JP4885569B2 - Style sheet generating apparatus and style sheet generating program - Google Patents

Description

  The present invention relates to a style sheet generation apparatus and a style sheet generation program, and more particularly to a style sheet generation apparatus and a style sheet generation program for easily generating an appropriate style sheet.

  Conventional style sheet description methods for web pages and the like are, for example, two methods such as CSS (Cascading Style Sheets) such as a method for uniquely selecting a template only by tag name and attributes, and XSLT (eXtended Style Language Transformations). There are known methods that allow an algorithm to be described for template selection, such as one or more branches, repetitive processing, and the like.

In general, in order to design a style sheet, knowledge of the style sheet description method and grammar as described above is required. Therefore, in order to make it possible to easily design a style sheet without knowledge of the grammar for describing the style sheet in the past, the candidate for the form part that has been made into a part is in a format that conforms to the GUI (Graphical User Interface) to the document designer. There is a technique to present (see, for example, Patent Document 1).
JP 2003-303183 A

  However, since the conventional CSS uniquely selects a template based on a tag name and an attribute value, complicated conditional branching and random selection cannot be described. In addition, in XSLT, complicated template selection such as conditional branching of if sentence and the like and iterative processing of for sentence and the like can be performed by an algorithm, but the algorithm has to be constructed manually, and advanced techniques are required. Therefore, the production of style sheets has been very difficult.

  Moreover, in patent document 1, since the user needs to make a detailed selection from form parts prepared in advance, the work is time-consuming and time-consuming.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a style sheet generation device and a style sheet generation program for easily generating an appropriate style sheet.

  In order to solve the above problems, the present invention employs means for solving the problems having the following characteristics.

According to a first aspect of the present invention, there is provided a style sheet generating apparatus for generating a style sheet in which a content effect is described, and content description data in which the content of learning content stored in advance is described, and the content description And using a style sheet including correct answer data completed corresponding to the data, a portion of the correct answer data corresponding to a parameter included in each element of the content description data is extracted from the style sheet, and the extracted correct answer A learning means for generating at least one template using data, and identification information is assigned to each template obtained by the learning means, and the number of calls by a calling condition set corresponding to the parameter for each identification information And call for each template based on the counted result. Includes a probability table generating means for generating a probability table containing the rate, and the template, and storing means for storing said probability table, each element of the content description data of the content of the unlearned is described input The identification information of one or a plurality of templates corresponding to the calling condition is extracted from the probability table obtained by the storage unit using the calling condition of the template having at least one parameter, and corresponds to the extracted identification information Style sheet generation means for selecting a predetermined template from the templates stored in the storage means using the call probability for each template to be generated, and generating the style sheet from the content description data using the selected template. It is characterized by that.

According to the first aspect of the present invention, an appropriate style sheet can be easily generated. In addition, by using the probability table, it is possible to give variation to the effect expression. Further, the appearance can be naturally performed by controlling the appearance frequency of the template by the probability. Furthermore, the appearance frequency of the template can be set with high accuracy.

According to a second aspect of the present invention, the probability table generating means includes a tag name, an attribute value, a text value, and superficial calling condition information for each element included in the content description data as the calling condition of the template. , and among the parameters of the detailed information of the content, and having at least one.

According to the second aspect of the present invention, a detailed probability table can be generated based on each calling condition.

The invention described in claim 3 is characterized in that the probability table generating means generates at least one template corresponding to the calling condition.

According to the third aspect of the present invention, the expression can be varied by generating at least one template corresponding to the calling condition. Further, the appearance can be naturally performed by controlling the appearance frequency of the template by the probability.

The invention described in claim 4 is characterized in that the learning means replaces a text portion included in an element of the content description data with a predetermined variable when the template is generated.

According to the invention described in claim 4 , a general-purpose template can be generated.

According to a fifth aspect of the present invention, when the style sheet generation unit selects a predetermined template from the templates stored in the storage unit using the call probability for each template corresponding to the extracted identification information. A random number is generated based on the call probability, and a predetermined template is selected from the templates stored in the storage means by the random number .

According to the invention described in claim 5 , the expression can be varied. Further, the appearance can be naturally performed by controlling the appearance frequency of the template by the probability.

The invention described in claim 6, region displaying means for inputting an instruction from the user, an area for displaying the content description data of the contents for the learning, a style sheet that contains the correct answer data, and the It has a display means for displaying a user operation screen having at least one area among edit areas for generating a template based on the instruction content input by the input means.

According to the sixth aspect of the present invention, an appropriate template can be generated easily and quickly by the user.

The invention described in claim 7 is a style sheet generation program for causing a computer to function as each unit included in the style sheet generation apparatus according to any one of claims 1 to 6 .

According to the seventh aspect of the present invention, an appropriate style sheet can be easily generated. In addition, by using the probability table, it is possible to give variation to the effect expression. Further, the appearance can be naturally performed by controlling the appearance frequency of the template by the probability. Further, the appearance frequency of the template can be set with high accuracy. Furthermore, a style sheet can be easily generated by installing an execution program in a computer.

  According to the present invention, an appropriate style sheet can be easily generated.

<Outline of the present invention>
Rather than letting the algorithm be described in the description of the style sheet, the present invention compares and learns the content description data (script data) in which the content part of the content is described and the ideal conversion result (correct answer data). Build a style sheet with That is, a set of templates and a probability table indicating which template is selected at the time of the script data are generated by learning. Further, when generating a style sheet, a probability table is referred to based on the input unlearned content description data, and a template to be selected is determined probabilistically. This makes it possible to generate an appropriate style sheet more easily.

  Hereinafter, embodiments in which a style sheet generation apparatus and a style sheet generation program according to the present invention having the above-described features are preferably implemented will be described in detail with reference to the drawings.

  Here, the following embodiment demonstrates the production | generation of the stochastic template of the presentation style sheet (APE) in TV4U, and its usage. The TV 4U is an automatic program production tool capable of describing a script of a program using a word processor type user interface and producing a program by applying an arbitrary program production engine to the script.

  In this embodiment, a program using TVML (TV program Marking Language) is described as an example of content. However, the type of content in the present invention is not limited to this, for example, music or A wide variety of contents such as movies, images and sounds can be handled.

  Here, TVML is an object-based description language for producing TV programs. Video and audio of TV programs can be described by dividing them into materials and scripts. Software or the like can read it and immediately view (present) it as a television program. The present invention is not limited to TVML, and an expression format other than TVML may be used.

<Style sheet generation device: device configuration>
FIG. 1 is a diagram illustrating an example of a device configuration of a style sheet generating device according to the present invention. 1 includes an input device 11, an output device 12, a drive device 13, an auxiliary storage device 14, a memory device 15, a CPU (Central Processing Unit) 16 that performs various controls, And a network connection device 17, which are connected to each other via a system bus B.

  The input device 11 has a pointing device such as a keyboard and a mouse operated by a user, and inputs various operation signals such as execution of a program from the user. Further, the output device 12 has a display for displaying various windows and data necessary for operating the device main body for performing processing in the present invention, a speaker for outputting sound, and the like, and the program of the program is controlled by the control program of the CPU 16. Execution progress and results can be displayed and output.

  In the present invention, an execution program (style sheet generation program) installed in the apparatus main body is provided by a recording medium 18 such as a CD-ROM. The recording medium 18 on which the program is recorded can be set in the drive device 13, and the execution program included in the recording medium 18 is installed in the auxiliary storage device 14 from the recording medium 18 via the drive device 13.

  The auxiliary storage device 14 is a storage means such as a hard disk, and can store an execution program in the present invention, a control program provided in the style sheet generation device 10 and the like, and can perform input / output as necessary.

  The CPU 16 performs processing of the entire apparatus such as various operations and input / output of data with each hardware component based on a control program such as an OS (Operating System) and an execution program read and stored by the memory device 15. Each process can be realized by controlling the above. Further, the CPU 16 can acquire various kinds of information necessary during the execution of the program from the auxiliary storage device 14, and can store processing results and the like in the auxiliary storage device 14.

  The network connection device 17 obtains an execution program from another terminal connected to the communication network or executes the program by connecting to a communication network such as the Internet or a LAN (Local Area Network). Thus, the result obtained by doing this or the execution program itself in the present invention can be provided to another terminal or the like.

<Style sheet generator: functional configuration>
Next, a functional configuration example of the style sheet generating apparatus 10 will be described. FIG. 2 is a diagram illustrating an example of a functional configuration of the style sheet generating apparatus according to the present invention. 2 includes an input unit 21, an output unit 22, a storage unit 23, a learning unit 24, a probability table generation unit 25, a style sheet generation unit 26, a transmission / reception unit 27, And a control means 28.

  The input unit 21 receives input of various processes such as a learning execution instruction, a probability table generation instruction, template selection, and conversion from a user or the like. Further, the output means 22 displays the contents of the instruction input by the input means 21 and the contents such as a probability table and a template generated based on the instruction contents.

  The accumulating unit 23 accumulates various information necessary for the present embodiment. Specifically, for example, learning script data 31, correct answer data 32, template 33, probability table 34, unlearned script data 35, presentation style sheet 36, and the like are stored in the storage means 23.

  The learning unit 24 inputs learning script data 31 as content description data stored in the storage unit 23 and correct answer data (TVML script) 32 as an ideal conversion result style sheet to perform learning. Generate a template. Details of the template generation method in the learning unit 24 will be described later. The learning unit 24 stores the learned template in the storage unit 23.

  In addition, the probability table generating unit 25 gives identification information to the template group learned by the learning unit 24. At this time, the duplicated template is deleted. In addition, the number of times the template to which the identification information is added is called based on a predetermined calling condition is counted, and the probability table 34 is generated based on the counted result. Details of the table generation method in the probability table generation means 25 will be described later. In addition, the probability table generating unit 25 stores the template 33 to which the identification information is added and the generated probability table 34 in the storage unit 23.

  The style sheet generation unit 26 is based on the template 33 obtained by the learning unit 24 and the probability table 34 generated by the probability table generation unit 25, and is an unlearned script as input unlearned content description data. A template corresponding to the template calling condition set from the data 35 is selected, and the unlearned script data 35 is converted into a style sheet format using the selected template. Details of template selection and conversion in the style sheet generation means 26 will be described later. The style sheet generating unit 26 stores the effect style sheet 36 generated by the selection and conversion described above in the storage unit 23.

  Further, the transmission / reception means 27 can acquire data necessary for generating a template from an external device connected via a communication network or the like. Further, the transmission / reception means 27 can transmit the produced effect style sheet 36 to another terminal or the like via a communication network.

  The control means 28 controls the entire components of the style sheet generating apparatus 10 according to the contents of instructions from the input means 21 such as a user. Specifically, the control unit 28 causes the learning unit 24 to perform template learning processing, the probability table generation unit 25 to generate a probability table, and the style sheet generation unit 26 to select and convert a template. Or the transmission / reception means 27 controls the transmission / reception of various data.

<Learning means 24>
Here, the learning means 24 described above will be described. The learning unit 24 generates a template by learning using a pair of learning content description data (learning script data 31) and an ideal conversion result (correct answer data 32). Specifically, the correct answer is input for each element (row) of the content description data, which part of the correct data corresponds to the content description data. Since each element of the content description data includes a tag name, an attribute value, a text value, and the like as parameters, information about what template corresponds to the combination is accumulated by learning.

  In the learning method, an operation screen for accumulating learning data is displayed on the output unit 22 of the style sheet generating apparatus 10, and a user or the like performs a predetermined operation using the input unit 21. A suitable template can be generated quickly.

  Here, an example of a template learning screen according to the present invention described above will be described with reference to the drawings. FIG. 3 is a diagram illustrating an example of an operation screen for accumulating learning data. The operation screen 40 shown in FIG. 3 includes a menu display area 41, a correct data display area 42, a content description data display area 43, and an editing area 44.

  The menu display area 41 inputs file format learning script data 31, correct answer data 32, and the like, and performs processing such as termination of screen display. The correct answer data display area 42 displays a TVML script which is correct answer data produced in advance. That is, the producer or the like edits in advance, and the completed TVML script is displayed as correct data. In the correct data display area 42, the sentence being edited in the editing area 44 can be highlighted (for example, the highlighted area 45 in FIG. 3) in bold or color. Thereby, a user etc. can be made to grasp | ascertain the present edit location clearly.

  In addition, the content description data display area 43 includes an event type (character) based on information included in the learning script data 31, for example, an action or a text by a CG (Computer Graphic) character appearing in the TVML script. Etc.) for each command (talk, etc.), the tag name (input) of each parameter, the attribute value (name) of a CG character or the like, and the text value (value), based on the preset description contents Are extracted and displayed based on the extraction conditions.

  In addition, the content description data display area 43 can highlight a portion corresponding to the sentence being edited in the editing area 44 (for example, the highlighted display area 46 in FIG. 3). This makes it possible for the user or the like to clearly understand the text value corresponding to the current editing location.

  The editing area 44 selects the description contents (elements) to be edited from the correct TVML script displayed in the correct data display area 42 (Pickup), and the selected element has variables such as lines and actions. For example, it is converted into a value (VALUE) indicating a variable value. The converted text is replaced (repressed) with the corresponding position of the correct TVML script displayed in the correct data display area 42. This makes it possible to generate a general-purpose template that can handle any type of script data.

  The edit area 44 has various button groups (such as a description description selection (Pick up) button 47, a VALUE conversion (This is Value) button 48, a replacement (Register) button 49, etc.) in order to quickly perform the above-described processes. May be provided.

  As described above, the user or the like can easily and quickly generate an appropriate template by using the screen described above. Note that the screen layout of each area shown in FIG. 3 is not limited to this in the present invention. For example, each area may be displayed using a plurality of screens.

  Moreover, the learning of the template in the learning means 24 described above may be performed based on the conversion rule by setting the editing contents for the command in advance.

<Probability table generation means 25>
Next, a probability table generation method in the probability table generation means 25 will be described. The probability table generation unit 25 gives an ID as identification information to the template group learned by the learning unit 24 and accumulated in the accumulation unit 23. In addition, the probability table generation unit 25 deletes a template having overlapping contents when assigning an ID. Next, the probability table generating means 25 counts which template is called for each calling condition based on correct answer data, and the number of times for each template ID.

  In addition, as a condition (parameter) for calling a template, in addition to the tag name, attribute value, and text value of each element of the content description data, which template is before the template call at that time (1 to N times) Surface parameters such as whether it was called (its template ID), which camera was selected at that time (camera name used for shooting), the name of the character (character name) hit immediately before, etc. Detailed information about the contents of the program can also be used as a parameter.

  Therefore, the probability table generation means 25 can generate a template group with IDs assigned and a probability table representing how many times the template has been called under various conditions by the above-described processing.

<Style sheet generation means 26>
Next, the contents of template selection / conversion in the style sheet generating means 26 will be described. The style sheet generation unit 26 uses the template 33 obtained by the learning unit 24 and the probability table 34 obtained by the probability table generation unit 25 as input to the script data 35 that is unlearned content description data. , Convert to stylesheet.

  Specifically, the style sheet generation means 26 compares the condition of the input content description data with the probability table 34, and if there is a condition in which all the parameters match, which template and how many under that condition. It can be obtained whether it is selected with probability. Therefore, a random number is generated based on the probability, a template is selected based on the random number, and style sheet generation processing is performed based on the selected template.

  Note that the style sheet generation means 26 does not select the template with the highest probability, but selects a template with a random number. This is because in the production of a television program or the like, when the same conditions are used, it is more natural to produce a variation in the representation and to control the frequency according to the probability than to perform the same representation.

  Note that the priority order may be set according to the counted value so that the template having a high probability is easily selected as compared with the other templates.

  Here, when the above-described method is used, it is only necessary that the conditions in which all the parameters match are in the learning data, but there are cases in which the conditions in which all the parameters match do not exist in the learned data. In particular, when there is little learning data or there are many kinds of parameters, the above-mentioned thing can occur frequently.

  Therefore, in such a case, even if there is no template that matches the condition, the following means are used to select the next best template that is similar to the condition.

  For example, when t parameters are taken as conditions, the parameters are Pt,..., P1, P0, respectively. At this time, the larger the subscript number, the higher the importance. (Importance: Pt>...> P1> P0). Next, a t-digit binary number S is prepared and the maximum value is input (all digits are set to 1). Here, in the condition search, first, the search is performed under the condition that all parameters match as usual. As a result of the search, if there is no corresponding template, the t-digit binary number S is decremented by -1 (decrement), and a parameter as a calling condition is set by a numerical value for every n digits of S.

  That is, when the nth (1 <n <t) digit of the binary number S is 1, the parameter Pn corresponding to that digit is included in the search, and when the nth digit is 0, Pn is the condition. Search without including. By repeating the above processing until the parameters and the conditions of the template match, it is possible to select a template under similar conditions even if there is no condition that completely matches the first calling condition.

<Example>
Next, processing contents in the present invention using TVML will be described using a flowchart and the like.

<Learning processing procedure and probability table generation procedure>
FIG. 4 is a flowchart showing an example of a learning process procedure and a probability table generation procedure in the present invention.

  First, a correct TVML script is input as learning script data and correct data (S01). Here, FIG. 5 is a diagram illustrating an example of the script data for learning. FIG. 6 is a diagram illustrating an example of a correct TVML script. Here, row numbers are given to the left end of FIGS. 5 and 6 for convenience of explanation.

  Next, a template corresponding to the input script data for learning is generated. Specifically, it is input which part of the correct TVML script shown in FIG. 6 corresponds to each element (row) of the script data for learning shown in FIG. Here, FIG. 7 is a diagram illustrating an example of the learned template. In FIG. 7, a row number is given to the left end for convenience.

  Here, for example, the element “<action name =“ A ”> camera line </ action>” in the (09) line in FIG. 5 is “character: look (name = name = number)” in the (11) line in FIG. A, what = camera, wait = no) ”.

  Therefore, when the line is input as learning data, as shown in lines (20) to (22) in FIG. 7, “<template input =“ action ”name =“ A ”value =“ camera line of sight ””> character: look (name = A, what = camera, wait = no) </ template> ”.

  In addition, FIG. 5 of the (11) th row "<text name =" A "> Hello, love </ text>!" Is the element that, in FIG. 6 (13) th row of the "character: talk (name = A, text = "Hello, love!", corresponds to the pitch = "1") ".

  Therefore, if you enter the line as data for learning, speech part is a text value "Hello, love!" Is, because it is possible substitution of templates, it demonstrates on the learning data that.

  Specifically, as shown in the (26) to (28) lines of FIG. 7, “<template input =“ text ”name =“ A ”> character: talk (name = A, text =” <VALUE / > ", Pitch =" 1 ") </ template>". Here, “<VALUE />” indicates a portion that can be replaced with the text value of the element of the script data as described above.

  The above processing is repeated to accumulate learning data as shown in FIG. 7, and when the accumulation is completed, an ID as identification information is given to the template group (S03). At this time, the duplicate template is deleted. Thereby, a template with an ID is generated.

  Here, FIG. 8 is a diagram illustrating an example of a template to which an ID is assigned. In FIG. 8, a row number is given to the left end for convenience. As shown in FIG. 8, (02), (06), (10), etc. are assigned tags having IDs for each template as “<TEMPLETE id =“ ”>”. Note that the description content of the tag attached to the template is not limited to this.

  Next, a probability table corresponding to each preset condition is generated from the template assigned the ID (S05). Specifically, for example, when learning is performed under a combination of calling conditions (parameters), it is possible to count how many templates are called and generate a probability table based on the count result. . Here, FIG. 9 is a diagram illustrating an example of the probability table. In FIG. 9, a line number is given to the left end for convenience.

  In the probability table shown in FIG. 9, there are four parameters: a script data tag name (input), an attribute value (name), a text value (value), and the last template called (before). Is used.

  Here, there are two types of pattern templates in the templates in the (41) to (44) lines shown in FIG.

  This is because there are two types of pattern templates (<pattern template_ref = “18” frequency = “2” />) for the template “<template_finder input =“ action ”value =“ hide picture ”before =“ 7 ”>”. , <Pattern template_ref = “12” frequency = “1” />). That is, under this condition, the template with ID = 18 is selected with a probability of twice ((42) line) and the template with ID = 12 is selected with a probability of once ((43) line).

  Thus, by generating not only one template but also a plurality of pattern templates for one template, it is possible to vary the expression. Further, the production can be naturally performed by controlling the frequency according to the probability. The probability table generated by the above-described procedure is stored in the storage means 23 for use in the template selection / conversion procedure.

<Template generation processing procedure>
Next, a template generation processing procedure for selecting a template from unlearned script data using the template generated by the above-described processing procedure and the probability table and performing conversion will be described using a flowchart. FIG. 10 is a flowchart showing an example of a template generation processing procedure in the present invention.

  In FIG. 10, first, a template group generated by the above-described processing procedure and a probability table are input (S11). Next, script data as unlearned content description data is input (S12). Note that unlearned script data means script data that is not used in learning in the above-described learning processing procedure.

  Here, parameters P0 to Pt for t digits and a binary number S are prepared (S13). For example, when there are four parameters (P3 = input, P2 = name, P1 = value, P0 = before), the number of digits of the binary number S is also four digits. Next, the maximum value is set in S, that is, 1 is set in all the digits (S14). In the case of the above-described example, S is 4 digits and thus becomes “1111”.

  Next, each element (row) of the script data input in S11 is read and substituted into the parameter P (S15), and the probability table is searched based on the preset condition (parameter P) to obtain the template. Call (S16).

  Here, it is determined whether or not there is a matching template (S17). If there is no matching template (NO in S17), the current value of S is set to -1 in order to easily and appropriately change the search condition ( S18). That is, when the current S is “1111”, S = 1111−1 = 1110.

Next, the search condition is reset based on the value of each digit of S (S19). In particular,
It is determined whether or not the nth digit of S is 0. If the nth digit is 0, Pn is removed from the search condition. That is, when S is 1110, the first digit P0 = before is excluded from the search conditions. In addition, after resetting the search conditions, the process returns to S16 to re-search the probability table.

  In the process of S17, if there is a matching template (YES in S17), a random number is generated based on the probability of the hit probability table, and a template is selected based on the random number (S20). Based on the script data input in S12, it is converted into an effect style sheet and output (S21).

  For example, suppose that P3 (input) = action, P2 (name) = null (empty data), P1 (value) = “hide picture”, and P0 (before) = 7. In this case, the condition matches the (41) th row of the probability table shown in FIG.

  Under this condition, the template with ID = 18 is selected with a probability of twice ((42) line) and the template with ID = 12 is selected with a probability of once ((43) line). Accordingly, the probability that the template with ID = 18 is selected is 2/3, the probability that the template with ID = 12 is selected is 1/3, a random number is generated, and one of the templates is selected based on the random number.

  Here, when ID = 12 is selected, (49) to (52) are selected from the templates to which the ID of FIG. 8 described above is assigned, and as a result, “video: switcher (source = studio) prop : Visible (name = SUBIMAGE, switch = off) ".

  As another example, for example, the search conditions are P3 (input) = subimage, P2 (name) = null (empty data), P1 (value) = “photo.jpg”, P0 (before) = 11, and P3- Assume that there is no condition in the probability table that satisfies all the parameters of P0. In that case, as described above, S is decremented and the search condition is reset (for example, if S = 1110, the first digit is 0, so P0 is removed from the condition), and the search is performed again. .

  In this condition, when S = 1100, the condition matches the (29) th row of the probability table shown in FIG. Therefore, the (32) to (39) lines in the template shown in FIG. 8 are selected, and the <VALUE /> part is replaced. As a result, “video: switcher (source = studio) prop: assign (Name = SUBIMAGE) prop: openimageplate (name = SUBIMAGE, filename = “photo.jpg”, platesize = 0.9, platesizev = 0.6) propire: visible (name = SUBimage = name) = SUBIMAGE, x = 0.01, y = 1.41, z = -2.40, pitch = 0.00, yaw = 0.00, roll = 0.00, scale = 1 45) camera: it is possible to obtain the results shown in switch (name = CamAB).

  Next, it is determined whether or not the element (row) that has been subjected to the above processing is the last row of the script data (S22). If it is not the last row (NO in S22), the process returns to S14, and S is the maximum value. Then, the process after reading the next element of the script data is continued in S15.

  In the process of S22, when the process reaches the last line (NO in S22), the template generation process ends.

  By performing the above processing for each element of the script data, an appropriate style sheet can be easily generated. Further, by using a binary number for setting the search condition, the search condition can be appropriately set without overlapping the previous search condition. Furthermore, by repeating the processing as described above, a template can be selected under similar conditions even when there is no completely matching condition.

  In the example shown in FIG. 10 described above, the value is decreased after setting S as the maximum value. However, the present invention is not limited to this. For example, the minimum value is set to S and the search condition is set. Sometimes, the value of S may be increased by 1 (incremented). In this case, when the number of digits is 0, the search condition is entered, and when the number of digits is 1, the search condition is removed.

<Program>
Here, the present invention generates an execution program (style sheet generation program) that can cause a computer to execute the processing in each configuration of the style sheet generation apparatus 10 described above. For example, the program is stored in a general-purpose personal computer or server. By installing, the style sheet generation processing in the present invention can be realized. Therefore, an appropriate style sheet can be easily generated. Moreover, a style sheet can be easily generated by installing an execution program in a computer.

  As described above, according to the present invention, an appropriate style sheet can be easily generated. In addition, by using the probability table, it is possible to give variation to the effect expression. Further, the appearance can be naturally performed by controlling the appearance frequency of the template by the probability.

  Specifically, the style sheet is constructed by comparing and learning the content part (script data) and the ideal conversion result (correct answer data), instead of describing the algorithm in the description of the style sheet. Further, a probability table indicating which template is selected when learning is obtained by collecting a set of templates and what kind of script data is generated in advance. In the conversion, the template to be selected is stochastically determined with reference to the probability table from the input script data. This makes it possible to produce a style sheet that performs simple and efficient complex processing.

  Therefore, if there is sufficient learning data for production of the production style sheet, it is possible to produce a natural production easily without describing a complicated algorithm. Note that the present invention can be widely applied in the field where television programs are produced and described in a script language and reproduced using computer graphics (CG), speech synthesis, or the like. Further, the present invention can be applied to the production editing of a style sheet in which only the effect portion is separated in order to divide the script describing the flow of the television program into the effect portion and the content portion.

  Although the preferred embodiment of the present invention has been described in detail above, the present invention is not limited to the specific embodiment, and various modifications, within the scope of the gist of the present invention described in the claims, It can be changed.

It is a figure which shows an example of the apparatus structure of the style sheet production | generation apparatus in this invention. It is a figure which shows an example of a function structure of the style sheet production | generation apparatus in this invention. It is a figure which shows an example of the operation screen for accumulating learning data. It is a flowchart which shows an example of the learning process procedure in this invention, and a probability table production | generation procedure. It is a figure which shows an example of the script data for learning. It is a figure which shows an example of a correct TVML script. It is a figure which shows an example of the learned template. It is a figure which shows an example of the template to which ID was provided. It is a figure which shows an example of a probability table. It is a flowchart which shows an example of the template production | generation process procedure in this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Style sheet production | generation apparatus 11 Input device 12 Output device 13 Drive apparatus 14 Auxiliary storage device 15 Memory device 16 CPU
DESCRIPTION OF SYMBOLS 17 Network connection apparatus 18 Recording medium 21 Input means 22 Output means 23 Storage means 24 Learning means 25 Probability table generation means 26 Style sheet generation means 27 Transmission / reception means 28 Control means 31 Script data for learning 32 Correct data 33 Template 34 Probability table 35 Not yet Learning script data 36 Production style sheet 40 Operation screen 41 Menu display area 42 Correct data display area 43 Content description data display area 44 Edit area 45, 46 Highlight display area 47 Description content selection button 48 VALUE conversion button 49 Replace button

Claims (7)

In a style sheet generating device that generates a style sheet in which content effects are described,
Included in each element of the content description data using content description data describing the content of the learning content stored in advance and a style sheet including correct answer data completed corresponding to the content description data Learning means for extracting a part of the correct answer data corresponding to a parameter to be generated from the style sheet and generating at least one template using the extracted correct answer data;
Identification information is given to each template obtained by the learning means, and further, the number of calls by the calling condition set corresponding to the parameter for each identification information is counted, and based on the counted result, Probability table generation means for generating a probability table including a call probability;
It means for storing said templates, and the probability table,
The calling condition from the probability table obtained by the storage means using the calling condition of the template having at least one parameter included in each element of the content description data describing the contents of the input unlearned content The identification information of one or a plurality of templates corresponding to is extracted, a predetermined template is selected from the templates stored in the storage means using the call probability for each template corresponding to the extracted identification information, and the selected template And a style sheet generating unit that generates the style sheet from the content description data using the content description data. The probability table generating means includes:
As a calling condition for the template, at least one of a tag name, an attribute value, a text value, superficial calling condition information for each element included in the content description data, and content detailed information parameters is included. The style sheet generating apparatus according to claim 1 . The probability table generating means includes:
Style sheet generating device according to claim 1 or 2, wherein the generating at least one template in association with said call condition. The learning means includes
4. The style sheet generating apparatus according to claim 1 , wherein a text portion included in the content description data element is replaced with a predetermined variable when the template is generated. The style sheet generating means includes
Using a call probability for each template corresponding to the extracted identification information, when a predetermined template is selected from the templates stored in the storage means, a random number is generated based on the call probability, and the storage is performed using the random number. style sheet generating device according to any one of claims 1 to 4, characterized in that selecting a predetermined template from the stored template means. An input means for inputting an instruction from the user;
Of the area for displaying the content description data of the learning content, the area for displaying the style sheet including the correct answer data , and the editing area for generating a template based on the instruction content input by the input unit, style sheet generating device according to any one of claims 1 to 5, characterized in that it has a display means for displaying a user operation screen having at least one region. A style sheet generation program for causing a computer to function as each unit included in the style sheet generation apparatus according to claim 1 .

Images