are specified, the determination unit 17 selects word data to be used as a search keyword from among the plurality of word data extracted by the extraction unit 15 for each of the specified plurality of supply destination categories. Select each one. In other words, when different word data is selected for each provider category, when the same word data is selected in common for all provider categories, or when there is a provider category where no word data is selected. There can be.

  For example, the determination unit 17 extracts the target query category from the query category table, and whether or not each word data extracted by the extraction unit 15 matches the query keyword associated with the target query category in the query word table. Respectively. The determination unit 17 selects word data that matches the query keyword associated with the target query category as a word data candidate to be used as a search keyword, and excludes other word data from the candidate.

  Further, the determination unit 17 calculates the similarity between the query keyword shown in the query word table and the feature amount of each word data extracted by the extraction unit 15, and uses the calculated similarity to extract the extraction unit 15. The determination may be performed for each word data extracted by the above. The determination unit 17 converts the query keyword data and each word data extracted by the extraction unit 15 into feature amount data, and calculates the similarity between the converted feature amount data. The notation method of the feature amount of the word data and the similarity calculation method between the words are as described above.

  For example, when a query keyword whose similarity with word data indicates a predetermined threshold or more is associated with the target query category, the determination unit 17 determines that the word data belongs to the query category. When the query word table stores average feature amount data of a plurality of query keywords for each query category, the determination unit 17 selects word data whose similarity with the average feature amount data of the target query category is a predetermined threshold value or more. It is determined that it belongs to the query category.

  The search control unit 18 specifies the site 7 corresponding to the provision destination category determined by the determination unit 17 and searches using the word data selected by the determination unit 17 for the information provided from the site 7 as a search keyword. Request. If there is a provision category for which no word data has been selected, the search control unit 18 does not specify a search keyword for the search for information provided from the site 7 corresponding to the provision category. Also good.

  If the identified site 7 supports in-site search, the search control unit 18 requests the identified site 7 to perform a search using the selected word data as a search keyword, and the search result. The information extracted as is received from the site 7. In addition, when the specified site 7 does not support search within the site, a computer other than the site 7 acquires all the information provided by the site 7 from the site 7, and for the acquired information The search may be performed. In this case, the search control unit 18 requests a search to the computer that performs the search. At this time, the search control unit 18 sends the address information of the site 7 and the word data selected as the search keyword to the computer. The computer that performs the search may be the analysis apparatus 10.

  The output processing unit 19 transmits information acquired as a search result by the search control unit 18 to the client 8. As a result, the client 8 outputs the information. The client 8 can output the information provided from the analysis device 10 by various methods such as display, printing, and voice.

[Operation example / Information analysis method]
Hereinafter, the information analysis method in the first embodiment will be described with reference to FIG. FIG. 9 is a flowchart showing an operation example of the analyzer 10 in the first embodiment. As shown in FIG. 9, the information analysis method in the first embodiment is executed by at least one computer such as the analysis apparatus 10. Each process shown in FIG. 9 is the same as the processing content of each software element included in the analyzer 10, and therefore details of each process are omitted as appropriate.

  The analysis apparatus 10 acquires word data generated due to the user's action (S91). Specifically, the analysis device 10 receives character string data or voice data from the client 8 and extracts word data from the received character string data or voice data using a known technique. This method of acquiring word data is as described in the description of the acquisition unit 11. The analysis device 10 may receive the word data itself from the client 8.

  The analysis apparatus 10 acquires the generation date / time and the scheduled date / time regarding the word data acquired in (S91) or the generation date / time (S92). The scheduled date and time may not be acquired. The meaning of the generation date / time and the scheduled date / time regarding the word data and the acquisition method thereof are as described in the description of the acquisition unit 11.

  The analysis apparatus 10 stores the word data acquired in (S91) in the user word DB 13 in association with the generation date / time and the scheduled date / time acquired in (S92) (S93). When the user word DB 13 is provided for each user, the analysis apparatus 10 stores the word data in the user word DB 13 corresponding to the user who generated the word data. The analysis device 10 may store the word data in the user word DB 13 together with the identification information of the user or the client 8.

  The analysis device 10 calculates a weight value for each word data stored in the user word DB 13 based on the generation date and the scheduled date and time associated with the word data (S94). The method for calculating the weight value is as described in the description of the updating unit 14. Specifically, for the word data for which the scheduled date and time is not set, the analysis device 10 calculates the weight value of the word data so as to decrease according to the elapsed time from the generation date and time. For the word data for which the scheduled date and time is set, the word data is reduced so that it decreases with each generation time and the elapsed time from the scheduled date and time and increases with the decrease in the remaining time until the scheduled date and time. A weight value is calculated. Further, for the word data acquired a plurality of times, the analysis device 10 calculates the weight value of the word data so as to increase as the number of times the word data is generated.

  The analysis device 10 extracts one or more word data for each user from the word data stored in the user word DB 13 based on the weight value of each word data calculated in (S94) (S95). The word data extraction method is as described in the description of the extraction unit 15.

  The analysis apparatus 10 determines an information providing destination based on the word data extracted in (S95) (S96). Further, the analysis apparatus 10 determines a search keyword based on the word data extracted in (S95) (S97). The method of determining the information providing destination and the search keyword is as described in the description of the determination unit 17.

  The analysis apparatus 10 requests a search using the word data determined in (S97) for the information provided from the site 7 determined as the information providing destination in (S96) (S98). The analysis device 10 may request a search from the identified site 7 or may request a search from another computer. Further, the analysis device 10 may perform a search by itself.

  The analysis apparatus 10 causes the client 8 to output information extracted as a result of the search (S99).

  Next, a specific example of the information analysis method in the first embodiment will be further described. FIG. 10 is a diagram illustrating a flow of one operation example of the analysis apparatus 10 and a data processing example in the first embodiment. Here, it is assumed that the user inputs the first schedule “buy pasta” and the second schedule “eat pasta” on the calendar screen displayed on the client 8.

  In this case, the analysis apparatus 10 receives two character string data “buy pasta” and “eat pasta” from the client 8 (S101). At this time, the analyzer 10 can acquire the time when the character string data is received and the schedule information on the calendar to which the character string data is input.

  The analysis apparatus 10 decomposes the received two character string data, respectively, to obtain noun word data “pasta” and verb word data “buy” and “eat” (S102). Here, the particle word data “O” is excluded from the processing target. Furthermore, the analysis apparatus 10 sets the generation date and time related to the word data “pasta” and “buy” to the date and time when the character string data “buy pasta” is received, and sets the scheduled date and time as the character string data “buy pasta”. "Is set to the scheduled date indicated by. The analysis apparatus 10 sets the generation date and time related to the word data “pasta” and “eat” to the date and time when the character string data “eating pasta” is received, and the character string data “buy pasta” is set as the scheduled date and time. Set to the scheduled date shown.

  The analysis apparatus 10 stores each word data acquired in (S102) in the user word DB 13 in association with the generation date / time and the scheduled date / time related to the word data (S103). In the example of FIG. 10, the word data “pasta” and “buy” are stored in association with the generation date “June 6, 2015” and the scheduled date “June 13, 2015”, and the word data “pasta”. And “eat” are stored in association with the generation date “June 7, 2015” and the scheduled date “June 13, 2015”.

  The analysis device 10 calculates a weight value for each word data stored in the user word DB 13 (S104). The analysis apparatus 10 calculates the weight values of the word data “buy” and “eat” by applying the generation date / time and scheduled date / time pairs to the time change model, respectively. In the example of FIG. 10, since the generation date and time of the word data “eat” is later than that of the word data “buy”, the weight value of the word data “eat” is larger than that of the word data “buy”. It is a value. Here, two pieces of word data “pasta” are stored in association with different generation dates and times. Therefore, the analysis device 10 applies the generation date / time and scheduled date / time pairs to the time change model, thereby calculating the weight value for each generation date / time, and adding the calculated weight values to each word. One weight value is calculated for the data “pasta”.

  The analysis apparatus 10 extracts word data whose weight value satisfies a predetermined condition from a plurality of word data stored in the user word DB 13 (S105). In the example of FIG. 10, the analysis apparatus 10 extracts word data “pasta” having a weight value “0.83” greater than a predetermined threshold “0.8”.

The analysis apparatus 10 determines an information providing destination (search site) and a search keyword (query) corresponding to the word data extracted in (S105) (S106).
The analysis apparatus 10 requests the search site determined in (S106) based on the query determined in (S106). The analysis apparatus 10 causes the client 8 to output information obtained as a result of the search (S107). In the example of FIG. 10, information related to “pasta” is output.

  9 and 10 illustrate a series of operation flows. However, the operation of the information analysis method and the analysis apparatus 10 in the first embodiment may not be such a continuous flow. For example, in FIG. 9, the processing flow of (S91), (S92) and (S93), the processing of (S94), and the processing flow after (S95) may be executed in parallel. Further, the processing flow after (S94) and the processing flows of (S91), (S92), and (S93) may be executed in parallel. Further, (S96) and (S97) may be executed in the reverse order or may be executed in parallel.

[Operation and effect of the first embodiment]
As described above, in the first embodiment, the word data generated due to the user's action is stored in the user word DB 13 for each user who generated the data. At this time, the word data is stored in association with the generation date and the scheduled date and time related to the word data. Based on the relationship between the generation date / time and the current date / time or the relationship between the generation date / time and the scheduled date / time and the current date / time, the weight value of each stored word data is calculated.

  In the first embodiment, the weight value of the word data is calculated so as to decrease with each elapsed time from the generation date and time and the scheduled date and time and to increase with a decrease in remaining time until the scheduled date and time. . As a result, the calculated weight value reflects the human characteristic that interest and interest decrease as the date passes, and the human characteristic that interest and interest increase as the scheduled date approaches.

  Furthermore, in the first embodiment, when word data indicating the same word is acquired with different generation dates and times, the word data is associated with each generation date and stored in the user word DB 13 in duplicate. Then, the weight value of the word data is calculated so as to increase as the number of word data generations increases. Thus, the calculated weight value is a value reflecting the user's interest and the degree of interest expressed by the number of times the user's action is performed on the word data.

  Thus, since the weight value is calculated by a method reflecting the interest and the temporal change related to the interest and the human behavioral characteristics, according to the first embodiment, the interest that the human holds the change in the weight value of each word data. And a change in the degree of interest. That is, according to the first embodiment, it is possible to manage the user's interest in each word and the degree of interest with high accuracy.

  Furthermore, in the first embodiment, one or more word data is extracted from a plurality of word data generated due to user behavior based on the weight value of each word data calculated as described above. Is done. Since each weight value reflects the degree of interest and interest of the user with high accuracy, the extracted word data is data of a word that is highly interested (strongly interested) by the user. In the first embodiment, the information providing destination and the search keyword are determined based on the extracted word data, and the search using the determined search keyword is performed on the information provided from the information providing destination. Information obtained as a result of the search is provided to the client 8 (user). As a result, information related to words of high interest to the user is provided to the user.

  Therefore, according to the first embodiment, it is possible to provide highly interested information at an appropriate timing for each user. Here, the appropriate timing means the time when the user is highly interested. As described above, the user's interests and interests decrease with the passage of time, so even a topic that is highly impressed at a certain point in time does not necessarily have a high interest at a certain point in time. In addition, even if the interest or interest in a topic related to a future schedule is low at a certain point in time, it increases again as the schedule approaches. According to the first embodiment, according to such a change in the interest and the degree of interest that a person has, information that the user is highly interested in can be provided when the interest is high.

[Second Embodiment]
In the second embodiment, the weight value of the word data is updated by further considering the user's evaluation on the information extracted by the search. Hereinafter, the information analysis device and the information analysis method in the second embodiment will be described focusing on the contents different from those in the first embodiment. In the following description, the same contents as those in the first embodiment are omitted as appropriate.

The hardware configuration of the analyzer 10 in the second embodiment may be the same as in the first embodiment.
[Processing configuration]
FIG. 11 is a diagram conceptually illustrating a processing configuration example of the analysis apparatus 10 in the second embodiment. The analyzer 10 in the second embodiment has software elements as shown in FIG. Specifically, the analyzer 10 further includes an evaluation acquisition unit 21 in addition to the processing configuration of the first embodiment. The evaluation acquisition unit 21 is also realized, for example, by executing a program stored in the memory 2 by the CPU 1. Further, the program may be installed from a portable recording medium such as a CD or a memory card or another computer on the network via the input / output I / F 3 or the communication unit 4 and stored in the memory 2. Good.

  The output processing unit 19 transmits information acquired as a search result by the search control unit 18 to the client 8. At this time, the output processing unit 19 can also transmit a screen for allowing the user of the client 8 to input the evaluation result of the information together with the information acquired as the search result. For example, the input screen may include a screen element for selecting whether it is good or not, and an evaluation score (for example, one of six levels from the best (5 points) to unnecessary (0 points)). May be included.

  The evaluation acquisition unit 21 acquires user evaluation information for the information acquired as a search result by the search control unit 18. The user evaluation information is information indicating the necessity or height of the provided information for the user. For example, the evaluation information of the user may be indicated by a binary value indicating whether or not it is necessary, or may be indicated by an evaluation value indicating the level of necessity. When the user evaluation is input by the operation of the client 8, the evaluation acquisition unit 21 acquires the input information as user evaluation information. Moreover, the evaluation acquisition part 21 can also acquire as a user evaluation information whether the user selected the summary of the provided information and the user browsed the detailed information. For example, the output processing unit 19 displays a screen of a summary list of information obtained as a search result on the client 8, and the evaluation acquisition unit 21 can determine a user's browsing status based on information regarding a selection operation on the screen. .

  The evaluation acquisition unit 21 specifies the search that is the target of the acquired evaluation information, and sets the evaluation value indicated by the evaluation information as the evaluation value of the word data that is the basis of the search. The word data used as the basis of the search means word data referred to in order to determine an information providing destination and a search keyword used for the search. The evaluation value of the word data may be stored in the user word DB 13 in association with the word data, or may be stored in another table together with the word data and the weight value calculated by the update unit 14.

  For example, the search for the evaluation information and the word data for which the evaluation value is to be set are specified as follows. For each search performed, the search control unit 18 holds information related to the identification information of the target user (or the client 8), the word data used as the basis of the search, and the information obtained by the search. The evaluation acquisition unit 21 acquires the relationship information between the identification information of the user (or client 8) and the information that is the target of the evaluation information together with the evaluation information. Based on the relationship information, the evaluation acquisition unit 21 can specify a search that is a target of evaluation information, and can specify word data for which an evaluation value is to be set.

The update unit 14 further updates the weight value of each word data in consideration of the evaluation information acquired by the evaluation acquisition unit 21. The updating unit 14 calculates the weight value so as to increase according to the necessity for the user indicated by the evaluation information. Such a calculation method by the update unit 14 can be implemented by the following calculation formula, for example. However, the weight value calculation method used by the updating unit 14 in the present embodiment is not limited to the following (Formula 4).

In the above (Formula 4), e _j represents an evaluation value for the word data s ′ _j . For example, e _j may, if not rated "1", if positive reviews "2", if negative rating takes a value such as "0". W _i indicating the weight value of the word data s _i is calculated by the above (formula 2) or the above (formula 3). According to the above (Formula 4), the weight value of the word data obtained with a positive evaluation is calculated higher than the word data obtained with a negative evaluation and the unevaluated word data.

FIG. 12 is a diagram illustrating an example of a time change model of weight values in the second embodiment. In this example, the weight value is set to 0 based on the evaluation information indicating the negative evaluation of the user when 4 days have elapsed. In this example, when the evaluation value e _j indicating a negative evaluation is set to "0", the same as the weight value calculated by the above equation (4). In addition, the weight value is increased based on the evaluation information indicating the positive evaluation of the user when 8 days have passed. In this example, when the evaluation value e _j indicating a positive evaluation is set to "2", the same as the weight value calculated by the above equation (4), has been updated weight values doubled .

[Operation example / Information analysis method]
Hereinafter, the information analysis method in the second embodiment will be described with reference to FIG. FIG. 13 is a flowchart showing an operation example of the analyzer 10 in the second embodiment. As shown in FIG. 13, the information analysis method in the second embodiment is executed by at least one computer such as the analysis apparatus 10. Each step shown in FIG. 13 is the same as the processing content of each software element included in the analysis apparatus 10, and therefore details of each step are omitted as appropriate.

  Steps (S91) to (S99) shown in FIG. 13 are the same as the steps in the first embodiment shown in FIG.

  In the second embodiment, the analysis apparatus 10 acquires user evaluation information for the information output as a search result in (S99) (S131). The analysis apparatus 10 is based on the identification information of the client 8 or the user who is the transmission destination of the information in (S99), the information transmitted in (S99), and the search basis in (S98) (in (S95)) Information on the relationship with the extracted word data can be held. The analysis apparatus 10 confirms that the evaluation information acquired in (S131) corresponds to the information transmitted in (S99), and determines word data related to the information as word data for which an evaluation value is to be set. . The analysis device 10 stores an evaluation value corresponding to the acquired evaluation information in association with the determined word data.

  The analysis device 10 further uses the stored evaluation value to update the weight value of the word data (S132). The calculation method of the weight value at this time is as described in the description of the updating unit 14.

  Next, a specific example of the information analysis method in the second embodiment will be further described. FIG. 14 is a diagram illustrating a flow of one operation example of the analysis apparatus 10 and a data processing example in the second embodiment. The precondition is the same as the scene shown in FIG. Further, (S101) to (S107) are the same as in FIG.

  The analysis apparatus 10 causes the client 8 to output pasta-related information as a search result (S107), and obtains user evaluation information for the information (S141). This user is a user who provides information in (S107) and a user who generates the character string data acquired in (S101). Here, it is assumed that the acquired evaluation information indicates a positive evaluation. Accordingly, the analysis apparatus 10 stores the evaluation value “2” corresponding to the evaluation information in association with the word data extracted in (S105).

  The analysis apparatus 10 further uses the evaluation value “2” to update the weight value of the word data (S104). Here, the analysis apparatus 10 calculates a new weight value “1.66” by multiplying the original weight value “0.83” by the evaluation value “2”. As a result, the pasta related information is highly useful for the user, and the pasta related information is provided again.

  The operation of the analysis apparatus 10 and the information analysis method in the second embodiment are not limited to the examples shown in FIGS. The analysis device 10 may not update the weight value of the word data corresponding to the evaluation information at the timing when the user evaluation information is acquired. In this case, after (S131), the analysis apparatus 10 stores the evaluation value corresponding to the acquired evaluation information in association with the word data and ends. When the analysis device 10 updates the weight value at an arbitrary timing, the weight value of the word data in which the evaluation value is stored may be calculated by further using the evaluation value.

[Operation and effect of the second embodiment]
As described above, in the second embodiment, information acquired as a search result is provided to the user (client 8), and user evaluation information for the information is acquired. Then, in addition to the generation date / time, the scheduled date / time, and the generation count (storage count), the weight value of each word data is calculated by further considering the acquired evaluation information. As described above, in the second embodiment, not only human psychological characteristics and behavioral characteristics related to interest, but also preference information directly fed back from the user to the search result is further reflected in the weight value of each word data. The

  Therefore, according to the second embodiment, since the weight value of the word data can be learned so as to be a value that can provide information that suits the user's preference, the user's interest and degree of interest in each word can be determined. It can be managed with higher accuracy. As a result, it is possible to provide only information that is highly interested in the user, that is, information that matches the user's preference.

[Modification of Second Embodiment]
The provision destination relationship table included in the keyword DB 16 described above may indicate the relationship between a plurality of information provision destinations and a plurality of feature vectors. In this case, the provision destination relationship table may be provided for each user, or may indicate the relationship between the user identification information, a plurality of information provision destinations, and a plurality of feature vectors. In this modification, the extraction unit 15 converts the extracted word data into a feature vector. Various known methods can be used for this conversion method. The extraction unit 15 may hold a list of feature vectors calculated in advance for a large number of words using a known method. The feature vector may be generated by a device other than the analysis device 10. In this case, the extraction unit 15 converts each extracted word data into a feature vector by using the retained feature vector list. Hereinafter, the converted feature vector is referred to as a word feature vector.

  The determining unit 17 calculates the similarity between each feature vector stored in the user providing destination relationship table and each word feature vector calculated by the extracting unit 15, thereby obtaining an information providing destination (to be searched) ( The combination of the site 7) and word data is specified. Based on a combination of a word feature vector and a feature vector whose calculated similarity is equal to or greater than a predetermined threshold, the determination unit 17 combines the word data corresponding to the word feature vector and the site 7 corresponding to the feature vector. Is identified. The predetermined threshold only needs to be held in advance by the determination unit 17, and the specific value is not limited.

  The search control unit 18 requests a search corresponding to the combination specified by the determination unit 17. Specifically, the search control unit 18 requests a search using word data in the combination for information provided from the site 7 in the combination specified by the determination unit 17 as a search keyword. When a plurality of combinations specified by the determination unit 17 include different word data for the same site 7, the search control unit 18 requests a plurality of searches for information provided from the site 7. .

  The update unit 14 further updates the feature vector stored in the provision destination relationship table of the keyword DB 16 in addition to the weight value of each word data. Specifically, based on the evaluation information acquired by the evaluation acquisition unit 21, the update unit 14 updates the feature vector associated with the information providing destination that is the search target in the providing destination relation table. . The update unit 14 updates the feature vector associated with the information providing destination of the search that is the target of the evaluation information so as to approach or separate from the feature vector of the word data used in the search. . The feature vector update method is not limited.

  According to this modification, the relationship between the information providing destination and the feature vector is automatically learned so as to suit each user's preference. As a result, the setting of the provision destination relation table can be automated, and the management effort of the table can be reduced.

[Third embodiment]
Hereinafter, an information analysis apparatus and an information analysis method according to the third embodiment will be described. The third embodiment may be a program that causes at least one computer to execute the information analysis method, or may be a recording medium that can be read by the at least one computer that records such a program. Good.

  FIG. 15 is a diagram conceptually illustrating a processing configuration example of the information analysis apparatus 100 in the third embodiment. As illustrated in FIG. 15, the information analysis apparatus 100 includes an acquisition unit 101, a storage processing unit 102, and an update unit 103. The information analysis apparatus 100 shown in FIG. 15 has, for example, the same hardware configuration as that of the above-described analysis apparatus 10 shown in FIG. Each software element shown in FIG.

  The acquisition unit 101 acquires word data generated due to the user's action. The user's operation includes all operations such as operation of the client 8 (keyboard operation, mouse operation, touch panel operation, moving the terminal of the client 8, etc.), speech, gesture and the like. The acquired word data may be word data obtained indirectly due to the user's action. For example, the acquisition unit 101 can also acquire the plurality of word data from character string data included in information provided from a server device (not shown) by a user browsing operation. The acquisition unit 101 may acquire the plurality of word data from the client 8, or apply predetermined processing (morphological analysis, speech recognition, gesture recognition, etc.) to data obtained from the client 8 or another device Thus, the plurality of word data may be acquired. The acquisition unit 101 can also acquire word data from a portable recording medium or a device other than the client 8.

  The storage processing unit 102 stores the word data acquired by the acquisition unit 101 in the user word storage unit in association with the generation date / time and the scheduled date / time related to the word data. The meanings of the generation date and time and the scheduled date and time are as described in the first embodiment. A specific example of the storage processing unit 102 is the above-described storage processing unit 12.

  The user word storage unit stores word data, generation date / time, and scheduled date / time in association with each other. However, the user word storage unit may store other information. In addition, the user word storage unit may store redundant word data indicating the same word, or may store only one word data indicating the same word, as in the above-described embodiments. In the latter case, the user word storage unit may store the latest generation date / time and scheduled date / time in association with the word data.

  As illustrated in FIG. 15, the information analysis apparatus 100 may or may not have a user word storage unit. The information analysis apparatus 100 can refer to the user word storage unit of another apparatus through communication via the communication unit 4.

  The update unit 103 updates the weight value of the word data based on the relationship between the generation date / time and the scheduled date / time and the current date / time. A specific example of the updating unit 103 is the updating unit 14 described above. In the present embodiment, the weight value updating method (if updated so as to decrease according to each elapsed time from the generation date / time and the scheduled date / time and to increase according to a decrease in the remaining time until the scheduled date / time ( The calculation method is not limited. For example, when the user word storage unit does not store redundant word data indicating the same word, the update unit 103 determines the latest generation date / time and scheduled date / time without considering the number of word data generations (number of acquisitions). The weight of word data may be calculated from the relationship with the current date and time.

  As illustrated in FIG. 15, the information analysis apparatus 100 may not include the extraction unit 15, the keyword DB 16, the determination unit 17, the search control unit 18, and the output processing unit 19. At least one other computer (device) has an extraction unit 15, a keyword DB 16, a determination unit 17, a search control unit 18, and an output processing unit 19, and these software elements are updated by the information analysis device 100. Please refer to.

  The information analysis method in the third embodiment is executed by at least one computer such as the information analysis apparatus 100. The information analysis method in this embodiment includes a first step, a second step, and a third step. Since these processes are the same as the processing contents of the software elements included in the information analysis apparatus 100, details of the processes are omitted as appropriate.

In the first step, the computer (CPU 1) acquires word data generated due to the user's action.
In the second step, the computer (CPU 1) stores the word data acquired in the first step in the user word storage unit in association with the generation date / time and the scheduled date / time related to the word data.
In the third step, the computer (CPU 1) updates the weight value of the word data based on the relationship between the generation date / time and the scheduled date / time and the current date / time.

[Operation and effect of the third embodiment]
In the third embodiment, the word data generated due to the user's action is stored in association with the generation date / time and the scheduled date / time regarding the word data, and based on the relationship between the generation date / time and the scheduled date / time and the current date / time. The weight value of the word data is updated. In this way, by taking into account the relationship between the generation date and time, the scheduled date and time, and the current date and time, the human data that interest and interest fades with the weight value of the word data and interest as the date and time approaches. And can reflect the human characteristics of increasing interest. Therefore, according to the third embodiment, the user's interest in each word and the degree of interest can be managed with high accuracy.

  In addition, in the some flowchart used by the above-mentioned description, although several process (process) is described in order, the execution order of the process performed by each embodiment is not restrict | limited to the order of the description. In each embodiment, the order of the illustrated steps can be changed within a range that does not hinder the contents. Moreover, each above-mentioned embodiment can be combined in the range in which the content does not conflict.

  The above contents can also be specified as follows. However, the above-mentioned content is not limited to the following description.

1. Acquisition means for acquiring word data generated due to the user's action;
Storage processing means for storing the word data in a user word storage unit in association with a generation date and a scheduled date and time related to the word data;
Updating means for updating the weight value of the word data based on the relationship between the generation date and time and the scheduled date and time and the current date and time;
An information analysis apparatus comprising:
2. The update means uses a time change model that decreases according to each elapsed time from the generation date and time and the scheduled date and time and increases according to a decrease in remaining time until the scheduled date and time. Calculate weight values,
1. Information analysis device described in 1.
3. The storage processing means stores certain word data in the user word storage unit in association with a plurality of different generation dates and times,
The update means calculates the weight value for each generation date and time by applying the generation date and time or a pair of the generation date and the scheduled date and time to the time change model, and adds the calculated weight values together. Calculating a weight value of the word data;
2. Information analysis device described in 1.
4). The updating means calculates a weight value of the word data so as to increase in accordance with an increase in the number of generations of word data;
1. To 3. An information analysis apparatus according to any one of the above.
5). Extraction means for extracting word data from a plurality of word data based on a weight value for each word data;
A determination means for determining an information providing destination and a search keyword based on the extracted word data;
Search control means for requesting a search using the determined search keyword for information provided from the determined information providing destination;
Evaluation acquisition means for acquiring evaluation information of the user for the information extracted by the search;
Further comprising
The update means updates the weight value of the extracted word data based on the user evaluation information.
1. To 4. An information analysis apparatus according to any one of the above.
6). The acquisition means extracts the word data by decomposing part-of-speech of character string data acquired due to the action of the user.
1. To 5. An information analysis apparatus according to any one of the above.
7). In an information analysis method executed by at least one computer,
Get word data generated due to user actions,
The word data is stored in the user word storage unit in association with the generation date and scheduled date and time related to the word data,
Updating the weight value of the word data based on the relationship between the generation date and time and the scheduled date and time and the current date and time;
Information analysis method.
8). The update is performed using a time change model that decreases according to each elapsed time from the generation date and time and the scheduled date and time and increases according to a decrease in remaining time until the scheduled date and time. Including calculating the value,
7). Information analysis method described in 1.
9. Certain word data is associated with a plurality of different generation dates and stored in the user word storage unit,
By applying the generation date / time or the generation date / time and scheduled date / time pair to the time change model, respectively, the weight value is calculated for each generation date / time,
The weight value of the word data is calculated by adding the calculated weight values.
Further includes: Information analysis method described in 1.
10. Calculating the weight value of the word data so as to increase as the number of generations of word data increases,
Further includes: To 9. The information analysis method as described in any one of these.
11. Extract word data from multiple word data based on the weight value for each word data,
Based on the extracted word data, an information providing destination and a search keyword are determined,
Requesting a search using the determined search keyword for information provided from the determined information providing destination;
Obtaining the user's evaluation information for the information extracted by the search;
Updating the weight value of the extracted word data based on the evaluation information of the user;
Further includes: To 10. The information analysis method as described in any one of these.
12 The obtaining includes decomposing part of speech of character string data obtained due to the user's action,
7). To 11. The information analysis method as described in any one of these.
13. 7). To 12. A program that causes at least one computer to execute the information analysis method according to any one of the above.

1 CPU
2 Memory 7 Information provision site (site)
8 Client device (client)
10,100 Information analyzer (analyzer)
11, 101 Acquisition unit 12, 102 Storage processing unit 13 User word DB
14,103 Update unit 15 Extraction unit 16 Keyword DB
17 Determination unit 18 Search control unit 19 Output processing unit 21 Evaluation acquisition unit

Claims (8)

Acquisition means for acquiring word data generated due to the user's action;
Storage processing means for storing the word data in a user word storage unit in association with a generation date and a scheduled date and time related to the word data;
Updating means for updating the weight value of the word data based on the relationship between the generation date and time and the scheduled date and time and the current date and time;
An information analysis apparatus comprising: The update means uses a time change model that decreases according to each elapsed time from the generation date and time and the scheduled date and time and increases according to a decrease in remaining time until the scheduled date and time. Calculate weight values,
The information analysis apparatus according to claim 1. The storage processing means stores certain word data in the user word storage unit in association with a plurality of different generation dates and times,
The update means calculates the weight value for each generation date and time by applying the generation date and time or the generation date and time / schedule date and time pair to the time change model, and adds the calculated weight values. Calculating a weight value of the word data;
The information analysis apparatus according to claim 2. The updating means calculates a weight value of the word data so as to increase in accordance with an increase in the number of generations of word data;
The information analysis apparatus according to any one of claims 1 to 3. Extraction means for extracting word data from a plurality of word data based on a weight value for each word data;
A determination means for determining an information providing destination and a search keyword based on the extracted word data;
Search control means for requesting a search using the determined search keyword for information provided from the determined information providing destination;
Evaluation acquisition means for acquiring evaluation information of the user for the information extracted by the search;
Further comprising
The update means updates the weight value of the extracted word data based on the user evaluation information.
The information analysis device according to any one of claims 1 to 4. The acquisition means extracts the word data by decomposing part-of-speech of character string data acquired due to the action of the user.
The information analysis device according to any one of claims 1 to 5. In an information analysis method executed by at least one computer,
Get word data generated due to user actions,
The word data is stored in the user word storage unit in association with the generation date and scheduled date and time related to the word data,
Updating the weight value of the word data based on the relationship between the generation date and time and the scheduled date and time and the current date and time;
Information analysis method.   A program for causing at least one computer to execute the information analysis method according to claim 7.

Images