JP6721926B1 - Content search interface, interface display method, and social network service using these - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a content search interface and a display method of the interface, and a social network service using these, which can discover another related topic simply by acting with interest. SOLUTION: When acquiring a vector expression from a database, a step of generating a large number of small models that refer only to contents including some text, in addition to a large model that targets all data, and a small model. And a step of organizing a large model into a nested relation on a database, and a method of displaying an interface is applied.

Description

The present invention relates to a content search interface, a method for displaying the interface, and a social network service using these.

Various methods are used in so-called social network services so that users can search for desired contents and exchange information.

For example, Patent Document 1 discloses a tag cloud configuration device capable of expanding and contracting a character string length, which makes it easier to use a search system using a tag cloud and further suppresses the time required to find a desired Web page. ..

JP 2013-77327 JP

According to the tag cloud configuration device of Patent Document 1, a search system that uses a tag cloud can be easily used, and a tag cloud that can expand and contract a character string length that can further suppress the time until finding a desired Web page is provided. To be done.
However, it is not possible to organically derive and propose another content related to the content of the user's interest.
The present invention has been made in view of the above circumstances, and a content search interface and a display method of the interface that can discover another related topic by merely acting with interest, and a method of using the same. One of the purposes is to provide the social network services that were provided.

According to the interface display method of the present invention, when a vector expression is obtained from a database, a large number of small models that refer only to the contents including some text are generated in addition to the large model that targets all data. And a step of organizing the small model into a nested relationship with the large model on the database.

Further, the interface display method of the present invention includes a step of creating a map of q×q=n cells (where q is an odd number of 3 or more), and an arbitrarily selected content X is placed at the center of the map. To create an important content list and fill the outermost one of the centers when the nearest neighbor search result (top n) of X is set to X2, X3, …, Xn from the highest rank. Steps of extracting two combinations Xa and Xb having the lowest cos similarity from among X2 to X8 necessary for arranging them diagonally, and two Xc, having the lowest cos similarity next to Xa and Xb, Regarding the step of extracting Xd and arranging it on the other diagonal, and for the remaining points, select 3 or 2 surrounding surrounding points, find the average value, and perform the nearest neighbor search from the average value. Then, the step of selecting and arranging the closest Xe, and the step of further executing the nearest neighbor search also in the outermost X9 to X24 and repeating this until reaching Xn are included. At this time, the diagonal is used only during X2 to X8 of the first round, and the diagonal operation is not used from the second round.

Further, the interface display method of the present invention includes a step of creating a map of q×q=n cells (where q is an odd number of 3 or more), and an arbitrarily selected content X is arranged at the center of the map. If step 1 and a list of important contents are created and the nearest neighbor search result (top n) of X is set to X2, X3, ..., Xn from the highest rank, the result is one outside the center. Step 2 of extracting the two combinations Xa and Xb with the lowest cos similarity degree and arranging them diagonally from among X2 to X9 required to fill in, and having the lowest cos similarity degree after Xa and Xb Step 3 where two Xc and Xd are extracted and placed on the other diagonal, and for the remaining points, select the surrounding 3 or 2 points that have already been determined, calculate the average value, and from that average value recently running neighbor searching includes the step 4 of placing select the nearest Xe, further executing the steps 3 to 5 even with the outside of X10～X25, step 5 repeating this until the Xn, a. At this time, the diagonal is used only in the first lap X2 to X9, and the diagonal operation is not used from the second lap.

Further, the content search interface of the present invention uses the display method of the above interface.

Further, the social network service of the present invention uses the above interface display method.

According to the present invention, it is possible to discover another related topic simply by acting with interest.

The figure which shows the content search interface of embodiment of this invention. The figure which shows the state which changed from FIG. The figure which shows the calculation method of a coordinate. The figure which shows the state which dragged the content to the area called a door. The figure which shows the state which dragged another content in the state of FIG. The figure which shows the portal screen of the content search interface of this invention. The figure which shows the social network service using the content search interface of embodiment of this invention. The figure which shows the interface displayed to another user in the state of FIG.

The content search interface according to the embodiment of the present invention will be described. FIG. 1 is a diagram showing a screen display of a content search interface according to the embodiment of the present invention. As shown in FIG. 1, the content X, that is, the character "tart" in this example, is displayed at the center of the screen. This content X is content that the user is interested in, and this is the starting point. For example, the content X is determined by a user inputting a search word and performing a search.

Below the "tart" is a colored circle, which represents itself. This circle is in the center of the screen, so even if the content moves, the circle does not move. That is, it moves relative to the content.

Other contents are arranged in a grid pattern on the top, bottom, left and right of the "tart". In this example, "cream puff", "white chocolate", "rare cheese", and "roll cake" are adjacent to the top, bottom, left and right of the "tart". Next to that, contents such as "Custard" and "Framboise" are lined up. The displayed size of the content is set to be smaller as it goes away from the center, and at the same time, the relevance to the “tart” that is the content X is set to be smaller.

The user can freely scroll the screen vertically or horizontally. When the screen is scrolled, the content moves, and the content becomes larger toward the center and becomes smaller away from the center. Therefore, the user feels that the circle at the center of the screen is himself/herself and is moving in a world where contents are lined up.

FIG. 2 shows a state in which the screen is scrolled. By dragging the screen from upper right to lower left, the circle moves relatively to the upper right. In this way, when the screen is scrolled and the content is scrolled, new content is generated in the traveling direction of the circle. New content loads endlessly, so users don't end up in a dead end. New content is being generated outside the visible range that is not displayed on the screen. As a result, users can move around the world indefinitely without being aware that the content is loaded. That is, by moving, the new content in the future is acquired.

At this time, the state of the screen is managed on the server side. The server manages the state of the user's location and the next generation.

The degree of relevance between contents is calculated by using an arbitrary word/sentence vectorization method, for example, a logic called word2vec. This is a method of analyzing a large amount of text data and converting the meaning of each word into a vector expression. In this system, the bot constantly crawls the Internet to create a database of words, and continuously calculates the closeness of meaning between words. By using this, the content becomes infinitely visible. Content will not be exhausted even if a word exists and the related word is traced. However, new content is not always generated. For example, when all the content is completely generated, or when the content is too distant, the content close to the corner is adopted. It will be. However, since the content in the visible area is excluded, the content is infinitely generated on the display.

In addition, when the user searches, the word may not exist in the database, but in that case, the result can be returned to the bot and used for future database construction. Furthermore, the constructed results generate a small model in real time, which can temporarily make the user feel as if the word exists.

However, if only a model for all data is generated, it will be displayed regardless of whether it is directly related to the content selected as the center. Therefore, it may not be effective when the user wants to search the content of the content of interest. Also, since it takes a considerable amount of time to generate the whole model, real-time updating is difficult only with the whole model. Therefore, it takes time to reflect the data requested by the user.

Therefore, in the interface display method of the present invention, when a vector expression is obtained from a database, a small model that refers only to contents including some text, in addition to a large model that targets all data Generate many. Then, it is organized in a nested relation on the database. For example, a small model “animal” that refers only to the content including the text “animal” in the large model is generated and organized as a nesting relationship. This allows users to experience a space of words that expresses the content of the words they are interested in. Although the vector representations for the small model and the large model are completely different, the point is that spatial representation is essential for the user, so this can be solved.

Here, since the words are associated with each other in a multidimensional manner in the database, dimension compression is necessary to represent them on a plane. However, when the conventional dimensional compression method is used, it is important that the closeness is expressed well as a whole, in terms of whether the important content X is emphasized.
Here, the criterion of the closeness with an emphasis on the important content X is defined by the rank when the nearest neighbor search of X is executed in the model O in which the vectorized content is indexed.
Let the result be X, X2, X3, ..., Xn.
When this is dimensionally compressed, it cannot be said that X2 is closest to X on the two-dimensional map. However, there are many cases where X2 and any Xk close to it are almost close.
In this case, the origin should be set freely by the user or the system, and it should be placed in the center, so it is important that something closer to the origin is closer in distance. .. As a result, it is more important that the important contents X and X2 are represented closer than the proximity of X2 and Xk.
In the dimension compression method, dimension compression is performed only within a specific range n. Everything can be dimensionally compressed, but it cannot be continuously generated, and when moving a certain amount, a wall is created.

In order to solve this, the interface display method of the present invention adopts the following method.
First, a map based on the center is generated.
Next, let X be an important content, and let the nearest neighbor search result (n) be X2, X3, ..., Xn from the highest rank. At this time, n chooses a number that is arbitrary q^2 (where q is an odd number of 3 or more). The subsequent flow is as follows.
1. Generate a simple map as shown in Figure 3.
2. Place X at (0, 0).
3. From X2 to X9 required to fill the outside, two Xa and Xb with the lowest cos similarity are extracted and placed diagonally.
4. Excluding Xa and Xb, the two Xc and Xd with the lowest cos similarity are extracted again and placed on the other diagonal.
5. Regarding the remaining points, select 3 or 2 points that have already been fixed and calculate the sum. The nearest neighbor search is executed from the sum, and the closest Xe is selected and arranged. This completes one round.
6. In the outermost X10 to X25, similarly, the nearest neighbor search of the above 3 to 5 is executed, and this is repeated until Xn is reached. At this time, the diagonal is used only in the first lap X2 to X9, and the diagonal operation is not used from the second lap.

For continuous generation, follow the procedure below.
1. Get the user position (x, y) that exists in the map.
2. Get all locations where the map has not yet been generated within an arbitrary distance (D) from the user position.
3. At each point, select the nearest 3 points or 2 points or 1 point and calculate the sum. The nearest neighbor search is performed from the sum to an arbitrarily large R, and the closest Xf is selected and arranged. At this time, select Xf that does not fall within an arbitrary E around the point (E <R).

The above is the basic configuration of the content search interface of the present invention. In addition to this, it is also possible to arrange contents starting from a plurality of elements, as in so-called AND search. For example, as shown in FIG. 4, drag the content "Agriculture" to the area called the door on the left side of the screen, search for the word "Venture", and then select the "Venture" content as shown in FIG. By dragging to the area called the door on the left side of the screen, the colors of the two contents of “agriculture” and “venture” are inverted to become the content X, as shown in FIG. In the vicinity thereof, contents such as "farming" and "venture" which are highly related to each other are displayed.

Further, FIG. 6 shows a portal screen of the content search interface of the present invention. On this portal screen, a circle is drawn in the center, and the circle at the center indicates the user himself. The circles around it are contents that the user is interested in, and the distance between the circles is determined by the degree of interest of the user himself. In order to calculate this degree of interest, past search words and the like are taken into consideration.
Click "Notice" on the right side of the screen to display the notice to the user. Clicking on "Rireki" on the right side of the screen displays the user's browsing history and comment history. When you click "Sukinato" on the right side of the screen, the contents evaluated by the user are displayed. You can create a comment by clicking "Create" on the right side of the screen. You can search contents etc. by clicking the magnifying glass mark on the left side of the screen.

Further, FIG. 6 is also a portal screen of a social network service using the content search interface as described above. The balloon displayed above the content is the conversation within the content. In addition, by using the switch on the upper right of the screen, conversation can be performed not only with characters but also with voice through a speaker and a microphone.
As for the conversation with characters, similar to the above, the content is displayed larger as it is closer, and by displaying the conversation larger as the content is closer, the conversation regarding the content of interest is displayed larger. It is possible to operate.
Also, with regard to voice conversation, it is possible to operate such that the volume of a conversation near the volume becomes loud, that is, when a content is approached, the conversation being performed in or around the content is heard.
FIG. 7 shows a social network service using the content search interface as described above. That is, in this state, there are a plurality of users, and the users are connected via the Internet via the server. Each user can have a conversation about the content, that is, a so-called chat by clicking the content and clicking a button "Talking" displayed.

The user is able to chat with characters or voice by displaying the other party as a dot of different color while existing in the map space where the appearance is different for each user.
For example, FIG. 7 shows the map space viewed from the user A. User B exists at the position of rare cheese content on the right side of user A located in the center of the screen, and as shown in FIG. 8, user A exists at the position of tart content on the upper side when viewed from user B. To do. Although the interface looks different for each user in this way, the distance between contents does not change for each user, and the positional relationship between each user and the contents does not change for each user.

A method for realizing such a view will be described.
Basically, the user position is solved by expressing the relative position from (-1,-1) to (1,1) based on the content closest to the current position. However, in this state, for example, the user A is not displayed from the space where only "roll cake" is displayed and "tart" is not displayed. Therefore, the processing is performed as follows.

1. The relative position is calculated from (-1, -1) to (1, 1) based on the content closest to the current position. Let C be the calculated value.
2. Calculate the relative position in the same way from the 6 surrounding points. At this time, the relative position is calculated from (-1, -1) to (1, 1). However, at this time, keep only the direction and discard the distance. The value calculated here is D.
3. Send the relative position C and D including the content ID to the server.
4. The server extracts the users whose C and D contents are visible and distributes the information.
5. In case of text, text data is delivered as it is.
6. In the case of audio, the distance to the user is calculated from the content position and relative position, and the user position. Based on the distance, the audio data is three-dimensionally processed. This is done for all surrounding users, and the added data is sent to the users.
7. The user side displays or plays the data sent from the server.

In this way, according to the content search interface of the present invention, the user can search for a content that includes an unknown thing to the user by simply scrolling the screen starting from the one that the user is interested in. Can be met.
Further, according to the social network service using the content search interface of the present invention, it is possible to easily exchange information about the content that has been encountered with another person.

The present invention is not limited to the above-described embodiment, and can be carried out with various variations. For example, so-called AR (augmented reality) or VR may be used so that the user can actually walk around the map space.

Claims (1)

Step 1 in which the server creates a map of q×q=n cells (where q is an odd number greater than or equal to 3),
The server placing the arbitrarily selected content X in the center of the map, 2 ;
X2 recently top n neighbor search results from higher rank X, X3, ..., when the Xn, in X2～X 9 needed to fill one outside of the central,
The server, cos similarity lowest two combinations Xa, step 3 arranged diagonally extracts Xb,
Step 4 in which the server extracts Xc and Xd having the lowest cos similarity next to Xa and Xb and places them on the other diagonal,
Said server, for the remaining points, select the already points around 3 has been finalized or two points, the average value calculated, step 5 executes the nearest neighbor search from the average value, placing select the closest Xe ,
Said server further performs the step 3-5 also with the outside of X10～X2 5, and Step 6 repeating this until the Xn,
How the server places content on the map, including.