JP2001134582A - Estimating device for genre of topic of news and topic presenting device for individual - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a estimating device for genre of topic of news which accurately estimate how much a topic of news is concerned in plural genres and a topic presenting device for an individual which can effectively present the topic of news so that it matches the taste of the individual. SOLUTION: According to information obtained from words included in a news original constituting a topic, how much the topic of the news is concerned in plural genres is estimated (3(3a, 3b)).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a news topic genre estimating apparatus for estimating a genre to which a news topic belongs, and a personal topic presenting apparatus for presenting a news topic in accordance with personal preference.

[Summary of the Invention] The present invention relates to information classification and retrieval using information included in news and personal preference information. That is, the present invention uses information obtained from words included in news manuscripts constituting a topic, so that the topic of the news is divided into a plurality of genres (for example, five genres of politics, economy, society, international, and sports). Is estimated. Using the estimation results and personal preferences input in the five genres of politics, economy, society, international and sports, news topics are presented effectively to match personal preferences. It is.

[0003]

2. Description of the Related Art In the category of news topics, conventionally, a method for uniquely determining which genre the user belongs to has been proposed.

[0004] In a personal TV program selection method,
Conventionally, a method has been proposed in which a genre or keyword is manually added to data to be searched, and personal news is selected by pattern matching between the added genre or keyword and a keyword included in personal preference information. In this method, matching with personal preference information
Pattern matching by perfect matching is used.

[0005]

However, when the genre of a news topic is classified by a conventional method of uniquely determining to which genre the news belongs, the actual news may belong to a plurality of genres. There is a problem in that it is difficult to accurately express how much of the news topic is related to a certain genre.

[0006] Further, in the conventionally proposed personal television program selection method, a keyword or genre must be manually added to each program, and the operation is complicated. In addition, since matching with personal preference information is based on pattern matching based on perfect matching, the priority of program selection cannot be clearly expressed. Therefore, it is difficult to present a personal news topic by a conventional method so as to match personal preferences.

The present invention has been made in view of the above circumstances, and has as its object to provide a news topic genre estimating apparatus capable of accurately estimating how much a news topic relates to a plurality of genres.

It is another object of the present invention to provide a personal topic presenting apparatus capable of effectively presenting news topics in accordance with personal preferences.

[0009]

In order to achieve the above object, according to the present invention, a news topic genre estimating apparatus according to claim 1 calculates a contribution degree of a word included in a news manuscript constituting a topic to a plurality of genres. A genre for extracting to which of the plurality of genres a news topic belongs based on the genre contribution calculating means for each of the appearing words and the calculated genre contribution for each word and the topics extracted from the news manuscript. Extraction means.

According to a second aspect of the present invention, in the news topic genre estimating apparatus according to the first aspect, the genre-specific contribution calculating means for the appearing word, the appearance frequency of the word included in the news manuscript constituting the topic is a predetermined number of times or more. And a means for calculating the genre-related contribution of the extracted words.

According to a third aspect of the present invention, in the news topic genre estimating apparatus according to the second aspect, the news topic genre extracting means includes a word serving as an element of data expressing the topic from a topic extracted from the news manuscript. Means for extracting the words for which the genre contributions are calculated by the genre contribution calculation means for the appearing words from the extracted words, and genre contributions for the selected words. And a means for calculating a topic-specific genre score by using the calculated topic-specific genre score.

According to the present invention, in order to accurately represent a genre of a news topic,
The genre to which the news belongs is not determined uniquely, and information obtained from words included in the news manuscript constituting the topic is used, so that the topic of the news is divided into a plurality of genres (for example, politics, economy, society, international, sports) The five genres are estimated.

According to a fourth aspect of the present invention, there is provided the news topic genre estimating apparatus according to any one of the first to third aspects, an individual preference input interface for inputting personal preferences for each of a plurality of genres, and the news topic genre estimating apparatus. It is characterized by comprising personal news topic presenting means for presenting a news topic so as to match personal preferences based on the obtained genre estimation result and the input information of the personal preference input interface.

According to the fourth aspect of the present invention, personal preferences are input using, for example, five genres of politics, economy, society, international and sports using the personal preference input interface, and the personal information and By using the topic genre estimation result, a news topic can be effectively presented so as to match personal preferences.

[0015]

FIG. 1 is a block diagram showing a configuration of a news topic genre estimating apparatus and a personal topic presenting apparatus according to an embodiment of the present invention.

In FIG. 1, a personalized topic presenting apparatus 1 of this embodiment receives a digitized news manuscript, classifies articles included therein into similar items, and extracts a topic by extracting a topic. And the past news manuscript and the output of the news topic extraction unit 2 (topics within a specific period (topic center of gravity vector)), the topic is divided into a plurality of genres (for example, five categories of politics, economy, society, international, and sports). A news topic genre extraction unit 3 for estimating the degree of involvement in each news genre using a word weighted by a degree of importance that changes with time, and Tastes, such as politics, economy,
Using the personal preference input interface 4 (see FIG. 8) for inputting in five genres of social, international, and sports, the result of the genre estimation of the news topic, and the personal preference input from the personal preference input interface 4 And a personalized news topic presenting unit 5 that effectively presents news topics so as to match personal preferences.

The news topic genre estimating apparatus 3 of the embodiment includes a genre-specific contribution calculator 3a for calculating a genre-specific contribution for each word included in a past news manuscript; Contribution calculator 3a
And a news topic genre extraction unit 3b that extracts a news topic genre using the word-based genre contribution degree output by the term and a topic (center-of-gravity vector) in a specific period.

Hereinafter, the operation of the news topic genre estimating apparatus and the personal topic presenting apparatus of the present embodiment will be described with reference to FIGS.
This will be described with reference to FIG. FIG. 2 is an example of topic extraction. FIG. 3 is an example of a centroid vector obtained by the news topic extraction unit. FIG. 4 is a flowchart of a genre-based contribution calculation process for appearing words. FIG. 5 is a flowchart of the news topic genre extraction process. FIG.
It is an example of the output result of the contribution to each genre of a topic. FIG. 7 shows a comparison result of the contribution degree of each news topic “missile launches of North Korea and the Democratic People's Republic of Korea” to each genre, compared with the conventional method. FIG. 8 is an example of the personal preference input interface. FIG. 9 and FIG. 10 are examples of results of personal news topic presentation.

First, a news topic extraction unit 2 extracts a news topic. That is, the news topic extraction unit 2
Is the weight of the words contained in the news manuscript, for example, the monthly importance Weight
(W) is obtained by the following equation (1), and articles included in the news manuscript are classified into similar items based on the equation (1), and topics are extracted.

[0020]

(Equation 1)

In the equation (1), n is the frequency at which the word W appears in a certain month, and e is the expected value. Also, N
Is the total number of news manuscripts in the target month, and df (W) is the number of manuscripts in which the word W appears.

The extracted topic in a specific period is a news genre estimating device 3 together with the center of gravity vector of the topic.
Is output to In FIG. 2 showing an example of topic extraction, 19
Eight topics in September 1998 are shown, including "Missile launches of North Korea and the Democratic People's Republic of Korea."
FIG. 3 shows an example of the centroid vector of the topic.

The news genre estimating apparatus 3 determines how much the topic extracted in this way relates to a plurality of genres, not a single genre. Use and estimate. Here, for example, five genres of politics, economy, society, international, and sports are taken as a plurality of genres.

In this estimating process, a numerical value is extracted as to how much the genre belongs to a plurality of genres. Therefore, the genre contribution of each word is calculated by the genre contribution calculation unit 3a of the appearing word, and the topic genre is determined by the news topic genre extraction unit 3b using the genre contribution of each word.

The genre contribution calculating section 3a of the appearing word obtains the genre contribution by word according to the procedure shown in FIG. 4 and generates a word table. In FIG. 4, words are extracted from past news manuscripts (step ST1), and the news manuscripts generated in the past are identified by the name of the department that created the news (political, economic, social, international, sports). Department) information. Here, a news manuscript for one year in 1997 is used as learning data, and for every word whose appearance frequency is a certain number of times or more (for example, 20 times or more) (step ST2), Is calculated by the following equation (2) (step ST3) and registered in the word table (step ST4).

The genre contribution of words indicating how much the word w contributes to each genre is represented by r1 as the frequency at which a word w appears in "politics", r2 as the frequency at which a word w appears in "economy", , The frequency of occurrences in "international" is r4, and the frequency of occurrences in "sports" is r5.

[0027]

## EQU2 ## Contribution (w, politics) = r1 / (rl + r2 + r3 + r4 + r5) Contribution (w, economy) = r2 / (rl + r2 + r3 + r4 + r5) Contribution (w, society) = r3 / (rl + r2 + r3 + r4 + r5) Contribution (w, international) = R4 / (rl + r2 + r3 + r4 + r5) Contribution (w, sports) = r5 / (rl + r2 + r3 + r4 + r5) (2)

By this processing, the genre-based contribution can be calculated for every word in which the frequency of appearance of the word that appeared in the 1997 news manuscript is equal to or more than a certain number of times, and the genre-based contribution is registered for each word. Can be completed (step ST5, step ST6).

Next, a news topic genre extraction unit 3
In FIG. 2B, a topic obtained by the news topic extraction unit 2 is used for the topic obtained by the news topic extraction unit 2 by using the genre-based contribution degree of a certain number of times or more (for example, 20 times or more) in the 1997 news manuscript. The degree of contribution to each genre is calculated by the procedure shown in FIG.

In FIG. 5, a topic for a specific period is acquired from the news topic extraction unit 2 (step ST11), and words which are elements of a center-of-gravity vector constituting the topic are extracted (step ST12, see FIG. 3).

If the extracted "word as an element of the center-of-gravity vector" exists in the word table (step ST13: Yes), the genre contribution degree (Wi, politics), contribution degree of the word (Wi, economy), contribution (Wi, society), contribution (Wi, international), contribution (W
i, sports) is applied to the following equation (3) to calculate scores Trl, Tr2, Tr3, Tr4, Tr5 for topics of politics, economy, society, international and sports (step ST14). Equation (3) defines words that are elements of the barycenter vector of one topic T as wl to Wn.

[0032]

(Equation 3)

When the process of calculating the genre-based scores by the equation (3) is completed for all the words that are the elements of the center-of-gravity vector (step ST15: Yes), the genre-based scores Trl, Tr2, Tr3, Tr4, and Tr5 are set as follows. The contribution T of the topic T to the politics, economy, society, international, and sports is calculated by applying the formula (4).

[0034]

[Equation 4] Contribution (T, politics) = Trl / (Tr1 + Tr2 + Tr3 + Tr4 + Tr5) Contribution (T, economy) = Tr2 / (Tr1 + Tr2 + Tr3 + Tr4 + Tr5) Contribution (T, society) = Tr3 / (Tr1 + Tr2 + Tr3 + TI4 + Tr5) Contribution (T, international) = Tr4 / (Tr1 + Tr2 + Tr3 + Tr4 + Tr5) Contribution (T, sports) = Tr5 / (Trl + Tr2 + Tr3 + Tr4 + Tr5) (4)

FIG. 6 is an example of an output result of the contribution of each topic to each genre in September 1998 shown in FIG. FIG. 7 shows a result of comparing the fifth topic “missile launching of North Korea and the Democratic People's Republic of Korea” with the conventional method.

In the conventional method, how much a word included in one news belongs to each genre is calculated based on past learning data, and the genre is estimated by the total ratio. Therefore, as shown in FIG. 7B, it belongs to "sports" and "economics" which have no relation at all, and it is difficult to accurately estimate the genre.

On the other hand, as shown in FIG. 7A, in the method according to the present embodiment, the ratios belonging to “sports” and “economy” are lower than those of the conventional method, and “politics” and “ You can estimate the percentages belonging to "economy", "international" and "society". Thus, it can be seen that the genre can be more accurately estimated by the method according to the present embodiment than by the conventional method.

Next, the personal preference input interface 4 is used to input personal preferences of the viewer. Personal preferences are input in five genres, for example, politics, economy, society, international, and sports. The preference is input as a numerical value from 0 to 100, for example. FIG. 8 shows the ratio of each genre when the preference of each of the five genres is input as a value of 30.

The personal news topic presenting section 5 searches for news that matches the personal preferences by using the input personal preferences and the genre extracted by the news topic genre extraction section 3b. .

First, a score Score (A, T) assigned to the topic T for the taste of the individual A is obtained according to the following equation (5). In equation (5), the value pol is the individual A for politics.
Is the value of the preference. The value eco is a value of the individual A's preference for the economy. The value soc is a value of individual A's preference for society. The value int is a value of the preference of the individual A for the international. The value spo is a value of personal A's preference for sports.

[0041]

(Equation 5)

In equation (5), the squared value is used in order to emphasize individual preference, but when emphasizing the importance of a topic, the following equation (6) is used, for example. By taking, for example, a logarithmic function (log) of equation (6), the importance of the topic can be further emphasized.

[0043]

(Equation 6)

A topic having a large value in the equations (5) and (6) can be determined to be a topic suited to individual tastes. 9 and 1
0 shows the result of personal news topic presentation calculated by equation (5).

In FIG. 9, the genre of the taste of the viewer A is
It is shown that international topics are mainly presented because of the high international prices. Also, in FIG.
Since the genre of the taste of the viewer B is given a large value for sports and society, it is shown that topics of domestic sports and society are mainly presented.

In the embodiment described above, five genres of politics, economy, society, international and sports are taken as genres, but it is needless to say that the present invention is not limited to these, and the number and range of genres are not limited to these. Is optional.

[0047]

As is clear from the above, according to the first to third aspects of the present invention, a plurality of news topics are obtained by utilizing information obtained from words included in a news manuscript constituting a topic. It is possible to estimate how much the genre is involved.

According to the fourth aspect of the present invention, a personal preference is input into a plurality of genres by using the personal preference input interface, and the personal information and the topic genre estimation result are used. Thereby, it is possible to effectively present the topic of the news so as to match the personal preference.

[Brief description of the drawings]

FIG. 1 is a configuration block diagram of a news topic genre estimation device and a personal topic presentation device according to a prevailing mode of the present invention.

FIG. 2 is an explanatory diagram illustrating an example of topic extraction.

FIG. 3 is an explanatory diagram illustrating a center-of-gravity vector obtained by a news topic extraction unit.

FIG. 4 is a flowchart of a genre-based contribution calculation process for appearing words.

FIG. 5 is a flowchart of a news topic genre extraction process.

FIG. 6 is an explanatory diagram illustrating an output result of a degree of contribution to each genre of a topic;

FIG. 7 is an explanatory diagram showing a comparison result of the contribution degree to the respective genres of the news topic “missile launches of North Korea and the Democratic People's Republic of Korea” with the conventional method.

FIG. 8 is an explanatory diagram illustrating a personal preference input interface.

FIG. 9 is an explanatory diagram exemplifying a result of personal news topic presentation.

FIG. 10 is an explanatory diagram exemplifying a result of personal news topic presentation.

[Explanation of symbols]

 1 Personal Topic Presentation Device 2 News Topic Extraction Unit 3 News Topic Genre Extraction Unit 3a Contribution Calculation Unit for Appearance Words by Genre 3b News Topic Genre Extraction Unit 4 Personal Preference Input Interface 5 Personal News Topic Presentation Unit

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] November 11, 1999 (1999.11.
11)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0037

[Correction method] Change

[Correction contents]

On the other hand, as shown in FIG. 7A, in the method according to the present embodiment, the ratios belonging to “sports” and “economy” are lower than those of the conventional method, and “politics” and “ It can estimate the ratio of "international" and "society".
Thus, it can be seen that the genre can be more accurately estimated by the method according to the present embodiment than by the conventional method.

Continuation of the front page (72) Inventor Masahiro Shibata F-term (Reference) 5B075 ND03 NR12 PR04 in the Japan Broadcasting Corporation Broadcasting Research Institute, 1-10-11 Kinuta, Setagaya-ku, Tokyo

Claims (4)

[Claims] 1. A genre-contribution calculating unit that calculates the contribution of a word included in a news manuscript constituting a topic to a plurality of genres, and an extracted word-based genre contribution and extraction from the news manuscript. A news topic genre estimating apparatus, comprising: genre extracting means for extracting which of the plurality of genres the news topic belongs to, based on the performed topic. 2. The news topic genre estimating apparatus according to claim 1, wherein the genre-specific contribution calculating means for the appearing words includes words whose appearance frequency is equal to or more than a predetermined number of times from words included in news manuscripts constituting the topic. A news topic genre estimating apparatus, comprising: means for extracting; and means for calculating a degree of contribution of each genre for the extracted word. 3. The news topic genre estimating apparatus according to claim 2, wherein the news topic genre extracting means extracts words which are elements of data expressing the topic from the topics extracted from the news manuscript. Means for selecting words whose genre contributions are calculated by the genre contribution calculation means for the appearing words from the extracted words; and topics using the genre contributions of the selected words. A news topic genre estimating apparatus, comprising: means for calculating a genre score for each topic; and means for calculating a contribution rate for each genre of the topic using the calculated score for each topic genre. 4. A news topic genre estimating apparatus according to claim 1, a personal preference input interface for inputting personal preferences for each of a plurality of genres, and a news topic genre estimating apparatus. Personal news topic presenting means for presenting a news topic so as to match personal preferences based on a genre estimation result and input information of the personal preference input interface.