JP5483442B2 - Information processing apparatus, dictionary updating method, and program - Google Patents

Description

  The present invention relates to an information processing apparatus, a dictionary updating method, and a program, and more particularly, to an information processing apparatus having a dictionary used for genre assignment, a dictionary updating method, and a program.

  In Patent Literature 1, the degree of association between a word extracted from classification target information, which is electronic information, and a genre word stored in a database (dictionary) is calculated. A genre classification device for classification is described.

JP 2008-65523 A

  The genre classification apparatus described in Patent Literature 1 classifies classification target information using genre words stored in advance in a database (dictionary). For this reason, the genre classification device described in Patent Literature 1 cannot properly classify electronic information unless genre words are appropriately set in the dictionary.

  On the other hand, the genre to which the word used in the electronic information belongs may change depending on the time. For example, it is appropriate that the word “South Africa” belongs to the “sports” or “soccer” genre before and after the 2010 World Cup, but otherwise belongs to the “sports” or “soccer” genre. That is no longer appropriate.

  Since the genre classification device described in Patent Document 1 continuously uses a genre word that has been set once, there is a problem that a genre cannot be appropriately assigned to a word whose genre changes depending on the time. It was.

  The objective of this invention is providing the information processing apparatus, the dictionary update method, and program which can solve the subject mentioned above.

  The information processing apparatus according to the present invention includes a storage unit that stores a dictionary in which a genre and a phrase related to the genre are associated with each other for each genre, and is registered on the network during the predetermined time every time the predetermined time elapses. And storing means for storing electronic information to which any one of the genres is assigned, and as each time the electronic information is stored, the phrase shown in the electronic information is associated with the assigned genre given to the electronic information. The number a of first electronic information which is electronic information stored in the storage means between the time when the electronic information is stored and the time before the time is stored, Of the first electronic information, the number b of the second electronic information to which the assigned genre is assigned, the number c of the third electronic information including the phrase in the first electronic information, and the third electronic information Control means for registering the word / phrase in the dictionary in association with the given genre when it is decided to register the word / phrase, based on the number d of the fourth electronic information to which the given genre is given; including.

  The dictionary update method of the present invention is a dictionary update method performed by the information processing apparatus, and stores a dictionary in which a genre and a word related to the genre are associated with each other in a storage unit, and a predetermined time. A storage step of storing in the storage means electronic information that is registered on the network and given any of the genres for each predetermined time, and each time the electronic information is stored, the electronic information is stored. The storage means between the time when the electronic information is stored and the time prior to the time when the electronic information is stored in the dictionary in association with the assigned genre given to the electronic information. The number a of first electronic information stored in the electronic information, the number b of second electronic information to which the given genre is assigned among the first electronic information, and the first power information. When the number c of third electronic information including the word in the information and the number d of fourth electronic information to which the assigned genre is given out of the third electronic information are determined and the word is registered And a control step of registering the word / phrase in the dictionary in association with the assigned genre when determined.

  The program of the present invention causes a computer to function as each unit included in the information processing apparatus.

  According to the present invention, it is possible to appropriately update a dictionary according to a phrase whose genre to which the time belongs changes.

It is the block diagram which showed one Embodiment of this invention. It is the figure which showed an example of the genre dictionary 24a. It is the figure which showed the word described in the strong search list | wrist 24a1. It is the figure which showed the phrase described in the learning list | wrist 24a2. It is a figure for demonstrating an example of the similarity calculation which the selection part 22 performs. It is a figure for demonstrating operation | movement of the genre provision server 2. FIG. It is a figure for demonstrating the procedure 1 of a genre dictionary learning. It is a figure for demonstrating the procedure 2 of a genre dictionary learning. It is a figure for demonstrating the procedure 3 of a genre dictionary learning.

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

  FIG. 1 is a block diagram showing a genre assignment system including a genre assignment server (information processing apparatus) according to an embodiment of the present invention.

  In FIG. 1, the genre assignment system includes a document collection server 1, a genre assignment server 2, a user PC (personal computer) 3, and a server 4. The document collection server 1, genre assignment server 2, user PC 3, and server 4 can communicate with each other via a network 5. The genre assignment server 2 includes an acquisition unit 21, a selection unit 22, a storage unit 23, a storage unit 24, a control unit 25, and an assignment unit 26. The document collection server 1 may be incorporated in the genre assignment server 2. Further, the number of user PCs 3 and servers 4 is not limited to one and may be a plurality.

  The document collection server 1 can be generally called a storage device.

  The document collection server 1 collects and stores electronic information registered on the network 5. The electronic information registered on the network 5 is electronic information having a document such as a new tweet (new mini blog) or a blog registered on the network 5 from the user PC 3. The document collection server 1 accumulates the collected electronic information in association with the collection date and time of the electronic information.

  Acquisition unit 21 can be generally referred to as acquisition means.

  The acquisition unit 21 acquires electronic information registered on the network 5 during 5 minutes from the document collection server 1 every 5 minutes. Note that 5 minutes can generally be called a predetermined time. The predetermined time is not limited to 5 minutes and can be changed as appropriate.

  Sorting unit 22 can be generally referred to as sorting means.

  The sorting unit 22 gives a genre corresponding to the content indicated by the electronic information to the electronic information acquired by the acquiring unit 21. The selection unit 22 deletes electronic information to which a plurality of genres are assigned from among the electronic information to which genres are assigned, and stores one or more pieces of electronic information to which only one genre is assigned in the storage unit 23. To do.

  In the present embodiment, the selection unit 22 uses the genre dictionary 24 a stored in the storage unit 24 to give a genre to the electronic information acquired by the acquisition unit 21. In the genre dictionary 24a, for each predetermined genre, a genre and a phrase related to the genre are associated with each other.

  FIG. 2 is a diagram showing an example of the genre dictionary 24a. The genre dictionary 24a can be generally called a dictionary.

In FIG. 2, the genre dictionary 24a has six genres of “politics”, “train”, “IT”, “food”, “soccer”, and “idol”. The six genres “politics”, “train”, “IT”, “food”, “soccer”, and “idol” can be generally referred to as a plurality of predetermined specific genres. The specific genre is not limited to “politics”, “train”, “IT”, “food”, “soccer”, and “idol”, but can be changed as appropriate.
The genre dictionary 24a has a strong search list 24a1 and a learning list 24a2 for each genre.

  FIG. 3 is a diagram illustrating words described in the strong search list 24a1 of the genre “soccer”.

  The strong search list 24a1 describes words (words not included in topics of other genres) that can be determined as topics words of the genre to which the strong search list 24a1 belongs. The strong search list 24a1 is prepared in advance by an administrator of the genre assignment server 2 or the like.

  FIG. 4 is a diagram illustrating words and phrases described in the learning list 24a2 of the genre “soccer”.

  In the learning list 24a2, words and phrases that are highly likely to be topical phrases of the genre to which the learning list 24a2 belongs, that is, words and phrases related to the genre to which the learning list 24a2 belongs are described. The words and phrases described in the learning list 24a2 are, for example, a word and a URL (Uniform Resource Locator). The URL may be a URL for specifying an image.

  In the present embodiment, the words / phrases described in the learning list 24a2 are determined by the control unit 25. It should be noted that the phrase described in the learning list 24a2 is given a point where the numerical value increases as the possibility that the topic is in the genre to which the learning list 24a2 belongs is higher.

  The sorting unit 22 gives the genre to which the word belongs to the electronic information having the word described in the strong search list 24a1 among the electronic information acquired by the acquiring unit 21.

  The sorting unit 22 calculates the similarity between the electronic information and the learning list 24a2 of each genre for each electronic information acquired by the acquiring unit 21.

  The selection unit 22 gives the electronic information acquired by the acquisition unit 21 the genre to which the learning list 24a2 whose similarity with the electronic information exceeds a certain threshold belongs. The constant threshold is set in the sorting unit 22.

  FIG. 5 is a diagram for explaining an example of similarity calculation performed by the selection unit 22.

  The selection unit 22 extracts words from the electronic information B1. A known technique is used as a word extraction method.

  Hereinafter, it is assumed that the sorting unit 22 has extracted “□□”, “Manufacturer”, “Easy”, “Mobile”, and “Touch” from the electronic information B1. In FIG. 5, the word extracted from the electronic information B1 is represented as B1 = (□□, maker, simple, mobile, touch). The sorting unit 22 adds 1 point to each word extracted from the electronic information B1.

  Further, it is assumed that “□□”, “maker”, “mobile”, and “touch” are described in the learning list R of a certain genre. In FIG. 5, the learning list R is represented as R = (□□, manufacturer, mobile phone, touch). In the learning list R, 10 points are added to “□□”, 2 points are added to “maker”, 1 point is added to “mobile”, and 0.5 point is added to “touch”. Suppose that

The selection unit 22 calculates the similarity sim (R, B1) according to the similarity calculation formula sim (R, B1) = | R · B1 | / | R || B1 |. In the example shown in FIG.
| R · B1 | = 10 + 2 + 1 + 0.5 = 13.5
| R | = sqrt (100 + 4 + 1 + 0.25) = 10.26,
| B1 | = sqrt (1 + 1 + 1 + 1 + 1) = 2.23, so
sim (R, B1) = 13.5 / (10.26 × 2.23) = 0.59.

  In the similarity calculation formula used in the present embodiment, the similarity is higher for electronic information that includes many high-point words and less irrelevant words.

  Note that the selection unit 22 may add a genre to the electronic information using only the strong search list 24a1 without using the learning list 24a2.

  When the genre assigning process using the genre dictionary 24a is completed, the sorting unit 22 adds, to the electronic information that is not given any genre shown in the genre dictionary 24a among the electronic information obtained by the obtaining unit 21. A genre of “others” is given.

  When the genre assignment to each electronic information acquired by the acquisition unit 21 is completed, the selection unit 22 deletes the electronic information to which a plurality of genres are assigned from the electronic information to which the genre is assigned. One or a plurality of pieces of electronic information to which only “A” is assigned are stored in the storage unit 23.

  Hereinafter, in the present embodiment, a description is given of a situation in which a plurality of electronic information registered every five minutes and assigned with any of a plurality of genres is stored in the storage unit 23 every five minutes.

  Storage unit 23 can generally be referred to as storage means.

  The storage unit 23 stores a plurality of pieces of electronic information registered every five minutes and assigned with any of a plurality of genres every five minutes.

  Storage unit 24 can generally be referred to as storage means.

  The storage unit 24 stores a genre dictionary 24a.

  Control unit 25 can be generally referred to as determining means.

  Each time a plurality of pieces of electronic information are stored in the storage unit 23, the control unit 25 sets the target word / phrase for each of the words / phrases (hereinafter referred to as “target words / phrases”) indicated in any of the plurality of pieces of electronic information. It is determined whether or not to register in the genre dictionary 24a (learning list 24a2) in association with the genre (hereinafter referred to as “grant genre”) given to the electronic information indicating the target word / phrase. The target phrase includes a word and a URL.

  First, the control unit 25 stores the electronic information stored in the storage unit 23 (hereinafter referred to as “first electronic information”) between the time when the plurality of pieces of electronic information are stored in the storage unit 23 and the time before the time. Of the first electronic information, the number b of the second electronic information to which the assigned genre is assigned, the number c of the third electronic information including the target phrase in the first electronic information, The number d of the 4th electronic information to which the grant genre was given among the three electronic information was calculated.

  In the present embodiment, each time a plurality of pieces of electronic information are stored in the storage unit 23, the control unit 25 includes the number aA of the plurality of pieces of electronic information and the piece of electronic information to which a given genre is assigned. The number bA of (eighth electronic information), the number cA of electronic information (the ninth electronic information) including the target word / phrase among the plurality of electronic information, and the target word / phrase with the given genre among the plurality of electronic information. The number dA of electronic information (tenth electronic information) to be included is calculated.

  Then, when a plurality of pieces of electronic information are first stored in the storage unit 23, the control unit 25 calculates aA as the number a of the first electronic information and bA as the number b of the second electronic information. Then, cA is calculated as the number c of the third electronic information, and dA is calculated as the number d of the fourth electronic information.

  Thereafter, each time a plurality of pieces of electronic information are stored in the storage unit 23, the control unit 25, as described below, the number a of the first electronic information, the number b of the second electronic information, and the third electronic information The number c and the number d of the fourth electronic information are calculated.

  The control unit 25 multiplies the first multiplication value, which is a value obtained by multiplying the number of first electronic information calculated previously by the first weighting factor, and the second weighting factor, the number aA of the plurality of electronic information calculated this time. The sum of the calculated value and the second multiplied value is calculated as the number of first electronic information a.

  The control unit 25 multiplies the number b of the second electronic information calculated previously by the first weighting factor and the third weighting value multiplied by the second weighting factor and the number bA of the plurality of electronic information calculated this time. The sum of the calculated value and the fourth multiplication value is calculated as the number b of the second electronic information this time.

  The control unit 25 multiplies the fifth weighted value, which is a value obtained by multiplying the number c of the third electronic information calculated last time by the first weighting factor, and the second weighting factor by the number cA of the plurality of electronic information calculated this time. The sum of the calculated value and the sixth multiplication value is calculated as the number c of the third electronic information at this time.

  The control unit 25 multiplies the seventh weight, which is a value obtained by multiplying the number d of the fourth electronic information calculated previously by the first weighting factor, and the second weighting factor, the number dA of the plurality of electronic information calculated this time. The sum of the calculated value and the eighth multiplication value is calculated as the number d of the fourth electronic information this time.

  The first weighting factor is obtained by subtracting the predetermined value α from the value of (n−1) / n, and the second weighting factor is obtained by adding the predetermined value α to the value of 1 / n. Is desirable. In addition, when the accumulated value (accumulated number) of the number of times when a plurality of electronic information is stored is smaller than a predetermined specified value, n is the accumulated number, and when the accumulated number is equal to or greater than the predetermined specified value, n is It becomes a predetermined specified value. The predetermined specified value is stored in the control unit 25, for example.

  For example, the control unit 25 calculates the number a of the first electronic information, the number b of the second electronic information, the number c of the third electronic information, and the number d of the fourth electronic information according to the following calculation formulas. To do.

  Note that t is a value incremented by 1 every 5 minutes, which is a predetermined time, and can be read as a cumulative value of the number of times a plurality of pieces of electronic information are stored in the storage unit 23. The T is 36, which is a value obtained by dividing 3 hours = 180 minutes, which is a calculation target period (past data learning target period), by 5 minutes, which is a predetermined time. T is an example of a predetermined specified value. The calculation target period is not limited to 3 hours and can be changed as appropriate.

  When calculating the number a of the first electronic information, the control unit 25 replaces “P” in the expressions (1) to (2) with “a”, and newly stores Ct when the cumulative number of times is t. By replacing the number of electronic information stored in the unit 23, the number a of the first electronic information is calculated.

  Hereinafter, an example in which the control unit 25 calculates the number a of the first electronic information using the expressions (1) to (2) will be described.

The control unit 25 sets the number a ₁ of the first electronic information as a number of a plurality of electronic information newly stored in the storage unit 23 when a ₁ = C ₁ , that is, the cumulative number t = 1.

When 1 ≦ t ≦ T, the control unit 25 calculates the number of first electronic information a _{t + 1} according to equation (1), and when t> T, the number of first electronic information according to equation (2). a _{t + 1} is calculated.

  When the control unit 25 calculates the number b of the second electronic information, the number c of the third electronic information, and the number d of the fourth electronic information, “P” in the expressions (1) to (2) ”Is replaced with“ b ”,“ c ”, and“ d ”, respectively, and Ct is a given genre among a plurality of electronic information newly stored in the storage unit 23 when the cumulative number of times is t. Of the plurality of electronic information newly stored in the storage unit 23 when the number of added electronic information is the cumulative number t, the number of electronic information including the target phrase is newly stored in the storage unit 23 when the cumulative number is t. Of the plurality of stored electronic information, the number is replaced with the number of electronic information that includes the target word / phrase and is assigned with a given genre.

  Whether or not the control unit 25 associates the target word / phrase with the given genre and registers it in the genre dictionary 24a (learning list 24a2) in the storage unit 24 based on a, b, c, and d having the same suffix for t. To decide.

  For example, if (b / a) <(d / c) holds, the control unit 25 determines to register the target word / phrase in the genre dictionary 24a in association with the assigned genre. On the other hand, if (b / a) <(d / c) does not hold, the control unit 25 determines not to register the target word / phrase in the genre dictionary 24a in association with the assigned genre.

  The control unit 25 satisfies (b / a) <(d / c), and the determination value F = (2 × (d / c) × (d / b)) / ((d / c) + If (d / b))> predetermined threshold value holds, it may be determined that the target word / phrase is associated with the assigned genre and registered in the genre dictionary 24a.

  In this case, when (b / a) <(d / c) does not hold or when the determination value F> the predetermined threshold does not hold, the control unit 25 associates the target phrase with the assigned genre in the genre dictionary 24a. Decide not to register.

  In addition, the control unit 25 satisfies (b / a) <(d / c), satisfies the determination value F> predetermined threshold, and satisfies the determination value F> the determination value Ff related to other genres. In this case, it may be determined that the target word / phrase is registered in the genre dictionary 24a in association with the assigned genre. At this time, the control unit 25 may assign the value of the determination value F × d to the target word / phrase as a point of the target word / phrase. However, Ff = (2 × (df / c) × (df / bf)) / ((df / c) + (df / bf))), and bf is given the other genre in the first electronic information. It is the number of electronic information, and df is the number of electronic information to which other genres are assigned among the third electronic information.

  In this case, the control unit 25 determines that (b / a) <(d / c) does not hold, or if the determination value F> the predetermined threshold value does not hold, or the determination value F> the determination value Ff. If not, it is determined that the target word / phrase is not registered in the genre dictionary 24a in association with the assigned genre.

  In addition, when the target word / phrase and the assigned genre are stored in the genre dictionary 24a and the plurality of pieces of electronic information are stored in the storage unit 23, the control unit 25 further includes the target words / phrases in the storage unit 24. Then, it is determined whether or not to delete the target word / phrase (hereinafter referred to as “past word / phrase”) not shown in any of the plurality of electronic information from the genre dictionary 24a.

  First, the control unit 25 includes the number a of the first electronic information, the number b1 of the fifth electronic information to which the related genre, which is the genre associated with the past phrase in the first electronic information, and the first electronic information. The number c1 of the sixth electronic information including the past phrase and the number d1 of the seventh electronic information to which the related genre is assigned among the sixth electronic information are calculated.

  The control unit 25 determines the number a of the first electronic information, the number b1 of the fifth electronic information, the number c1 of the sixth electronic information, and the number d1 of the seventh electronic information (1) to (2) described above. Calculate according to formula.

  When the control unit 25 calculates the number b1 of the fifth electronic information, the number c1 of the sixth electronic information, and the number d1 of the seventh electronic information, “P” in the expressions (1) to (2) ”Is replaced with“ b1 ”,“ c1 ”, and“ d1 ”, respectively, and the related genre among the plurality of electronic information newly stored in the storage unit 23 when Ct is the cumulative number of times t, respectively. Of the plurality of electronic information newly stored in the storage unit 23 when the number of added electronic information is the cumulative number t, the number of electronic information including past words is newly stored in the storage unit 23 when the cumulative number is t. Of the plurality of stored electronic information, the number is replaced with the number of electronic information including a past word and having a related genre.

  The control unit 25 determines whether or not to delete the past phrase from the genre dictionary 24a based on a, b1, c1 and d1 having the same suffix for t.

  For example, if (b1 / a) <(d1 / c1) does not hold, the control unit 25 determines to delete the past word / phrase from the genre dictionary 24a. On the other hand, when (b1 / a) <(d1 / c1) holds, the control unit 25 determines not to delete the past word / phrase from the genre dictionary 24a.

  The controller 25 determines that (b1 / a) <(d1 / c1) does not hold, or the determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / If c1) + (d1 / b1))> predetermined threshold does not hold, it may be determined that the past phrase is deleted from the genre dictionary 24a.

  In this case, if (b1 / a) <(d1 / c1) holds and the determination value F1> the predetermined threshold holds, the control unit 25 determines not to delete the past phrase from the genre dictionary 24a.

  In addition, the control unit 25 determines that (b1 / a) <(d1 / c1) does not hold, or if the determination value F1> predetermined threshold does not hold, or the determination value F1> the determination value related to other genres. If Ff1 does not hold, it may be determined to delete the past phrase from the genre dictionary 24a. However, Ff1 = (2 × (df1 / c1) × (df1 / bf1)) / ((df1 / c1) + (df1 / bf1)), and bf1 is assigned the other genre in the first electronic information. It is the number of electronic information, and df1 is the number of electronic information to which other genres are assigned among the third electronic information.

  In this case, if the control unit 25 satisfies (b1 / a) <(d1 / c1), the determination value F1> the predetermined threshold, and the determination value F1> the determination value Ff1, the past It is determined not to delete the phrase from the genre dictionary 24a.

  When it is determined that the target word / phrase is stored in the genre dictionary 24a in association with the given genre, the control unit 25 associates the target word / phrase with the given genre and stores it in the learning list 24a2 in the genre dictionary 24a.

  In addition, when the value of the determination value F × d is given as a point to the target word / phrase determined to be stored in the genre dictionary 24a in association with the given genre, the control unit 25 gives the target word / phrase together with the point. The information is stored in the learning list 24a2 in the genre dictionary 24a in association with the genre.

  When the control unit 25 determines to delete the past phrase from the genre dictionary 24a, the control unit 25 deletes the past phrase from the learning list 24a2 in the genre dictionary 24a.

  The granting unit 26 can generally be referred to as granting means.

  In the situation where the genre request has been received from the user PC 3 or the server 4, if the word / phrase is stored in the storage unit 24 (genre dictionary 24 a), the assigning unit 26 is associated with the word / phrase. The assigned genre is transmitted to the transmission source (user PC 3 or server 4) of the phrase.

  The genre assignment server 2 may be realized by a computer. In this case, the computer reads and executes a program recorded on a recording medium such as a CD-ROM (Compact Disk Read Only Memory) readable by the computer, thereby obtaining the acquisition unit 21, the selection unit 22, and the storage unit 23. , Function as a storage unit 24, a control unit 25, and a grant unit 26. The recording medium is not limited to the CD-ROM and can be changed as appropriate.

  Next, the operation will be described.

  FIG. 6 is a diagram for explaining the operation of the genre assignment server 2.

  The acquisition unit 21 acquires electronic information (new arrival tweets) for 5 minutes from the document collection server 1 every 5 minutes (step S601).

  When the acquisition unit 21 acquires the electronic information for 5 minutes, the selection unit 22 uses the genre dictionary 24a in the storage unit 24 to display the electronic information acquired by the acquisition unit 21 in the content indicated by the electronic information. A genre corresponding to the above is assigned (steps S602 and S603).

  In the present embodiment, the selection unit 22 gives the genre to which the word belongs to the electronic information having the word described in the strong search list 24a1 among the electronic information acquired by the acquisition unit 21. In addition, the selection unit 22 gives the electronic information acquired by the acquisition unit 21 the genre to which the learning list 24a2 whose similarity with the electronic information exceeds a certain threshold belongs.

  When the genre assigning process using the genre dictionary 24a is completed, the selecting unit 22 selects electronic information that is not assigned any genre shown in the genre dictionary 24a among the electronic information acquired by the acquiring unit 21. , The genre of “others” is given.

  When the genre assignment to each electronic information acquired by the acquisition unit 21 is completed, the selection unit 22 deletes the electronic information to which a plurality of genres are assigned from the electronic information to which the genre is assigned. One or a plurality of pieces of electronic information to which only is assigned is stored in the storage unit 23 (steps S604 and S605).

  Here, the point that electronic information to which only one genre is assigned is stored in the storage unit 23 will be described.

  In the present embodiment, a process is performed for determining whether or not to write the phrase shown in the electronic information stored in the storage unit 23 in the learning list 24a2 of the genre given to the electronic information.

  For this reason, when electronic information to which a plurality of genres are assigned is stored in the storage unit 23, it is unclear to which genre of the plurality of genres the word / phrase indicated in the electronic information corresponds.

  For example, the storage unit 23 stores electronic information that is given the genres of “soccer” and “IT”, “Mr. XX who likes soccer and IT often goes around holidays, futsal, or visits to computer shops.” If this happens, the following problems will occur. It is assumed that “futsal” and “computer shop” are not registered in the genre dictionary 24a.

  There is a problem that the phrase “futsal” in the electronic information is misrecognized as the topic (genre) of “IT”, or the phrase “computer shop” is misrecognized as the topic (genre) of “soccer”.

  For this reason, in this embodiment, the electronic information to which only one genre is assigned is stored in the storage unit 23, thereby preventing the accuracy of the genre dictionary 24a from deteriorating.

  Each time a plurality of pieces of electronic information are stored in the storage unit 23, the control unit 25 sets a, b, c, d, bf, and df for each of the target words indicated in any of the plurality of pieces of electronic information. Then, a, b1, c1, d1, bf1, and df1 are calculated for each of the past words (step S606).

  Hereinafter, a process of calculating a, b, c, d, bf, and df for each of the target words and phrases, and calculating a, b1, c1, d1, bf1, and df1 for each of the past words, This is referred to as update processing.

  When the statistic update process ends, the control unit 25 updates the words in the learning list 24a2 based on a, b, c, d, bf, df, b1, c1, d1, bf1, and df1. Learning is executed (step S607).

  Next, genre dictionary learning will be described with a specific example.

  In this embodiment, steps 1 to 3 are executed as genre dictionary learning.

  FIG. 7 is a diagram for explaining the procedure 1 of genre dictionary learning.

  In the following, as shown in FIG. 7, the genre of the “soccer” is given to the tweet (miniblog) A, the genre of “IT” is given to the tweet B, and the tweet C "Other" genre is given to Tweet D, "Soccer" and "IT" genres are given to Tweet D, "Soccer" genre is given to Tweet E, and "IT" genre is given to Tweet F To do.

  A plurality of genres are assigned to the tweet D. For this reason, the tweet D becomes noise of genre dictionary learning, and therefore the tweet D is deleted by the selection unit 22.

  For the sake of simplicity, it is assumed that tweets A, B, C, E, and F are selected as the first electronic information by the statistical amount update process. For this reason, the number a of the first electronic information is 5 (tweets A, B, C, E, and F).

  As shown in FIG. 7, the phrase “South Africa” is included in tweets A and E, not included in tweets B, C, and F, and the phrase “America” is included in tweets A, B, and C. And are not included in tweets E and F.

  The control unit 25 determines whether or not the target phrase “South Africa” described in the tweet A to which the “soccer” genre (given genre) is assigned is a candidate to be stored in the learning list 24a2 of the “soccer” genre. Determine as follows.

  In this case, the number “b” of the second electronic information to which “soccer” as the grant genre is given among the first electronic information is 2 (tweets A and E).

  Further, the number c of the third electronic information including “South Africa” which is the target word in the first electronic information is 2 (tweets A and E).

  In addition, the number d of the fourth electronic information to which “soccer” as the grant genre is given among the third electronic information is 2 (tweets A and E).

  The control unit 25 represents the genre share indicating the ratio of the number b of the second electronic information to the number a of the first electronic information (the number b of the electronic information to which the assigned genre “soccer” is assigned among the first electronic information). Calculate the rate b / a. In this case, b / a = 2/5 = 40%.

  Subsequently, the control unit 25 selects the third number d for the number d of fourth electronic information (the number of pieces of electronic information to which “soccer” is given as the grant genre including the target phrase “South Africa” in the first electronic information). A post-classification share rate d / c indicating the ratio of the number of electronic information c (the number of electronic information including “South Africa” as the target word in the first electronic information) is calculated. In this case, d / c = 2/2 = 100%.

  Since b / a (= 40%) <d / c (= 100%) holds, the control unit 25 determines “South Africa” as a candidate to be listed in the soccer genre learning list 24a2. Note that the control unit 25 may determine that “South Africa” is listed in the soccer genre learning list 24a2 when (b / a) <(d / c) holds.

  Subsequently, the control unit 25 associates the target word “USA” described in Tweet A with the genre (given genre) of “Soccer” to which Tweet A belongs, and stores it in the learning list 24a2 of the genre of “Soccer”. Whether or not to be a candidate is determined as follows.

  In this case, the number b of the second electronic information to which “soccer” as the grant genre is given among the first electronic information is 2 (tweets A and E).

  In addition, the number c of the third electronic information including the target word “America” in the first electronic information is 3 (tweets A, B, and C).

  In addition, the number d of the fourth electronic information to which “soccer” as the grant genre is assigned among the third electronic information including “America” is 1 (tweet A).

  The control unit 25 represents the genre share indicating the ratio of the number b of the second electronic information to the number a of the first electronic information (the number b of the electronic information to which the assigned genre “soccer” is assigned among the first electronic information). Calculate the rate b / a. In this case, b / a = 2/5 = 40%.

  Subsequently, the control unit 25 determines the third number d for the fourth electronic information d (the number of electronic information including the target word “America” and the assigned genre “soccer” among the first electronic information). A post-classification share rate d / c indicating the ratio of the number of electronic information c (the number of electronic information including the target word “America” in the first electronic information) is calculated. In this case, d / c = 1/3 = 33%.

  Since b / a (= 40%) <d / c (= 33%) does not hold, the control unit 25 determines that “USA” is not included in the soccer genre learning list 24a2.

  FIG. 8 is a diagram for explaining procedure 2 of genre dictionary learning.

  In step 2, the control unit 25 calculates an F value (determination value) from the tweet (electronic information) including the words (words and URLs) listed as candidates to be listed in the list of each genre in step 1. .

  Note that F value = (2 × compliance rate × reproducibility) / (adaptation rate + reproducibility).

  However, the relevance rate = [d (phrase × genre)] / [c (phrase)], and the reproduction rate = [d (phrase × genre)] / [b (genre)].

  The control unit 25 leaves a word / phrase having an F value larger than a predetermined threshold (predetermined threshold) as a candidate to be stored in the genre learning list attached to the tweet (electronic information) including the word / phrase. Note that the control unit 25 may determine to store a phrase whose F value is larger than a predetermined threshold (predetermined threshold) in a genre learning list attached to a tweet (electronic information) including the phrase. . The threshold is 0.4, for example. The threshold value is not limited to 0.4 and can be changed as appropriate.

  For example, for “South Africa” determined as a candidate to be included in the soccer genre learning list 24a2 in step 1, the control unit 25 determines the F value as follows.

  In “South Africa” determined as a candidate to be included in the soccer genre learning list 24 a 2, relevance rate = d (the first electronic information includes the target word “South Africa” and the given genre “soccer” is assigned. Number of electronic information) / c (number of electronic information including “South Africa” as the target word in the first electronic information) = 2/2 = 1.

  In addition, in “South Africa” determined as a candidate to be included in the soccer genre learning list 24a2, recall rate = d (the first electronic information includes the target phrase “South Africa” and the given genre “soccer” Number of electronic information provided) / b (number of electronic information provided with “soccer” as an assigned genre in the first electronic information) = 2/2 = 1.

  Therefore, the F value of “South Africa” determined as a candidate to be included in the soccer genre learning list 24a2 is F value = (2 × 1 × 1) / (1 + 1) = 2/2 = 1.

  Therefore, the F value = 1 of “South Africa” determined as a candidate to be included in the soccer genre learning list 24a2 is larger than the threshold = 0.7. Therefore, the control unit 25 recognizes the soccer genre learning list 24a2. “South Africa” determined as a candidate to be described in (1) is left as a candidate to be described in the soccer genre learning list 24a2.

  Further, as shown in FIG. 8, the F value of “Spain” determined as a candidate to be included in the soccer genre learning list 24a2 is F value = 0.8, which is described in the soccer genre learning list 24a2. The F value of “England” determined as a candidate to be played is F value = 0.67, which is left as a candidate to be described in the soccer genre learning list 24a2.

  FIG. 9 is a diagram for explaining the procedure 3 of genre dictionary learning.

  In step 3, the control unit 25 verifies whether the words (words, URLs) remaining as candidates to be written in the learning list of each genre in step 2 are appropriate.

  In the present embodiment, the control unit 25 uses the F value of a word / word remaining as a candidate in the learning list of a genre in which the word / word remaining as a candidate is “other”. When it is larger than the F value (Ff), it is determined in step 2 that the words (words, URLs) remaining as candidates to be described in the learning list of each genre are appropriate.

  In FIG. 9, it is assumed that in addition to the tweets A, B, C, E, and F, the first electronic information 91 includes a genre of “others” and tweets G and H including the phrase “Germany”.

  In this case, the F value of “Germany” determined as a candidate to be listed in the soccer genre learning list 24a2 is F = 2/4 = 0.5 and recall = 2/2 = 1. Value = (2 × 0.5 × 1) / (0.5 + 1) = 0.66.

  On the other hand, the F value (Ff) when “Germany” is included in the learning list of the genre of “others” is the precision rate = 2/4 = 0.5 and the recall rate = 2/3 = 0.66. Therefore, F value (Ff) = (2 × 0.5 × 0.66) / (0.5 + 0.66) = 0.57.

  Therefore, for the phrase “Germany”, the F value of the “soccer” genre is higher than the F value (Ff) of the “other” genre, so the control unit 25 learns “Germany” as a soccer genre. It is determined to be listed in list 24a2.

  At this time, the control unit 25 adds the number d of the fourth electronic information d (“soccer”, which is the genre of the first electronic information, to the F value (F value = 0.66) related to soccer, and is called “Germany” A value (1.32) obtained by multiplying the number of electronic information including words) is set as the point value of “Germany” in the soccer genre learning list 24a2. In this case, “Germany” with 1.32 points added is registered in the soccer genre learning list 24a2.

  In FIG. 9, as the first electronic information 92, in addition to the tweets A, B, C, E, F, G, and H, the genre “soccer” is given, and the tweets I and I include the phrase “Italy”. Suppose that J exists.

  In this case, the F value of “Italy” determined as a candidate to be included in the soccer genre learning list 24a2 is the precision rate = 2/4 = 0.5 and the recall rate = 2/4 = 0.5. F value = (2 × 0.5 × 0.5) / (0.5 + 0.5) = 0.5.

  On the other hand, the F value (Ff) in the case where “Italy” is listed in the learning list of the genre “others” is the precision rate = 2/4 = 0.5 and the recall rate = 2/3 = 0.66. Therefore, F value (Ff) = (2 × 0.5 × 0.66) / (0.5 + 0.66) = 0.57.

  Therefore, regarding the phrase “Italy”, the F value of the “soccer” genre is equal to or less than the F value (Ff) of the “other” genre, so the control unit 25 designates “Italy” as a learning list 24a2 of soccer genres. It is determined that it is not described in.

  Note that the process in which the control unit 25 determines whether to delete the past phrase from the learning list 24a2 is based on the fact that in Steps 1 to 3, it is determined that the phrase is replaced with the past phrase and the phrase is not described in the learning list 24a2. It is explained that the phrase is determined to be deleted from the learning list 24a2, that the F value is replaced with the F1 value, and the Ff value is replaced with the Ff1 value.

  When it is determined that the target word / phrase is associated with the assigned genre and registered in the genre dictionary 24a, the control unit 25 associates the target word / phrase with the given genre and registers it in the learning list 24a2 in the storage unit 24. Further, when the control unit 25 determines to delete the past phrase from the genre dictionary 24 a, the control unit 25 deletes the past phrase from the learning list 24 a 2 in the storage unit 24.

  In the situation where the genre request has been received from the user PC 3 or the server 4, if the word / phrase is stored in the storage unit 24 (genre dictionary 24 a), the assigning unit 26 is associated with the word / phrase. The assigned genre is transmitted to the transmission source (user PC 3 or server 4) of the phrase.

  According to the present embodiment, the control unit 25 calculates a value calculated for the electronic information stored in the storage unit 23 from the time when a plurality of pieces of electronic information are stored to the time before the time. Based on (a, b, c and d), it is determined whether or not the target word / phrase is associated with the given genre and registered in the genre dictionary 24a.

  For this reason, the target word / phrase stored in the genre dictionary 24a is updated based on the newly registered electronic information. Therefore, the target word / phrase in the genre dictionary 24a can be updated to an appropriate state.

  The target words / phrases stored in the genre dictionary 24a are used when a genre is assigned to a word / phrase requested to be given a genre.

  For this reason, it becomes possible to update the target word / phrase used when assigning a genre to a word / phrase for which genre assignment is requested based on newly registered electronic information.

  Therefore, it is possible to appropriately assign a genre to a phrase whose genre changes according to time.

  Further, when a URL that specifies an image is used as the target phrase, a genre can be appropriately assigned to the image specified by the URL.

  In the present embodiment, when (b / a) <(d / c) holds, the control unit 25 determines to register the target word / phrase in the genre dictionary 24a in association with the assigned genre. In this case, it is possible to register words / phrases that are highly likely to be topics of the assigned genre in the genre dictionary 24a.

  In the present embodiment, the control unit 25 associates the target phrase with the assigned genre and holds the genre dictionary when (b / a) <(d / c) holds and the determination value F> the predetermined threshold holds. It is decided to register in 24a. In this case, it is possible to register words / phrases that are more likely to be topics of the assigned genre in the genre dictionary 24a.

  Further, in the present embodiment, the control unit 25 satisfies (b / a) <(d / c), the determination value F => the predetermined threshold, and the determination value F> other genres. When the determination value Fα holds, it is determined that the target word / phrase is associated with the assigned genre and registered in the genre dictionary 24a. In this case, it is possible to register words / phrases that are more likely to be topics of the assigned genre in the genre dictionary 24a.

  In the present embodiment, the control unit 25 is a value calculated for electronic information stored in the storage unit 23 between the time when a plurality of pieces of electronic information are stored and the time before the time. Based on (a, b1, c1, and d1), it is determined whether or not to delete the past phrase from the genre dictionary 24a.

  For this reason, the past phrase memorize | stored in the genre dictionary 24a is updated based on the newly registered electronic information. Therefore, it is possible to update the word / phrase in the genre dictionary 24a to an appropriate state.

  The past words / phrases stored in the genre dictionary 24a are used when a genre is assigned to a word / phrase requested to be given a genre.

  For this reason, it becomes possible to update the past phrase used when assigning a genre to a phrase for which genre assignment is requested, based on newly registered electronic information.

  Therefore, it is possible to appropriately assign a genre to a phrase whose genre changes according to time.

  In the present embodiment, the control unit 25 determines to delete the past phrase from the genre dictionary 24a when (b1 / a) <(d1 / c1) does not hold. In this case, it is possible to leave past words / phrases that are highly likely to be topics of the assigned genre in the genre dictionary 24a.

  Further, in the present embodiment, the control unit 25, when (b1 / a) <(d1 / c1) does not hold or when the determination value F1> predetermined threshold does not hold, the past phrase is retrieved from the genre dictionary 24a. Decide to delete. In this case, it is possible to leave a past phrase that is more likely to be a topic of the assigned genre in the genre dictionary 24a.

  In the present embodiment, the control unit 25 determines that (b1 / a) <(d1 / c1) does not hold, or if the determination value F1 => predetermined threshold does not hold, or the determination value F1>. When the judgment value Fα1 regarding other genres does not hold, it is determined that the past phrase is deleted from the genre dictionary 24a. In this case, it is possible to leave a past word / phrase that is more likely to be a topic of the assigned genre in the genre dictionary 24a.

  In the present embodiment, the control unit 25 calculates aA, bA, cA, and dA each time a plurality of pieces of electronic information are stored.

  Then, when a plurality of pieces of electronic information are stored for the first time, the control unit 25 calculates aA as the number a of the first electronic information, bA as the number b of the second electronic information, and cA It is calculated as the number c of the third electronic information, and dA is calculated as the number d of the fourth electronic information.

  In addition, each time a plurality of pieces of electronic information are stored, the control unit 25 performs the following steps: a number of first electronic information a, number b of second electronic information, and number c of third electronic information, as described below. And the number d of the fourth electronic information are calculated.

  The control unit 25 multiplies the first multiplication value, which is a value obtained by multiplying the number of first electronic information calculated previously by the first weighting factor, and the second weighting factor, the number aA of the plurality of electronic information calculated this time. The sum of the calculated value and the second multiplied value is calculated as the number of first electronic information a.

  The control unit 25 multiplies the number b of the second electronic information calculated previously by the first weighting factor and the third weighting value multiplied by the second weighting factor and the number bA of the plurality of electronic information calculated this time. The sum of the calculated value and the fourth multiplication value is calculated as the number b of the second electronic information this time.

  The control unit 25 multiplies the fifth weighted value, which is a value obtained by multiplying the number c of the third electronic information calculated last time by the first weighting factor, and the second weighting factor by the number cA of the plurality of electronic information calculated this time. The sum of the calculated value and the sixth multiplication value is calculated as the number c of the third electronic information at this time.

  The control unit 25 multiplies the seventh weight, which is a value obtained by multiplying the number d of the fourth electronic information calculated previously by the first weighting factor, and the second weighting factor, the number dA of the plurality of electronic information calculated this time. The sum of the calculated value and the eighth multiplication value is calculated as the number d of the fourth electronic information this time.

  For this reason, the control unit 25 calculates the number a of the first electronic information, the number b of the second electronic information, the number c of the third electronic information, and the fourth electron from the previous calculation result and the current calculation result. The number d of information is calculated.

  Therefore, the past data used to calculate the number a of the first electronic information, the number b of the second electronic information, the number c of the third electronic information, and the number d of the fourth electronic information are calculated. The amount of processing required to calculate the number a of the first electronic information, the number b of the second electronic information, the number c of the third electronic information, and the number d of the fourth electronic information can be reduced. It becomes possible.

  In this case, the storage unit 23 only needs to store the latest plurality of electronic information. For example, the selection unit 22 immediately before storing the latest plurality of electronic information in the storage unit 23 The electronic information stored in may be deleted.

  Therefore, the cost required for storage for storing data can be reduced.

  In this embodiment, the first weighting factor is obtained by subtracting the predetermined value α from the value of (n−1) / n, and the second weighting factor is obtained by adding the predetermined value α to the value of 1 / n. It is a thing. For this reason, as the predetermined value α is larger, new information can be emphasized.

  Moreover, in this embodiment, the acquisition part 21 acquires the electronic information registered during the 5 minutes from the document collection server 1 every 5 minutes (predetermined time). The sorting unit 22 assigns a genre corresponding to the content indicated by the electronic information to the electronic information acquired by the acquiring unit 21, and the electronic information to which a plurality of genres are assigned among the electronic information to which the genre is assigned. And one or more pieces of electronic information to which only one genre is assigned are stored in the storage unit 23.

  In the present embodiment, the words and phrases shown in the electronic information stored in the storage unit 23 are described in the learning list 24a2 of the genre given to the electronic information. For this reason, when electronic information to which a plurality of genres are assigned is stored in the storage unit 23, it is unclear to which genre of the plurality of genres the word / phrase indicated in the electronic information corresponds.

  In the present embodiment, it is possible to prevent deterioration of the accuracy of the genre dictionary 24a by storing electronic information to which only one genre is assigned in the storage unit 23.

  In the present embodiment, the calculation method of a, b, c, d, bf, df, b1, c1, d1, bf1, and df1 is not limited to the method using equations (1) to (2). For example, electronic information for the past several times is stored in the storage unit 23, and a, b, c, d, bf, df, b1, c1, d1, and bf1 are stored based on the stored electronic information. And df1 may be calculated.

  Further, in the present embodiment, the acquisition unit 21 and the selection unit 22 cooperate with each other, and every five minutes, a plurality of electronic devices that are registered on the network 5 and given any one of a plurality of genres during the five minutes. The information is stored in the storage unit 23. For example, a user or another server is registered on the network 5 every 5 minutes and is assigned with any of a plurality of genres. Information may be stored in the storage unit 23.

  In the embodiment described above, the illustrated configuration is merely an example, and the present invention is not limited to the configuration.

DESCRIPTION OF SYMBOLS 1 Document collection server 2 Genre assignment server 21 Acquisition part 22 Sorting part 23 Storage part 24 Storage part 25 Control part 26 Giving part 3 User PC
4 servers 5 networks

Claims (24)

Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. and, when it is determined to register the phrase, looking contains a control means for registering in the dictionary in association with the phrase to the grant genre,
The information processing apparatus according to claim 1, wherein the control unit determines that the word / phrase is registered in the dictionary in association with the assigned genre when (b / a) <(d / c) holds . Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. And, when it is determined to register the word / phrase, control means for registering the word / phrase in the dictionary in association with the given genre,
The control means satisfies (b / a) <(d / c), and determination value F = (2 × (d / c) × (d / b)) / ((d / c) + (d / B))> An information processing apparatus that determines that the word / phrase is registered in the dictionary in association with the given genre when the predetermined threshold is satisfied. Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. And, when it is determined to register the word / phrase, control means for registering the word / phrase in the dictionary in association with the given genre,
In the dictionary, a plurality of predetermined genres and other genres that are not any of the plurality of genres are stored as the genres,
The control means includes
(B / a) <(d / c) holds, and
Determination value F = (2 × (d / c) × (d / b)) / ((d / c) + (d / b))> predetermined threshold value, and
The determination value F> the determination value Ff relating to other genres (where Ff = (2 × (df / c) × (df / bf)) / ((df / c) + (df / bf))), bf Is the number of electronic information to which the other genre is assigned in the first electronic information, and df is the number of electronic information to which the other genre is assigned in the third electronic information. The information processing apparatus determines to register the word / phrase in the dictionary in association with the given genre. The information processing apparatus according to any one of claims 1 to 3 ,
When the electronic information is stored, the control means further determines whether or not to delete from the dictionary a past phrase that is not indicated by the electronic information among words in the dictionary. A number a1, a number b1 of fifth electronic information to which a related genre that is a genre associated with the past phrase is included in the first electronic information, and a sixth including the past phrase in the first electronic information. When it is determined based on the number c1 of electronic information and the number d1 of seventh electronic information to which the related genre is assigned among the sixth electronic information, and when it is determined to delete the past phrase from the dictionary, Deleting the past phrase from the dictionary;
The information processing apparatus, wherein the control unit determines to delete the past phrase from the dictionary when (b1 / a) <(d1 / c1) does not hold. The information processing apparatus according to any one of claims 1 to 3 ,
When the electronic information is stored, the control means further determines whether or not to delete from the dictionary a past phrase that is not indicated by the electronic information among words in the dictionary. A number a1, a number b1 of fifth electronic information to which a related genre that is a genre associated with the past phrase is included in the first electronic information, and a sixth including the past phrase in the first electronic information. When it is determined based on the number c1 of electronic information and the number d1 of seventh electronic information to which the related genre is assigned among the sixth electronic information, and when it is determined to delete the past phrase from the dictionary, Deleting the past phrase from the dictionary;
In the case where (b1 / a) <(d1 / c1) does not hold, or the determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / c1) + If (d1 / b1))> predetermined threshold does not hold, the information processing apparatus determines to delete the past word / phrase from the dictionary. The information processing apparatus according to any one of claims 1 to 3 ,
When the electronic information is stored, the control means further determines whether or not to delete from the dictionary a past phrase that is not indicated by the electronic information among words in the dictionary. A number a1, a number b1 of fifth electronic information to which a related genre that is a genre associated with the past phrase is included in the first electronic information, and a sixth including the past phrase in the first electronic information. When it is determined based on the number c1 of electronic information and the number d1 of seventh electronic information to which the related genre is assigned among the sixth electronic information, and when it is determined to delete the past phrase from the dictionary, Deleting the past phrase from the dictionary;
In the dictionary, a plurality of predetermined genres and other genres that are not any of the plurality of genres are stored as the genres,
The control means includes
When (b1 / a) <(d1 / c1) does not hold, or the determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / c1) + (d1 / b1) )> Predetermined threshold value does not hold, or the determination value F1> the determination value Ff1 relating to other genres (where Ff1 = (2 × (df1 / c1) × (df1 / bf1)) / ((df1 / c1) ) + (Df1 / bf1)), bf1 is the number of electronic information to which the other genre is assigned in the first electronic information, and df1 is the number of electronic information to which the other genre is assigned in the third electronic information The number of electronic information) is not established, the information processing apparatus determines to delete the past phrase from the dictionary.   Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
  Control means for registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information each time the electronic information is stored,
  When the electronic information is stored, the control means further includes the number of first electronic information a stored in the storage means between the time when the electronic information is stored and the time before the time a. And among the first electronic information, a number b1 of fifth electronic information to which a related genre that is a genre related to a past phrase that is a word in the dictionary and is not indicated in the electronic information is given, Based on the number c1 of the sixth electronic information including the past phrase in the first electronic information and the number d1 of the seventh electronic information to which the related genre is assigned among the sixth electronic information, (b1 If / a) <(d1 / c1) does not hold, it is determined to delete the past word / phrase from the dictionary, and when it is determined to delete the past word / phrase from the dictionary, the past word / phrase is deleted from the dictionary. , Information processing equipment .   Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
  Control means for registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information each time the electronic information is stored,
  When the electronic information is stored, the control means further includes the number of first electronic information a stored in the storage means between the time when the electronic information is stored and the time before the time a. And among the first electronic information, a number b1 of fifth electronic information to which a related genre that is a genre related to a past phrase that is a word in the dictionary and is not indicated in the electronic information is given, Based on the number c1 of the sixth electronic information including the past phrase in the first electronic information and the number d1 of the seventh electronic information to which the related genre is assigned among the sixth electronic information, (b1 / A) <(d1 / c1) does not hold, or determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / c1) + (d1 / b1))> If the predetermined threshold does not hold, the past phrase is An information processing apparatus that determines to delete from the dictionary and deletes the past phrase from the dictionary when it is determined to delete the past phrase from the dictionary.   Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
  Control means for registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information each time the electronic information is stored,
  In the dictionary, a plurality of predetermined genres and other genres that are not any of the plurality of genres are stored as the genres,
  When the electronic information is stored, the control means further includes the number of first electronic information a stored in the storage means between the time when the electronic information is stored and the time before the time a. And among the first electronic information, a number b1 of fifth electronic information to which a related genre that is a genre related to a past phrase that is a word in the dictionary and is not indicated in the electronic information is given, Based on the number c1 of the sixth electronic information including the past phrase in the first electronic information and the number d1 of the seventh electronic information to which the related genre is assigned among the sixth electronic information, (b1 / A) <(d1 / c1) does not hold, or determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / c1) + (d1 / b1))> If the predetermined threshold value does not hold, or the determination value F 1> a judgment value Ff1 regarding other genres (where Ff1 = (2 × (df1 / c1) × (df1 / bf1)) / ((df1 / c1) + (df1 / bf1))), and bf1 is the first value The number of electronic information to which the other genre is assigned in one electronic information, and df1 is the number of electronic information to which the other genre is assigned in the third electronic information). An information processing apparatus that determines to delete a past phrase from the dictionary, and deletes the past phrase from the dictionary when it is determined to delete the past phrase from the dictionary. Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. And, when it is determined to register the word / phrase, control means for registering the word / phrase in the dictionary in association with the given genre,
The control means includes
Each time the electronic information is stored, the number aA of the electronic information, the number bA of the eighth electronic information to which the assigned genre is assigned among the electronic information, and the ninth of the electronic information including the phrase Calculating the number cA of electronic information and the number dA of the tenth electronic information to which the assigned genre is given and including the word / phrase among the electronic information,
When the electronic information is stored for the first time, the aA is calculated as the number a of the first electronic information, the bA is calculated as the number b of the second electronic information, and the cA is calculated as the third electronic information. Calculating as the number c of electronic information, and calculating the dA as the number d of the fourth electronic information,
Thereafter, each time the electronic information is stored, a first multiplication value, which is a value obtained by multiplying the number a of the first electronic information calculated previously by a first weighting factor, and the plurality of electronic information calculated this time. The sum of the second multiplication value, which is a value obtained by multiplying the number aA by the second weighting factor, is calculated as the current number a of the first electronic information, and the previously calculated number b of the second electronic information is calculated. A sum of a third multiplication value that is a value obtained by multiplying the first weighting factor and a fourth multiplication value that is a value obtained by multiplying the number bA of the plurality of pieces of electronic information calculated this time by the second weighting factor, A fifth multiplication value that is a value obtained by multiplying the number c of the third electronic information calculated last time by the number c of the third electronic information calculated this time and the first weighting factor, The sum of the number of electronic information cA and the sixth multiplication value, which is a value obtained by multiplying the second weighting coefficient, is calculated this time. A seventh multiplication value that is a value obtained by multiplying the number d of the fourth electronic information calculated previously by the number d of the third electronic information multiplied by the first weighting factor, and the plurality of electronic information calculated this time An information processing apparatus that calculates a sum of an eighth multiplication value that is a value obtained by multiplying the number dA of the second weighting coefficient by the number dA of the fourth electronic information. The information processing apparatus according to claim 10 ,
The first weighting factor is (n-1) / n (however, when the accumulated number of times the electronic information is stored is smaller than a predetermined specified value, n is the accumulated number, and the accumulated number is When the value is equal to or greater than a predetermined specified value, n is the predetermined specified value.) The predetermined value α is subtracted from the value, and the second weighting coefficient is 1 / n. An information processing apparatus that is an addition. Storage means for storing a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
Storage means for storing electronic information registered on the network and given any of the genres for each predetermined time;
Control means for registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information each time the electronic information is stored,
When the electronic information is stored, the control means further determines whether or not to delete from the dictionary a past phrase that is not indicated by the electronic information among words in the dictionary. A number a1, a number b1 of fifth electronic information to which a related genre that is a genre associated with the past phrase is included in the first electronic information, and a sixth including the past phrase in the first electronic information. When it is determined based on the number c1 of electronic information and the number d1 of seventh electronic information to which the related genre is assigned among the sixth electronic information, and when it is determined to delete the past phrase from the dictionary, Deleting the past phrase from the dictionary;
The control means includes
Each time the electronic information is stored, the number aA of the electronic information, the number b1A of the eleventh electronic information to which the related genre is assigned out of the electronic information, and the past phrase of the electronic information are included. Calculating the number c1A of 12 electronic information and the number d1A of 13th electronic information to which the related genre is given and including the past phrase in the electronic information;
When the electronic information is stored for the first time, the aA is calculated as the number a of the first electronic information, the b1A is calculated as the number b1 of the fifth electronic information, and the c1A is calculated as the sixth electronic information. Calculated as the number c1 of electronic information, and calculated d1A as the number d1 of the seventh electronic information,
Thereafter, each time the electronic information is stored, a first multiplication value, which is a value obtained by multiplying the number a of the first electronic information calculated previously by a first weighting factor, and the plurality of electronic information calculated this time. The sum of the second multiplication value, which is a value obtained by multiplying the number aA by the second weighting factor, is calculated as the current number a of the first electronic information, and the previously calculated number b1 of the fifth electronic information is calculated. A sum of a ninth multiplication value that is a value obtained by multiplying the first weighting factor and a tenth multiplication value that is a value obtained by multiplying the number b1A of the plurality of pieces of electronic information calculated this time by the second weighting factor, An eleventh multiplication value that is a value obtained by multiplying the number c1 of the sixth electronic information calculated previously by the number c1 of the sixth electronic information calculated this time and the first weighting factor, and the plurality of the plurality of the electronic information calculated this time. A twelfth multiplied value that is a value obtained by multiplying the number c1A of electronic information by the second weighting factor. , The 13th multiplication value that is a value obtained by multiplying the previously calculated number d1 of the seventh electronic information by the first weighting factor, and the current number of the sixth electronic information c1 An information processing apparatus that calculates a sum of the calculated number d1A of the plurality of pieces of electronic information and a fourteenth multiplied value that is a value obtained by multiplying the second weighting coefficient as the number d1 of the seventh electronic information. The information processing apparatus according to claim 12 ,
The first weighting factor is (n-1) / n (however, when the accumulated number of times the electronic information is stored is smaller than a predetermined specified value, n is the accumulated number, and the accumulated number is When the value is equal to or greater than a predetermined specified value, n is the predetermined specified value.) The predetermined value α is subtracted from the value, and the second weighting coefficient is 1 / n. An information processing apparatus that is an addition. The information processing apparatus according to any one of claims 1 to 13 ,
An acquisition means for acquiring electronic information registered during the predetermined time from the storage device that collects and stores electronic information registered on the network for each predetermined time;
A genre corresponding to the content indicated by the electronic information is given to the electronic information obtained by the obtaining means, and the electronic information to which only one genre is given among the electronic information to which the genre is given, An information processing apparatus further comprising sorting means stored in the storage means. The information processing apparatus according to any one of claims 1 to 14 ,
In a situation in which a phrase requested to be assigned a genre is received, when the phrase is stored in the storage unit, a granting unit that transmits a grant genre associated with the phrase to the transmission source of the phrase is further included. Information processing device. A dictionary updating method performed by an information processing device,
A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. and, when it is determined to register the word, look including a control step of registering the dictionary in association with the word to the grant genre, and
In the control step, when (b / a) <(d / c) is satisfied, the dictionary update method determines that the word / phrase is registered in the dictionary in association with the assigned genre .   A dictionary updating method performed by an information processing device,
  A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
  Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. And, when it is determined to register the word / phrase, a control step of registering the word / phrase in the dictionary in association with the given genre,
  In the control step, (b / a) <(d / c) is satisfied, and the determination value F = (2 × (d / c) × (d / b)) / ((d / c) + (d / B))> a predetermined threshold, a dictionary update method that determines that the word is registered in the dictionary in association with the given genre.   A dictionary updating method performed by an information processing device,
  A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
  Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. And, when it is determined to register the word / phrase, a control step of registering the word / phrase in the dictionary in association with the given genre,
  In the dictionary, a plurality of predetermined genres and other genres that are not any of the plurality of genres are stored as the genres,
  In the control step,
  (B / a) <(d / c) holds, and
  Determination value F = (2 × (d / c) × (d / b)) / ((d / c) + (d / b))> predetermined threshold value, and
  The determination value F> the determination value Ff relating to other genres (where Ff = (2 × (df / c) × (df / bf)) / ((df / c) + (df / bf))), bf Is the number of electronic information to which the other genre is assigned in the first electronic information, and df is the number of electronic information to which the other genre is assigned in the third electronic information. A dictionary updating method for determining that the word is registered in the dictionary in association with the assigned genre. A dictionary updating method performed by an information processing device,
A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
Each time the electronic information is stored, a control step of registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information,
In the control step, when the electronic information is stored, the number a of the first electronic information stored in the storage unit between the time when the electronic information is stored and the time before the time is a. And among the first electronic information, a number b1 of fifth electronic information to which a related genre that is a genre related to a past phrase that is a word in the dictionary and is not indicated in the electronic information is given, Based on the number c1 of the sixth electronic information including the past phrase in the first electronic information and the number d1 of the seventh electronic information to which the related genre is assigned among the sixth electronic information, (b1 If / a) <(d1 / c1) does not hold, it is determined to delete the past word / phrase from the dictionary, and when it is determined to delete the past word / phrase from the dictionary, the past word / phrase is deleted from the dictionary. ,dictionary New method.   A dictionary updating method performed by an information processing device,
  A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
  Each time the electronic information is stored, a control step of registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information,
  In the control step, when the electronic information is stored, the number a of the first electronic information stored in the storage unit between the time when the electronic information is stored and the time before the time is a. And among the first electronic information, a number b1 of fifth electronic information to which a related genre that is a genre related to a past phrase that is a word in the dictionary and is not indicated in the electronic information is given, Based on the number c1 of the sixth electronic information including the past phrase in the first electronic information and the number d1 of the seventh electronic information to which the related genre is assigned among the sixth electronic information, (b1 / A) <(d1 / c1) does not hold, or determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / c1) + (d1 / b1))> If the predetermined threshold does not hold, the past phrase is A dictionary update method for deleting a past phrase from the dictionary when it is determined to delete from the dictionary and the past phrase is determined to be deleted from the dictionary.   A dictionary updating method performed by an information processing device,
  A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
  Each time the electronic information is stored, a control step of registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information,
  In the dictionary, a plurality of predetermined genres and other genres that are not any of the plurality of genres are stored as the genres,
  In the control step, when the electronic information is stored, the number a of the first electronic information stored in the storage unit between the time when the electronic information is stored and the time before the time is a. And among the first electronic information, a number b1 of fifth electronic information to which a related genre that is a genre related to a past phrase that is a word in the dictionary and is not indicated in the electronic information is given, Based on the number c1 of sixth electronic information including the past phrase in the first electronic information and the number d1 of seventh electronic information to which the related genre is assigned among the sixth electronic information, (b1 / A) <(d1 / c1) does not hold, or determination value F1 = (2 × (d1 / c1) × (d1 / b1)) / ((d1 / c1) + (d1 / b1))> If the predetermined threshold does not hold, or Constant value F1> judgment value Ff1 regarding other genres (where Ff1 = (2 × (df1 / c1) × (df1 / bf1)) / ((df1 / c1) + (df1 / bf1))), and bf1 is the above-mentioned value The first electronic information is the number of electronic information to which the other genre is assigned, and df1 is the number of electronic information to which the other genre is assigned among the third electronic information). A dictionary update method, wherein a past word / phrase is determined to be deleted from the dictionary, and the past word / phrase is deleted from the dictionary when it is determined to delete the past word / phrase from the dictionary.   A dictionary updating method performed by an information processing device,
  A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
  Every time the electronic information is stored, whether or not the words shown in the electronic information are registered in the dictionary in association with the assigned genre given to the electronic information from the time when the electronic information is stored. The number a of first electronic information that is electronic information stored in the storage means before the time point before the time point, and the number of second electronic information to which the assigned genre is assigned among the first electronic information. b, the number c of the third electronic information including the phrase in the first electronic information, and the number d of the fourth electronic information to which the assigned genre is assigned among the third electronic information. And, when it is determined to register the word / phrase, a control step of registering the word / phrase in the dictionary in association with the given genre,
  In the control step,
  Each time the electronic information is stored, the number aA of the electronic information, the number bA of the eighth electronic information to which the assigned genre is assigned among the electronic information, and the ninth of the electronic information including the phrase Calculating the number cA of electronic information and the number dA of the tenth electronic information to which the assigned genre is given and including the word / phrase among the electronic information,
  When the electronic information is stored for the first time, the aA is calculated as the number a of the first electronic information, the bA is calculated as the number b of the second electronic information, and the cA is calculated as the third electronic information. Calculating as the number c of electronic information, and calculating the dA as the number d of the fourth electronic information,
  Thereafter, each time the electronic information is stored, a first multiplication value, which is a value obtained by multiplying the number a of the first electronic information calculated previously by a first weighting factor, and the plurality of electronic information calculated this time. The sum of the second multiplication value, which is a value obtained by multiplying the number aA by the second weighting factor, is calculated as the current number a of the first electronic information, and the previously calculated number b of the second electronic information is calculated. A sum of a third multiplication value that is a value obtained by multiplying the first weighting factor and a fourth multiplication value that is a value obtained by multiplying the number bA of the plurality of pieces of electronic information calculated this time by the second weighting factor, A fifth multiplication value that is a value obtained by multiplying the number c of the third electronic information calculated last time by the number c of the third electronic information calculated this time and the first weighting factor, The sum of the number of electronic information cA and the sixth multiplication value, which is a value obtained by multiplying the second weighting coefficient, is calculated this time. A seventh multiplication value that is a value obtained by multiplying the number d of the fourth electronic information calculated previously by the number d of the third electronic information multiplied by the first weighting factor, and the plurality of electronic information calculated this time A dictionary updating method for calculating a sum of an eighth multiplication value that is a value obtained by multiplying the number dA of the second weighting coefficient by the number dA of the fourth electronic information.   A dictionary updating method performed by an information processing device,
  A storage step of storing in the storage means a dictionary in which the genre and a phrase related to the genre are associated with each other for each genre;
  A storage step of storing in the storage means electronic information that is registered on the network and given any one of the genres for each predetermined time,
  Each time the electronic information is stored, a control step of registering the word / phrase indicated in the electronic information in the dictionary in association with a given genre given to the electronic information,
  In the control step, when the electronic information is stored, it is further determined whether the past words / phrases that are not indicated by the electronic information are deleted from the dictionary among the words / phrases in the dictionary. A number a1, a number b1 of fifth electronic information to which a related genre that is a genre associated with the past phrase is included in the first electronic information, and a sixth including the past phrase in the first electronic information. When it is determined based on the number c1 of electronic information and the number d1 of seventh electronic information to which the related genre is assigned among the sixth electronic information, and when it is determined to delete the past phrase from the dictionary, Deleting the past phrase from the dictionary;
  In the control step,
  Each time the electronic information is stored, the number aA of the electronic information, the number b1A of the eleventh electronic information to which the related genre is assigned out of the electronic information, and the past phrase of the electronic information are included. Calculating the number c1A of 12 electronic information and the number d1A of 13th electronic information to which the related genre is given and including the past phrase in the electronic information;
  When the electronic information is stored for the first time, the aA is calculated as the number a of the first electronic information, the b1A is calculated as the number b1 of the fifth electronic information, and the c1A is calculated as the sixth electronic information. Calculated as the number c1 of electronic information, and calculated d1A as the number d1 of the seventh electronic information,
  Thereafter, each time the electronic information is stored, a first multiplication value, which is a value obtained by multiplying the number a of the first electronic information calculated previously by a first weighting factor, and the plurality of electronic information calculated this time. The sum of the second multiplication value, which is a value obtained by multiplying the number aA by the second weighting factor, is calculated as the current number a of the first electronic information, and the previously calculated number b1 of the fifth electronic information is calculated. A sum of a ninth multiplication value that is a value obtained by multiplying the first weighting factor and a tenth multiplication value that is a value obtained by multiplying the number b1A of the plurality of pieces of electronic information calculated this time by the second weighting factor, An eleventh multiplication value that is a value obtained by multiplying the number c1 of the sixth electronic information calculated previously by the number c1 of the sixth electronic information calculated this time and the first weighting factor, and the plurality of the plurality of the electronic information calculated this time. A twelfth multiplied value that is a value obtained by multiplying the number c1A of electronic information by the second weighting factor. , The 13th multiplication value that is a value obtained by multiplying the previously calculated number d1 of the seventh electronic information by the first weighting factor, and the current number of the sixth electronic information c1 A dictionary updating method for calculating a sum of the calculated number d1A of the plurality of pieces of electronic information and a fourteenth multiplied value, which is a value obtained by multiplying the second weighting factor, as the number d1 of the seventh electronic information. Computer
The program for functioning as each means which the information processing apparatus of any one of Claim 1 to 15 has.