JPH1063679A - Information provision device - Google Patents

Abstract

(57) [Summary] [Problem] To provide information, grasp a range of interest of a user, and provide information in a field of interest of the user without waste and without omission as much as possible. For a keyword in a profile 5, in addition to a weight (weight A) for a specific keyword given to narrow down information, a degree of association between keywords is used to capture a user's interest in a wide range. By calculating another weight (weight B) and using two types of weights to calculate a matching score of the information provided to the user,
Provide information by estimating the range of interest of the user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information providing apparatus for providing information of interest to a user from among information stored in a database without requiring a user to input a search key. It is about.

[0002]

2. Description of the Related Art A conventional information providing system for searching and providing information on a user's interests or interests from information added to a database without a user inputting a search key is described below. Such a method was taken.

(1) A matching score of provided information for a user is calculated using a keyword held in a profile and attribute data of provided information and a keyword weight, and information having a high matching score is provided.

At this time, as a method of determining the weight of the keyword in the attribute data of the provided information, the weight of a specific keyword (a keyword having a narrow meaning such as a proper noun) is increased in order to capture the characteristics of each provided information. In addition, as a method of determining the weight of the keyword in the profile,
The information provided to the user is narrowed down to a narrow range related to the specific keyword, and the information is provided to the user by using the weight of the keyword, which is the attribute data of the provided information obtained so far, as it is.

(2) On the other hand, with respect to each keyword held in the profile, a thesaurus (a higher word of the word,
There has been a method that seeks to provide information by widely catching the interests of the user by tracing the upper words using a dictionary in which lower words are written) and searching by the derived words.

[0006]

However, the conventional information providing apparatus has the following problems. (1) When the weight of a specific keyword is increased in order to capture the characteristics of each provided information, the information to be provided is narrowed down to a narrow range related to the specific keyword. there were.

For example, in the profile of a person who is interested in the field α, a keyword β of a proper noun related to the field α,
It is assumed that γ and δ have been registered. According to the conventional method, β, γ, and δ have large weights, and a search using β, γ, and δ can provide only information related to β, γ, and δ regarding the field α. Therefore, when the user is interested in the field α in general, there is a drawback that information cannot be completely provided to the interest of the user.

(2) In a method of searching for keywords in a profile by using a thesaurus and tracing the upper terms thereof to capture a wide range of users' interests, keywords used for the search are registered in the thesaurus in advance. Need to be kept.

In addition, since the thesaurus itself is merely for deriving the high-order words and low-order words of a word, it is not known how many high-order words on the thesaurus should be traced from the keywords in the profile. There is a drawback that the range of information to be provided becomes too wide if the path is traced too much, and unnecessary information is provided.

[0010] An object of the present invention is to provide a keyword in a profile with a weight (weight A) assigned to provide information by narrowing down to a narrow range related to a specific keyword so far, as well as a user's interest. In order to provide information in a broader sense, another weight (weight B) is calculated using the relevance between keywords obtained from the co-occurrence relationship of keywords, etc., and the information provided to the user by using two types of weights properly The purpose of the present invention is to calculate a matching score of, estimate the range of interest of the user, and provide information on the field of interest of the user as much as possible without omission.

In the above example, if the user's interest is in the field α
To determine whether it is directed only to the keywords β, γ, and δ or to the general field α. If the user's interest is directed to the general field α,
If information other than the information on the keywords β, γ, and δ is provided in the information on the field α, and the user's interest is directed only to the keywords β, γ, and δ, the keyword β is used from the information on the field α. , Γ, δ only.

[0012]

In order to solve the above-mentioned problems, the present invention is configured as follows. (1) An attribute data storage unit for storing attribute data of provided information, a profile as a means for holding a keyword and a weight of the keyword as information indicating an object that the user is interested in and the degree of the interest, A matching score calculating unit that calculates a matching score that is an index of the degree to which the provided information matches the user's interests or interests, using the attribute data and the attribute data. When the provided information is output to the output unit, the fitness evaluation unit evaluates the relevance to the user's provided information, and with reference to the relevance, adds / deletes keywords stored in the profile and weights the keywords In addition to the conventional information providing device consisting of a profile correction unit that changes the -A related word holding unit that holds words and their relevance, a word analysis unit that determines whether a specified word (keyword) is a semantically broad word or a semantically narrow word, and the user's interest The weight calculation unit calculates the weight B of the keywords in the profile in order to provide the information by widely grasping the target of the object. An interest range estimation control unit that performs control to estimate the range and provide information. (2) Alternatively, in place of the above-mentioned related word holding unit, for the keyword added to the information obtained by the user,
A co-occurrence relation holding unit that holds a co-occurrence relation that is a ratio of two keywords appearing in the same information, an association degree calculation unit that calculates an association degree between keywords using the keyword co-occurrence relation, And a related word holding unit that holds the degree of association.

The operation of the present invention is as follows. In the invention of the above (1), it is difficult to obtain information in a specific field by using a word having a broad meaning such as a common noun alone when searching, but it is difficult to obtain a combination of a plurality of words having a broad meaning. By using the property that it is easy to obtain information in a specific field by performing a search, it is relevant to a specific object (a narrow keyword such as a proper noun) that the user is interested in, and , It is possible to provide a wide range of objects of interest to the user.

Information obtained by searching using a plurality of words having a wide range of meanings is provided to the user, and the suitability of the information to the user is evaluated based on how to handle the information provided by the user. Here, when the degree of matching is evaluated to be low, it can be determined that the user's interest is limited to information on a keyword having a specific narrow meaning. On the other hand, if the relevance is highly evaluated, it can be determined that the user's interest may extend to the entire field including the keyword of the word having the broad meaning.

In the invention of the above (2), in order to widely grasp the interest of the user, the related words of the keywords registered in the profile and the degree of their relevance are accumulated from the history of the information acquisition of the user. As described above, without using a synonym dictionary such as a thesaurus or related word dictionary, (1)
In this invention, it is possible to realize that the degree of association between keywords is maintained.

Therefore, by using the present invention, it is possible to finally grasp the range of interest of the user and to provide information on the object of interest to the user as much as possible without omission.

[0017]

BEST MODE FOR CARRYING OUT THE INVENTION

[First Embodiment] FIG. 1 is a block diagram showing one embodiment of a device configuration for implementing the present invention.

An apparatus for carrying out the present invention comprises, as shown in FIG. 1, for example, a data storage unit 1 for storing provided information composed of text, voice, images, etc., its summary information, and its attribute data. Search unit 2 for searching stored information, and information display unit 3 for outputting provided information
And an input unit 4 for receiving a user operation on the device
And a profile 5 that retains keywords and their weights with reference to attribute data of provided information obtained by the user in the past.
A matching score calculating unit 6 for calculating a matching score from the profile 5 and attribute data of the information stored in the data storage unit 1; and a method of handling the information provided by the user and evaluating the information provided by the user. A fitness evaluation unit 7 for evaluating the fitness to the information of the user, and a profile modification unit 8 for modifying the keywords and the weights of the keywords stored in the profile 5 with reference to the fitness evaluated by the fitness evaluation unit 7. In addition to the conventional information providing device composed of, a related word storage unit 9, a word analysis unit 10, a weight calculation unit 11, and the like.

The related word holding unit 9 holds a degree of association between words registered in advance. The word analysis unit 10 determines whether the keyword in the profile 5 is a word having a broad meaning or a word having a narrow meaning.
It is determined whether the keyword in the profile 5 is a common noun or a proper noun, for example.

The weight calculation unit 11 refers to the related word holding unit 9 based on the determination result of the word analysis unit 10 and provides the profile 5
Is calculated.

The interest range estimation control unit 12 includes the weight calculation unit 1
Using the weight B calculated in step (1), information provision control is performed by estimating the range of interest of the user. FIG.
2 shows an example of storage of provision information, its summary information, and its attribute data in a data storage unit 1. The data storage unit 1 stores a summary of the information and attribute data in addition to the body of the provided information as shown in FIG. The attribute data is a keyword and its weight.

FIG. 3 shows an example of storing keywords and their weights in the profile 5. The profile 5 stores a first weight A and a second weight B for each keyword, for example, as shown in FIG. Note that the weight B is dynamically calculated from the weight A when the information is provided, and therefore does not necessarily need to be stored in the profile 5 in advance.

FIG. 4 is an explanatory diagram of the semantic relevance.
The semantic relevance of the related word held by the related word holding unit 9 is a numerical value indicating how close a certain keyword is to another keyword and is, for example, n keywords {k ₁ , For k ₂ , ..., k _n }, the keyword k
_Assuming that the degree of association between _i and the keyword k _j is w _ij , it is expressed as a symmetric matrix as shown in FIG. Where w _ij =
_wji .

FIG. 5 is a basic flow of the processing of the present apparatus.
In FIG. 5, first, in a process P1, the total number of information provided by the interest range estimation control unit 12 is determined. The total number of information to be provided may be a case where the user inputs the total number or a case where the total number is provided in advance.

In the process P2, the interest range estimation control unit 12 determines the ratio of providing information with the weight A and the ratio of providing information with the weight B. The method of setting the weight A / weight B will be described later.

In providing information using the weight A, processing P3
And the profile 5 by the matching score calculation unit 6
Is compared with the keywords stored in the attribute data, and the weights of the matching keywords are multiplied (the weight of the keyword in profile 5 is weight A). Is calculated.

In providing information using the weight B, the weight calculator 11 calculates the weight B from the weight A in the process P4. This calculation is performed as follows, for example. When the keywords in profile 5 are composed of m proper noun keywords (P ₁ , P ₂ ,..., P _m ) and n common noun keywords (Q ₁ , Q ₂ ,..., Q _n ) , The weight A of the proper noun is (p ₁ , p ₂ ,..., P _m ), and the weight A of the common noun is (q ₁ , q ₂ ,..., Q _n ).

Let R be the degree of association between proper noun P _i and common noun Q _j
Assuming that the weights B of the proper noun P _i and the common noun Q _j are x _i and y _j , respectively, x _i and y _j are obtained by the following equations.

[0029] _{x i = 0 (i = 1,2} , ..., m) y j = q j ΣR kj p k (j = 1,2, ..., n) [Dadashi, sigma from k = 1 to m Next, in the process P5, the matching score calculation unit 6
By comparing the keyword held in the profile 5 with the keyword of the attribute data, multiplying each weight of the matching keyword (using the weight B for the weight of the keyword of the profile 5), and taking the sum thereof, Calculate the matching score.

In the process P6, the summary information is presented to the information display unit 3 based on the matching scores calculated in the processes P3 and P5. Next, in process P7, the user is directly requested to evaluate the provided information. Or, as shown in FIG.
Quantitatively measure the content of the acquired information and the information browsing time, and evaluate the relevance. That is, as shown in FIG. 6, when the user requests and obtains not only the summary information but also the body of the provided information, it is evaluated that the user has a high degree of conformity. In addition, it is evaluated that the longer the information viewing time is, the higher the user's fitness is. The information browsing time can be measured by, for example, the time from when the provided information is transmitted to when there is a request from the user next time.

In the process P8, the profile 5 is corrected by the process shown in FIG. 7 based on the result evaluated by the suitability evaluation section 7. In the process P9, it is determined whether or not the information provision is continued. If the information provision is continued, the process returns to the process P1 and the process is repeated similarly. If the information provision is not continued, the process ends.

FIG. 7 shows a flow of modifying the profile 5 in the process P8 shown in FIG. Since the weight B is calculated every time based on the weight A, it is not always necessary to save the weight B. The correction of the weight in the profile 5 is performed only for the weight A. However, for the information provided by calculating the matching score using the weight B, only the “common noun” is considered when calculating the matching score, and therefore, the weight A of the keyword of the common noun among the corresponding keywords in the profile is used. Only change.

In the process P21 of FIG. 7, it is determined whether or not the degree of conformity to the user is large. If the degree of adaptation to the user is high, the process proceeds to process P22, and if it is low, the process proceeds to process P27.

In the process P22, the profile 5 is searched with the keyword of the attribute data obtained from the data storage unit 1. The profile 5 is determined by the next process P23.
If a corresponding keyword exists in the profile 5, the weight of the corresponding keyword in the profile 5 is increased by a process P24. If there is no corresponding keyword, process P2
5, the corresponding keyword and its weight are added to the profile 5.

In the next process P26, it is determined whether there is any other information to be evaluated, and if there is no more information to be evaluated, the process ends. If there is information to be evaluated, the processes in and after the process P22 are repeated in the same manner.

On the other hand, if the degree of conformity to the user is low, the profile 5 is searched by the keyword of the attribute data in process P27. Then, by processing P28, the weight of the corresponding keyword in the profile 5 is reduced. Next, if it is determined in the process P29 that there is no more information to be evaluated, the process is terminated. If there is, the process returns to the process P27 to repeat the same process.

An example of a method of setting the weight A / weight B relating to the processing P2 and the like in FIG. 5 is a case where the user determines the total number of information to be provided and a case where all information is provided as a list in descending order of the matching score. Will be described.

When the User Determines the Total Number of Information Provided In this case, first, the ratio of the information provided by weight A to the information provided by weight B is stored. When providing information to a user, information provided by calculating a matching score with weight A and information provided by calculating a matching score with weight B are mixed in accordance with the stored ratio. And correct the ratio based on the user's fitness.

For example, it is assumed that the total number of information to be provided is 10 and the ratio of the provided information with the weight A and the weight B is 7: 3. The fitness score is calculated with the weight A, and seventy (70%) pieces of information are provided in descending order of the fitness score. The fitness score is calculated with the weight B, and three (3)
Percent) information.

The degree of conformity is calculated, and the ratio calculated as follows is used when the next time is provided. For example, the ratio of information to be provided next time is determined by the following equation. Information provision by weight A: Information provision by weight B = (sum of matching scores of information provided by weight A) / (weight A
Ratio): (sum of matching scores of information provided by weight B) / (ratio of weight B) These calculations are performed by the interest range estimation control unit 12.

When all information is provided as a list in descending order of the matching score In this case, the ratio of the information provided by weight A and the information provided by weight B is not determined. A part for calculating the matching score with the weight A and presenting the summary information as a list in descending order of the matching score, and a part for calculating the matching score with the weight B and presenting the summary information as a list in descending order of the matching score , Provided and provided on the same screen.

As described above, the interest range estimation control unit 12
Provides information using the weight B, and performs control to estimate the user's interest in the entire field or only specific keywords when evaluating the user's fitness, and to provide the information. .

As described in the present embodiment, a method of providing information by narrowing down to a narrow range by giving a large weight to proper nouns, and a method of providing common nouns having a high degree of relevance to proper nouns in profile 5 Giving a large weight to provide information by combining the method of providing information with a wider view of the user's interests, providing feedback by evaluating the user's fitness at that time, and providing feedback By specifying the range of interest of the user, it becomes possible to provide information on the object that the user is interested in with as little waste as possible.

[Second Embodiment] Next, a second embodiment of the present invention will be described.
An embodiment will be described. FIG. 8 is a block diagram showing a device configuration of the second embodiment in which address information indicating the location of the provided information is stored instead of storing the provided information itself in the data storage unit 1.

The data storage unit 1 stores address information, summary information, and attribute data. As address information,
For example, there is a URL on the Internet. In the case of this embodiment, when a matching score is calculated based on the attribute data in the data storage unit 1 and information is provided, the information is acquired via the communication unit 14 based on the address information in the data storage unit 1. .

The other points are the same as in the first embodiment. [Third Embodiment] FIG. 9 is a block diagram showing the configuration of a device according to a third embodiment in the case where a keyword and text information without its weight are externally stored as provided information.

In this embodiment, text information is input from the information input unit 15 to the data storage unit 1. Keywords are extracted by the information analysis unit 16, weighted by the keywords, and stored in the data storage unit 1.

The co-occurrence relation holding unit 17 holds the co-occurrence relation of the keywords assigned to the information acquired by the user.
The co-occurrence relationship refers to the rate at which two keywords appear in the same information when handling a plurality of pieces of information. Here, the co-occurrence rate is the rate at which two keywords appear in common with respect to information obtained by a user.

The relevance calculating unit 18 uses the co-occurrence relationship of keywords in the information obtained by the user, for example, in the method described in the following document [1], pp. 63-66.
The degree of relevance between keywords is calculated by replacing the document with the provided information and the degree of similarity with the degree of relevance, and stores the calculated degree in the relevant word storage unit 9.

That is, a set of keywords in the profile 5 is set as K = {k ₁ , k ₂ ,..., K _M }, and a set of information acquired by a user in which any of these appears is D = { d _1, d _2, ..., and d _N}. Then, the appearance of the keywords k _u and k _v ∈K in the information set D is expressed as follows using the vector of information obtained by the user.

K _u : v _ku = (δ ₁ ^u , δ ₂ ^u ,..., Δ _{i
^{u, ..., δ N u)}} k v: v kv = (δ 1 v, δ 2 v, ..., δ i v, ..., δ N
^{v) _{However, δ i u (δ i v}} ), 1 ≦ i ≦ N becomes 0 to 1, in other cases when the keyword k _u (k _v) is included in the information d _i. From this, the degree of semantic relevance between keywords is a related measure s: K × K → [0, 1] that gives a value proportional to the number of pieces of information acquired by the user that occurs in the information set D in common. Can be calculated by determining To formulate the related measure s, for example, a well-known method such as a set-theoretic measure or a cosine function, which has a simple calculation method and a wide application range, may be used.

The following is the same as the above-described first embodiment. References [1] "Information Search", pp.63-66, Tetsuro Ito, Shokodo (1986).

[0053]

As described above, according to the present invention,
Without inputting a search key, information related to a user's interest can be provided from the information stored in the database without waste and without omission as compared with the conventional device. .

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an apparatus configuration according to a first embodiment.

FIG. 2 is a diagram illustrating an example of storage of provided information, its summary information, and its attribute data in a data storage unit according to the first and third embodiments.

FIG. 3 is a diagram illustrating a storage example of keywords and their weights in a profile according to the first, second, and third embodiments.

FIG. 4 is an explanatory diagram of a semantic association degree.

FIG. 5 is a basic flow of a process in the first, second, and third embodiments.

FIG. 6 is a diagram illustrating an example of calculating a matching score in the first, second, and third embodiments.

FIG. 7 is a flowchart of profile correction in the first, second, and third embodiments.

FIG. 8 is a block diagram illustrating a device configuration according to a second embodiment.

FIG. 9 is a block diagram illustrating a device configuration according to a third embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Data storage part 2 Search part 3 Information display part 4 Input part 5 Profile 6 Relevance score calculation part 7 Relevance evaluation part 8 Profile correction part 9 Related word holding part 10 Word analysis part 11 Weight calculation part 12 Interest range estimation control part 14 Communication unit 15 Information input unit 16 Information analysis unit 17 Co-occurrence relationship holding unit 18 Relevance calculation unit

Claims (2)

[Claims] 1. An information providing apparatus for automatically searching and providing information on a user's interests or interests from information added to a database without a user inputting a search key. Means for holding, as attribute data, the keywords assigned to the information provided to the user and the weights of the keywords, and the keywords and the first weights of the keywords as the objects representing the objects of interest and the degree of interest. Means for storing a profile as means for storing, means for storing semantic relevance for a specified keyword, and means for determining whether a word is a word having a broad meaning or a word having a narrow meaning with respect to the keywords in the profile Judgment and, based on the judgment result, referencing the relevance held by the means for retaining semantic relevance to capture the user's interest in a wide range Means for calculating a second weight for the information to be provided, and using the weight obtained from the attribute data of the provided information and the first and second weights related to the keywords in the profile, to provide the provided information with the interest or interest of the user. Calculates the relevance score, which is an index of the degree to which the information matches, and provides information based on the calculated relevance score, means for evaluating the relevance of the information provided to the user, and refer to the relevance. Means for adding / deleting a keyword registered in the profile and changing the weight of the keyword, and by selectively using the first weight and the second weight, the interest of the user is increased. An information providing apparatus characterized in that information is provided by estimating a range in which the information is provided. 2. An information providing apparatus for automatically retrieving and providing information on a user's interests or interests from information added to a database without a user inputting a search key. Means for holding, as attribute data, the keywords assigned to the information provided to the user and the weights of the keywords, and the keywords and the first weights of the keywords as the objects representing the objects of interest and the degree of interest. About the profile which is the means to hold and the keyword given to the information acquired by the user, 2
Means for retaining a co-occurrence relationship that is a ratio of two keywords appearing in the same information; means for calculating the relevance between keywords using the keyword co-occurrence relationship; and means for retaining the relevance between keywords And determining whether the word in the profile is a word having a broad meaning or a word having a narrow meaning with respect to the keyword in the profile, and storing the degree of relevance between the keywords based on the result of the determination. Means for calculating a second weight for capturing a user's interest in a wide range with reference to the degree of association, and the first and second weights obtained from attribute data of the provided information and the keywords in the profile. Using the weights of 2, a relevance score, which is an index of how much the provided information matches the interests and interests of the user, is calculated, and the information is calculated based on the calculated relevance score. Comprising means for providing, and means for evaluating the fitness of providing information to the user, and means for changing the weight of the added or deleted, and keywords of the keywords with reference to fit registered in the profile,
An information providing apparatus characterized in that information is provided by estimating a range of interest of a user by selectively using the first weight and the second weight.