JP6995698B2 - Information processing equipment, information processing methods, and information processing programs - Google Patents

Description

The present invention relates to an information processing apparatus, an information processing method, and an information processing program.

Conventionally, various techniques related to character string conversion such as translation have been provided. For example, there is known a technique for realizing more accurate language identification by using the series information of language-likeness at the frame level. In such a technique, a plurality of learning data as a set of a character string before conversion and a character string after change are used. Further, in the prior art, for example, a plurality of learning data in which voice data in a plurality of languages and language labels representing the languages of the voice data are combined are used.

Japanese Unexamined Patent Publication No. 2016-156870

However, with the above-mentioned prior art, it may be difficult to appropriately provide a service related to character information conversion. For example, in such character information conversion, if the combination of the character information before conversion and the character information after conversion (teacher data) is insufficient, it is difficult to appropriately provide the service related to the character information conversion. There is a problem such as.

The present application has been made in view of the above, and an object of the present application is to provide an information processing apparatus, an information processing method, and an information processing program that can appropriately provide services related to character information conversion.

The information processing apparatus according to the present application includes an acquisition unit for acquiring first character information which is character information of the first type and second character information which is character information of the second type different from the first type. Based on learning that the first character information is input information and the first character information is correct information, and learning that the second character information is input information and the second character information is correct information. It is characterized by including a learning unit that generates a conversion model that outputs the converted character information obtained by converting the target character information into the second type in response to the input of the target character information of the first type. And.

According to one aspect of the embodiment, there is an effect that a service related to character information conversion can be appropriately provided.

FIG. 1 is a diagram showing an example of information processing according to an embodiment. FIG. 2 is a diagram showing an example of information processing according to an embodiment. FIG. 3 is a diagram showing a configuration example of an information processing system according to an embodiment. FIG. 4 is a diagram showing a configuration example of the information processing apparatus according to the embodiment. FIG. 5 is a diagram showing an example of the first type information storage unit according to the embodiment. FIG. 6 is a diagram showing an example of the second type information storage unit according to the embodiment. FIG. 7 is a diagram showing an example of a model information storage unit according to an embodiment. FIG. 8 is a flowchart showing an example of information processing according to the embodiment. FIG. 9 is a flowchart showing an example of information processing according to the embodiment. FIG. 10 is a diagram showing an example of information processing according to a modified example. FIG. 11 is a diagram showing an example of information processing according to a modified example. FIG. 12 is a diagram showing a configuration example of the information processing apparatus according to the modified example. FIG. 13 is a diagram showing an example of a model information storage unit according to a modified example. FIG. 14 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing device.

Hereinafter, the information processing apparatus, the information processing method, and the embodiment for implementing the information processing program (hereinafter referred to as “the embodiment”) according to the present application will be described in detail with reference to the drawings. It should be noted that this embodiment does not limit the information processing apparatus, information processing method, and information processing program according to the present application. Further, in each of the following embodiments, the same parts are designated by the same reference numerals, and duplicate explanations are omitted.

(Embodiment)
[1. Information processing (learning process)]
First, an example of information processing according to an embodiment will be described with reference to FIG. FIG. 1 is a diagram showing an example of information processing according to an embodiment. Specifically, FIG. 1 describes an example of learning processing among the information processing according to the embodiment. In FIG. 1, the information processing apparatus 100 refers to character information of the first type (hereinafter, also referred to as “first character information”) and character information of a second type different from the first type (hereinafter, also referred to as “second character information”). ) And the case of generating a model is shown. It should be noted that the first type and the second type referred to here are the types in the same language. Further, in FIG. 1, the first type is a news article and the second type is an article for children as an example, but the first type and the second type are different types in the same language. If there is, any type may be used. In FIG. 1, a model in which the information processing apparatus 100 converts a type of character information based on a learning process using the first character information and a learning process using the second character information (hereinafter, also referred to as a “conversion model”). ) Is generated. That is, in FIG. 1, a conversion model in which the information processing apparatus 100 outputs character information (hereinafter also referred to as “post-conversion character information”) obtained by converting the type of input character information (hereinafter also referred to as “character information before conversion”). The case of generating is shown.

Further, the information processing apparatus 100 uses the generated conversion model to generate the converted character information obtained by converting the type of the pre-conversion character information when a certain pre-conversion character information is acquired. Further, the information processing apparatus 100 provides a service based on the generated converted character information, and this point will be described with reference to FIG.

[Information processing system configuration]
First, prior to the description of FIG. 1, the information processing system 1 shown in FIG. 3 will be described. As shown in FIG. 3, the information processing system 1 includes a terminal device 10, an information providing device 50, and an information processing device 100. The terminal device 10, the information providing device 50, and the information processing device 100 are connected to each other via a predetermined network N so as to be communicable by wire or wirelessly. FIG. 3 is a diagram showing a configuration example of an information processing system according to an embodiment. The information processing system 1 shown in FIG. 3 may include a plurality of terminal devices 10, a plurality of information providing devices 50, and a plurality of information processing devices 100.

The terminal device 10 is an information processing device used by the user. The terminal device 10 is realized by, for example, a smartphone, a tablet terminal, a notebook PC (Personal Computer), a desktop PC, a mobile phone, a PDA (Personal Digital Assistant), or the like. In the following, the terminal device 10 may be referred to as a user. That is, in the following, the user can be read as the terminal device 10.

The terminal device 10 accepts operations by the user. Further, the terminal device 10 transmits various information to the information processing device 100 and the information providing device 50 according to the operation of the user. For example, the terminal device 10 transmits information or the like requesting content such as character information to the information processing device 100 or the information providing device 50. Further, the terminal device 10 receives various information from the information processing device 100 and the information providing device 50. For example, the terminal device 10 receives the converted character information from the information processing device 100 and the information providing device 50. For example, the terminal device 10 displays character information received from the information processing device 100 and the information providing device 50.

The information providing device 50 is an information processing device used by a provider that provides a predetermined service. The provider may be an individual or a corporation or the like. For example, the information providing device 50 is used by the information provider CP1. The information provider CP1 may be a content provider such as a content provider. The information provider CP1 may be a business operator that provides a content distribution service. The information provider CP1 may be a business operator that provides an advertisement distribution service.

For example, the information provider CP1 uses the information providing device 50 to transmit the character information provided in the content distribution service to the information processing device 100. The information provider CP1 uses the information providing device 50 to transmit the converted character information to the information processing device 100 in order to generate the character information. Further, the information providing device 50 is realized by, for example, a smartphone, a tablet terminal, a notebook PC, a desktop PC, a mobile phone, a PDA, or the like. In the following, the information providing device 50 may be referred to as a provider. That is, in the following, the provider can be read as the information providing device 50. Specifically, FIGS. 1 and 2 show a case where the information providing device 50 is a notebook PC used by the information provider CP1. The information providing device 50 may be integrated with the information processing device 100. That is, the information processing apparatus 100 may provide a content distribution service, an advertisement distribution service, or the like.

The information processing device 100 is an information processing device that generates a conversion model that outputs converted character information in which the target character information is converted into the second type in response to input of the target character information of the first type. The information processing device 100 is an information processing device that generates a conversion model that outputs other converted character information obtained by converting other target character information into the first type in response to input of other target character information of the second type. Is. For example, the information processing device 100 is a learning device that performs learning processing.

The information processing device 100 is an information processing device that generates converted character information in which the type of the character information is converted by inputting the character information into the conversion model. For example, the information processing device 100 is a generation device that performs a character information generation process. For example, the information processing device 100 is a conversion device that performs conversion processing of character information.

Further, the information processing apparatus 100 provides a service based on the generated character information after conversion. In the example of FIG. 2, the information processing apparatus 100 provides the generated converted character information to the information providing apparatus 50. The information processing apparatus 100 may provide information to the terminal apparatus 10 used by the user by using the generated converted character information. The information processing device 100 may distribute the generated content including the converted character information to the terminal device 10. In this case, the information processing apparatus 100 acquires various information from an external information processing apparatus or the like that provides a content distribution service, and performs information processing or information provision to an external information processing apparatus based on the acquired various information. May be good.

First, the information processing apparatus 100 acquires the first character information as the learning data LD1 (step S11). Then, the information processing apparatus 100 adds the first character information included in the learning data LD1 acquired in step S11 as learning data (step S12). Specifically, the information processing apparatus 100 adds the learning data LD1 to the first type information storage unit 121.

The first type information storage unit 121 shown in FIG. 1 includes items such as "character information ID" and "first character information".

The "character information ID" in the first type information storage unit 121 of FIG. 1 indicates identification information for identifying character information. The "first character information" in the first type information storage unit 121 of FIG. 1 indicates character information identified by the character information ID.

In the example shown in FIG. 1, the character information identified by the character information ID "FID11" indicates that it is the first character information FCH11, and the character information identified by the character information ID "FID12" is the first character information. Indicates that it is FCH12. Further, the character information identified by the character information ID "FID13" indicates that it is the first character information FCH13, and the character information identified by the character information ID "FID14" is the first character information FCH14. show. The first character information FCH11 to FCH14, etc. is any character information as long as it is general news not for children (hereinafter, also referred to as "normal news") or character information distributed by ordinary content distribution. May be. For example, the first character information FCH11 to FCH14 and the like may be ordinary news about politics and economy, ordinary news about sports and entertainment, and the like.

Further, the information processing apparatus 100 acquires the second character information as the learning data LD2 (step S13). Then, the information processing apparatus 100 adds the second character information included in the learning data LD2 acquired in step S13 as learning data (step S14). Specifically, the information processing apparatus 100 adds the learning data LD2 to the second type information storage unit 122.

The second type information storage unit 122 shown in FIG. 1 includes items such as "character information ID" and "second character information".

The "character information ID" in the second type information storage unit 122 of FIG. 1 indicates identification information for identifying character information. The "second character information" in the second type information storage unit 122 of FIG. 1 indicates character information identified by the character information ID.

In the example shown in FIG. 1, the character information identified by the character information ID "SID11" indicates that it is the second character information SCH11, and the character information identified by the character information ID "SID12" is the second character information. Indicates that it is SCH12. Further, the character information identified by the character information ID "SID13" indicates that it is the second character information SCH13, and the character information identified by the character information ID "SID14" is the second character information SCH14. show. The second character information SCH11 to SCH14 and the like may be any character information as long as it is character information such as news created for children (hereinafter, also referred to as "news for children"). For example, the second character information SCH11 to SCH14 and the like may be children's news about politics and economy created for children, children's news about sports and entertainment, and the like.

Then, the information processing apparatus 100 generates a model by learning processing based on the learning data. For example, the information processing apparatus 100 learns data (teacher data) such as first character information FCH11 to FCH14 in the first type information storage unit 121 and second character information SCH11 to SCH14 in the second type information storage unit 122. As a result, a transformation model is generated by training. It is assumed that there is no correspondence between the first character information FCH11 to FCH14 and the like and the second character information SCH11 to SCH14 and the like in the second type information storage unit 122. For example, the first character information FCH11 to FCH14 and the like and the second character information SCH11 to SCH14 and the like in the second type information storage unit 122 are assumed to be news with different contents.

First, a conversion model (learner) generated by the information processing apparatus 100 will be briefly described. In the example of FIG. 1, the model M1 in the model information storage unit 123 is a model of the architecture as shown in the structural information ARC1. In the example of FIG. 1, the model M1 which is a conversion model generated by the information processing apparatus 100 is a model including a plurality of first partial models M1-1, M1-2 and the like. As described above, when "model M * (* is an arbitrary numerical value)" is described, it means that the model is a model identified by the model ID "M *". For example, when described as "model M1", the model is a model identified by the model ID "M1".

Further, as described above, when the description is described as "first part model M * (* is an arbitrary numerical value)", the first part model is a model identified by the first part model ID "M *". Is shown. For example, when described as "first partial model M1-1", the first partial model is a model identified by the first partial model ID "M1-1".

Further, as described above, when the description is "second part model M * (* is an arbitrary numerical value)", the second part model is a model identified by the second part model ID "M *". Is shown. For example, when described as "second part model M11", the second part model is a model identified by the second part model ID "M11". When the first partial model and the second partial model are described without particular distinction, a "partial model" may be simply described.

The first partial model M1-1 is a coding unit, and includes a encoder (hereinafter, also referred to as “encoder”) that extracts the features of the input information by compressing the dimensional amount of the input information. Corresponds to the partial model. As shown in FIG. 1, the first partial model M1-1 corresponds to a model in which the input information IN1 is input and the intermediate information INT1 is output. The first partial model M1-1 corresponds to a model in which first character information and second character information are input as input information IN1 and intermediate information INT1 is output.

The first partial model M1-1 includes a second partial model M11 which is an encoder for inputting the first character information. The second partial model M11 extracts the features of the input first character information by compressing the dimension amount of the input first character information. The second partial model M11 outputs, as the intermediate information INT1 in FIG. 1, intermediate information indicating the characteristics of the first character information in which the dimension amount of the first character information is compressed. For example, the intermediate information INT1 in FIG. 1 is an intermediate output (encoded vector or the like) showing the characteristics of the input information, and may be, for example, a distributed representation or a real value vector. For example, the intermediate information INT1 includes information regarding the meaning (concept) of the character information input to the model M1.

Further, the first partial model M1-1 includes a second partial model M12 which is an encoder for inputting the second character information. The second partial model M12 extracts the features of the input second character information by compressing the dimension amount of the input second character information. The second partial model M12 outputs, as the intermediate information INT1 in FIG. 1, intermediate information indicating the characteristics of the second character information in which the dimension amount of the second character information is compressed.

For example, the intermediate information INT1 is set with specifications such as the number of dimensions, and the second partial models M11 and M12 in the first partial model M1-1 output the intermediate information INT1 based on the common specifications. The second partial models M11 and M12 output intermediate information compressed to a common number of dimensions. In the example of FIG. 1, the second partial models M11 and M12 are shown individually for the sake of explanation, but the second partial models M11 and M12 may share a common part. For example, the second partial models M11 and M12 may be integrated. In other words, the second partial models M11 and M12 may be one model. For example, the first partial model M1-1 may be configured by removing overlapping portions from the configurations for realizing the functions of the second partial models M11 and M12.

Further, the first partial model M1-1 has a function of determining which of the second partial models M11 and M12 to input the input character information according to the type of the input character information. For example, the first partial model M1-1 accepts information (specific information) that specifies the type of character information to be input, and inputs character information to the second partial models M11 and M12 that correspond to the specific information. Then, it has a function to determine. For example, in the first partial model M1-1, character information is input to the input layer of the model based on the specific information among the second partial models M11 and M12 by a predetermined switching function (switching function) (logic). ) The circuit etc. may be switched. For example, in the first partial model M1-1, at the time of learning, the selection information (control vector) described later may be used as specific information to determine a model for inputting character information among the second partial models M11 and M12. ..

For example, the first partial model M1-1 is a model in which character information is input from the second partial models M11 and M12 by using the information obtained by inverting the selection information described later as specific information at the time of generation (conversion). May be determined. For example, in the first partial model M1-1, when the selection information is the information indicating the "first type" at the time of generation (conversion), the information indicating the "second type" obtained by reversing the information is specified. Of the second partial models M11 and M12, which may be used as information, a model for inputting character information may be determined. For example, in the first partial model M1-1, when the selection information is the information indicating the "second type" at the time of generation (conversion), the information indicating the "first type" obtained by reversing the information is specified. Of the second partial models M11 and M12, which may be used as information, a model for inputting character information may be determined. The above is an example, and the information processing apparatus 100 may realize the determination of the model to be input by any function as long as the input to the desired model is possible.

The first partial model M1-2 is a decoding unit, and includes a decoder (hereinafter, also referred to as “decoder”) that increases the dimensional amount of the feature extracted by the encoder and outputs character information of a predetermined type. Corresponds to a partial model. That is, the first partial model M1-2 is a partial model including a decoder that inputs intermediate information output by an encoder, increases the dimension amount of the input intermediate information, and outputs character information of a predetermined type. handle. As shown in FIG. 1, the first partial model M1-2 corresponds to a model in which the intermediate information INT1 is input and the output information OUT1 is output.

The first partial model M1-2 includes a second partial model M13 which is a decoder for inputting intermediate information. The second partial model M13 generates the first character information corresponding to the input intermediate information by increasing the dimension amount of the input intermediate information. The second partial model M13 outputs the first character information corresponding to the input intermediate information as the output information OUT1 in FIG.

Further, the first partial model M1-2 includes a second partial model M14 which is a decoder for inputting intermediate information. The second partial model M14 generates the second character information corresponding to the input intermediate information by increasing the dimension amount of the input intermediate information. The second partial model M14 outputs the second character information corresponding to the input intermediate information as the output information OUT1 in FIG. In the example of FIG. 1, the second partial models M13 and M14 are shown individually for the sake of explanation, but the second partial models M13 and M14 may share a common part. For example, the second partial models M13 and M14 may be integrated. In other words, the second partial models M13 and M14 may be one model. For example, even if the second partial models M13 and M14 are integrated (same model), the first partial model M1-2 generates information according to the selection by the selection information by inputting different selection information (the same model). Output. For example, the first partial model M1-2 may be configured by removing overlapping portions from the configurations for realizing the functions of the second partial models M13 and M14.

Further, the first partial model M1-2 accepts selection information (control vector) for selecting the type of character information to be output. In the first partial model M1-2, selection information SL1 for selecting the type of character information to be output is input. That is, the first partial model M1-2 has a function of determining whether the input intermediate information is input to the second partial model M13 or M14 according to the selection information SL1. For example, the first partial model M1-2 has a function of accepting selection information and determining that intermediate information is input to the second partial models M13 and M14 corresponding to the selection information. For example, in the first partial model M1-2, the intermediate information is input to the input layer of the model based on the intermediate information among the second partial models M13 and M14 by a predetermined switching function (switching function) (logic). ) The circuit etc. may be switched.

For example, the first partial model M1-2 determines that intermediate information is input to the second partial model M13 when the selection information is information indicating "first type". For example, the first partial model M1-2 determines that intermediate information is input to the second partial model M14 when the selection information is information indicating "second type". The above is an example, and the information processing apparatus 100 may realize the determination of the model to be input by any function as long as the input to the desired model is possible.

The encoder and decoder are not only a neural network that simply changes the number of dimensions of the input information amount, but also an RNN (Recurrent Neural Network) such as a GRU (Gated Recurrent Unit) and an LSTM (Long Short-Term Memory). ) Etc. can be adopted.

The information processing apparatus 100 generates a conversion model by learning processing using the first character information (step S15). The information processing apparatus 100 generates the model M1 by performing learning using the first character information FCH11 to FCH14 and the like in the first type information storage unit 121 as learning data (teacher data). In FIG. 1, as shown in the first learning FLR1, the selection information FSL1 for selecting the first type is set. As a result, in the first learning FLR1, among the models M1, the second partial model M11 and the second partial model M13 are controlled to be used. Then, the information processing apparatus 100 receives the input information FIN1 which is the first character information as an input, and outputs the output information FOUT1 which is the first character information via the second partial model M11 and the second partial model M13. To do.

In the first learning FLR1, the information processing apparatus 100 performs learning processing so that the first character information input to the model M1 is output from the model M1. For example, when the first character information FCH 11 is input to the model M1, the information processing apparatus 100 performs a learning process so that the first character information FCH 11 is output from the model M1. For example, when the first character information FCH 12 is input to the model M1, the information processing apparatus 100 performs a learning process so that the first character information FCH 12 is output from the model M1.

For example, in the first learning FLR1, the information processing apparatus 100 performs learning processing using the second partial model M11 and the second partial model M13 as an autoencoder of the first character information. For example, the information processing apparatus 100 performs learning processing by a method such as backpropagation (error back propagation method) so that the character information output by the model M1 approaches the character information input to the model M1. For example, the information processing apparatus 100 modifies the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, thereby making the first part model M11 and the second part model M13 first. Learn the characteristics of textual information. In this way, the information processing apparatus 100 learns the model M1 by processing such as backpropagation that corrects the parameter (connection coefficient) so that the error between the output and the input in the model M1 is reduced. For example, the information processing apparatus 100 generates the model M1 by performing processing such as backpropagation so as to minimize a predetermined loss function. As a result, the information processing apparatus 100 can perform learning processing on the second partial model M11 and the second partial model M13 using only the first character information.

Further, the information processing apparatus 100 generates a conversion model by a learning process using the second character information (step S16). The information processing apparatus 100 generates the model M1 by performing learning using the second character information SCH11 to SCH14 and the like in the second type information storage unit 122 as learning data (teacher data). In FIG. 1, as shown in the second learning SLR1, the selection information SSL1 for selecting the second type is set. As a result, in the second learning SLR1, the second partial model M12 and the second partial model M14 are controlled to be used among the models M1. Then, the information processing apparatus 100 receives the input information SIN1 which is the second character information as an input, and outputs the output information SOUT1 which is the second character information via the second partial model M12 and the second partial model M14. To do.

In the second learning SLR1, the information processing apparatus 100 performs learning processing so that the second character information input to the model M1 is output from the model M1. For example, when the second character information SCH11 is input to the model M1, the information processing apparatus 100 performs a learning process so that the second character information SCH11 is output from the model M1. For example, when the second character information SCH12 is input to the model M1, the information processing apparatus 100 performs a learning process so that the second character information SCH12 is output from the model M1.

For example, in the second learning SLR 1, the information processing apparatus 100 performs learning processing using the second partial model M12 and the second partial model M14 as an autoencoder of the second character information. For example, the information processing apparatus 100 performs learning processing by a method such as backpropagation so that the character information output by the model M1 approaches the character information input to the model M1. For example, the information processing apparatus 100 modifies the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes to the second part model M12 and the second part model M14. Learn the characteristics of textual information. As a result, the information processing apparatus 100 can perform learning processing on the second partial model M12 and the second partial model M14 using only the second character information.

As described above, the information processing apparatus 100 performs learning processing on the second partial model M11 and the second partial model M13 in the first learning FLR1, and the second partial model M12 and the second in the second learning SLR1. The learning process is performed on the two-part model M14. As a result, the information processing apparatus 100 can learn about each of the second partial models M11 to M14 included in the model M1.

As described above, the information processing apparatus 100 associates the first character information with the second character information by performing the learning process using the first character information and the learning process using the second character information. Even if there is no learning data (teacher data), the learning process can be performed appropriately. That is, it is difficult for the information processing apparatus 100 to appropriately provide a service related to character information conversion when the combination (teacher data) of the first type character information and the second type character information is insufficient. It is possible to solve such problems. Then, the information processing apparatus 100 can appropriately convert the character information type by converting the character information type using the learned model, and can perform various services.

(1-1. 1st type, 2nd type)
The model learning method is not limited to the above-mentioned method, and any known technique can be applied. Further, for example, even if the information processing apparatus 100 has a small number, if there is teacher data (hereinafter, also referred to as "pair data") in which the first character information and the second character information are associated with each other, the pair data is used. You may use it. For example, when the information processing apparatus 100 has pair data, learning may be performed using one of the pair data as input information and the other as output correct answer information.

The information processing apparatus 100 controls and learns to use the second partial model M11 and the second partial model M14 when the first character information of the pair data is input and the second character information is output. Perform processing. Then, the information processing apparatus 100 receives the first character information as an input and learns to output the second character information via the second partial model M11 and the second partial model M14. Further, the information processing apparatus 100 controls to use the second partial model M12 and the second partial model M13 when the second character information of the pair data is input and the first character information is output. , Perform learning process. Then, the information processing apparatus 100 receives the second character information as an input, and learns to output the first character information via the second partial model M12 and the second partial model M13. As a result, the information processing apparatus 100 can improve the conversion accuracy of the model M1.

(1-2. Selection information (control vector))
In the example of FIG. 1, the case where which decoder is used is controlled by the selection information has been described as an example, but the information processing apparatus 100 may be expanded so that the control vector can be received. For example, the information processing apparatus 100 may combine a word vector and a control vector and regard it as a normal word vector. Further, the information processing apparatus 100 may use various control vectors. For example, the information processing apparatus 100 may use a different control vector for each word. For example, the information processing apparatus 100 may combine different control vectors for each word vector. For example, the information processing apparatus 100 may use a control vector corresponding to the number of remaining characters when summarizing. For example, when summarizing, the information processing apparatus 100 may use information on how many remaining characters and how many characters are output as a control vector (referred to as "control vector CX"). For example, the information processing apparatus 100 may perform character string conversion processing using a vector in which a word vector and a control vector CX are combined.

(1-3. 1st type, 2nd type)
In the example of FIG. 1, the case where the first type is a news article and the second type is an article for children is shown, but the information processing apparatus 100 may perform processing for various types. The first type may be character information including characters having a first number of characters or more, and the second type may be character information including characters having a number of first characters or less and less than the second number of characters. For example, the first type may be the text (article text), and the second type may be a summary. Further, there may be two or more types of signals for controlling the output.

(1-4. 3 or more types)
In the example of FIG. 1, the conversion of two types, the first type and the second type, has been described as an example, but the information processing apparatus 100 learns a conversion model for mutually converting character information of three or more types. You may. For example, a conversion model may be generated that mutually converts three types of character information, the first type for adults, the second type for elementary school students, and the third type for high school students.

[2. Information processing (provided processing)]
An example of information processing according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram showing an example of information processing according to an embodiment. Specifically, FIG. 2 describes an example of the provision process among the information processing according to the embodiment. FIG. 2 shows a case where the information processing apparatus 100 provides the converted character information obtained by converting the type of the character information acquired from the information providing apparatus 50 to the information providing apparatus 50.

First, the information processing device 100 acquires character information to be generated as converted character information from the information providing device 50 used by the information provider CP1 (step S21). In the example of FIG. 2, the information processing apparatus 100 acquires the target character information INF21, which is the character information of the first type, from the information providing apparatus 50.

Then, the information processing apparatus 100 acquires the conversion model (step S22). In the example of FIG. 2, the information processing apparatus 100 acquires the model M1 from the model information storage unit 123. Then, the information processing apparatus 100 uses the model M1 to generate the converted character information INF22 in which the type of the target character information INF21 is converted. Specifically, the information processing apparatus 100 generates the converted character information INF22 of the second type by inputting the target character information INF21 of the first type into the model M1.

In the example of FIG. 2, the information processing apparatus 100 generates the converted character information INF22 by the processing as shown in the processing group PS1. Since the target character information INF21 is the first type, the information processing apparatus 100 sets the selection information SSL1 for selecting the second type (step S23). Further, since the target character information INF21 is the first type, the information processing apparatus 100 sets specific information for selecting the first type. As a result, in the processing group PS1, the second partial model M11 and the second partial model M14 of the model M1 are controlled to be used.

Then, the information processing apparatus 100 inputs the target character information INF21 into the model M1 (step S24). The information processing apparatus 100 takes the input information TIN1 which is the target character information INF21 as an input, and outputs the output information TOUT1 which is the converted character information INF22 of the second type via the second partial model M11 and the second partial model M14. Let me. Specifically, the information processing apparatus 100 inputs the target character information INF21 into the second partial model M11. Then, the information processing apparatus 100 inputs the intermediate information INT1 output by the second partial model M11 into the second partial model M14. Then, the information processing apparatus 100 causes the second partial model M14 to output the converted character information INF22 (step S25). As a result, the information processing apparatus 100 generates the converted character information INF22 in which the type of the target character information INF21 of the first type is converted into the second type.

Then, the information processing apparatus 100 provides the converted character information INF22 to the information providing apparatus 50 (step S26). In the example of FIG. 2, the information processing apparatus 100 transmits the converted character information INF 22 to the information providing apparatus 50.

As described above, the information processing apparatus 100 appropriately provides the service related to the character information conversion by providing the converted character information INF22 in which the type of the target character information INF21 is converted into the second type by using the model M1. be able to.

[3. Information processing device configuration]
Next, the configuration of the information processing apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 4 is a diagram showing a configuration example of the information processing apparatus 100 according to the embodiment. As shown in FIG. 4, the information processing apparatus 100 includes a communication unit 110, a storage unit 120, and a control unit 130. The information processing device 100 includes an input unit (for example, a keyboard, a mouse, etc.) that receives various operations from the administrator of the information processing device 100, and a display unit (for example, a liquid crystal display, etc.) for displaying various information. You may have.

(Communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card) or the like. Then, the communication unit 110 is connected to the network by wire or wirelessly, and transmits / receives information to / from the terminal device 10 and the information providing device 50.

(Memory unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory (Flash Memory), or a storage device such as a hard disk or an optical disk. As shown in FIG. 4, the storage unit 120 according to the embodiment includes a first-type information storage unit 121, a second-type information storage unit 122, and a model information storage unit 123.

(First type information storage unit 121)
The first type information storage unit 121 according to the embodiment stores various information related to the first type. FIG. 5 is a diagram showing an example of the first type information storage unit according to the embodiment. For example, the first type information storage unit 121 stores teacher data related to the first type used for generating a model. The first type information storage unit 121 shown in FIG. 5 includes items such as "character information ID" and "first character information".

The "character information ID" indicates identification information for identifying character information. "First character information" indicates character information identified by the character information ID. In FIG. 5, an example in which conceptual information such as "FID11" is stored in "first character information" is shown, but in reality, a character string (character information) such as a specific sentence or its storage location is shown. The file path name indicating is stored. The first character information FCH 11 may be any character information as long as it is normal character information that is not intended for children. The first character information FCH 11 may be news related to politics and economy, news related to sports and entertainment, and the like. Further, the first character information FCH 11 may be an instruction manual of a product, an article of a law, or the like.

In the example shown in FIG. 5, it is shown that the character information identified by the character information ID "FID11" is the first character information FCH11.

The first type information storage unit 121 is not limited to the above, and may store various information depending on the purpose. For example, the first type information storage unit 121 may store information regarding the date and time when each character information (learning data) is added. Further, for example, the first type information storage unit 121 may store information indicating the date and time when each character information was created. Further, the first type information storage unit 121 may store information that specifies the type of each character information.

Further, when the first type is character information including characters of the first character number (for example, 200 characters), the first type information storage unit 121 stores character information of the first character number or more as the first character information. You may. For example, when the first type is the article body, the first type information storage unit 121 may store the article body as the first character information.

(Type 2 information storage unit 122)
The second type information storage unit 122 according to the embodiment stores various information related to the second type. FIG. 6 is a diagram showing an example of the second type information storage unit according to the embodiment. For example, the second type information storage unit 122 stores teacher data related to the second type used for generating a model. The second type information storage unit 122 shown in FIG. 6 includes items such as "character information ID" and "second character information".

The "character information ID" indicates identification information for identifying character information. "Second character information" indicates character information identified by the character information ID. In FIG. 6, an example in which conceptual information such as "SID11" is stored in "second character information" is shown, but in reality, a character string (character information) such as a specific sentence or its storage location is shown. The file path name indicating is stored. The second character information SCH11 may be any character information as long as it is character information for children. The second character information SCH11 may be any character information as long as it is character information created for children. The second character information SCH11 may be a text of a book for children or the like. Further, the second character information SCH11 may be news on politics and economy created for children, news on sports and entertainment, and the like.

In the example shown in FIG. 6, it is shown that the character information identified by the character information ID "SID11" is the second character information SCH11.

The second type information storage unit 122 is not limited to the above, and may store various information depending on the purpose. For example, the second type information storage unit 122 may store information regarding the date and time when each character information (learning data) is added. Further, for example, the second type information storage unit 122 may store information indicating the date and time when each character information was created. Further, the second type information storage unit 122 may store information that specifies the type of each character information.

Further, when the second type is character information including characters less than the second character number (for example, 15 characters), the second type information storage unit 122 stores character information less than the second character number as the second character information. You may. For example, when the second type is a summary, the second type information storage unit 122 may store the summary as the second character information.

(Model information storage unit 123)
The model information storage unit 123 according to the embodiment stores information about the model. For example, the model information storage unit 123 stores model information (model data) generated by information processing. FIG. 7 is a diagram showing an example of a model information storage unit according to an embodiment. The model information storage unit 123 shown in FIG. 7 includes items such as "model ID", "first partial model ID", "second partial model ID", "use", and "model data". Although only the model M1 is shown in FIG. 7, a plurality of model information such as models M3 and M4 may be stored.

The "model ID" indicates identification information for identifying the model. For example, the model identified by the model ID "M1" corresponds to the model M1 shown in the example of FIG. The "first part model ID" indicates identification information for identifying the first part model of the corresponding model. For example, the first partial model identified by the partial model ID "M1-1" corresponds to the first partial model M1-1 shown in the example of FIG.

The "second part model ID" indicates identification information for identifying the second part model of the corresponding model. The "second partial model ID" indicates a model in which the corresponding first partial model is further subdivided. For example, the partial model identified by the second partial model ID "M11" corresponds to the second partial model M11 shown in the example of FIG. "Use" indicates the use of the corresponding partial model (second part model). Further, the "model data" indicates the data of the corresponding partial model (second partial model). For example, "model data" includes information including nodes in each layer, functions adopted by each node, connection relationships of the nodes, and connection coefficients set for connections between the nodes.

For example, in the example shown in FIG. 7, it is shown that the model M1 includes the first partial models M1-1, M1-2, and the like. It is shown that the first partial model M1-1 includes the second partial models M11 and M12. It is shown that the first partial model M1-2 includes the second partial models M13 and M14.

Further, for example, the model (second part model M11) identified by the second part model ID "M11" has a purpose of "first character information ⇒ intermediate information" and is intermediate from the input first character information. Indicates that it is used to generate information. That is, it is shown that the second partial model M11 is used for extracting features from the input first character information. Further, it is shown that the model data of the second partial model M11 is the model data MDT11.

Further, for example, the model identified by the second part model ID "M14" (second part model M14) has a purpose of "intermediate information ⇒ second character information", and is used for generating second character information from the intermediate information. Indicates that it will be used. That is, it is shown that the second partial model M14 is used to generate the second character information from the intermediate information which is the information indicating the characteristics of the extracted character information. Further, it is shown that the model data of the second partial model M14 is the model data MDT14.

The model M1 (each model data MDT11 to MDT14, etc.) is an input layer into which a predetermined type of character information is input, an output layer, and any layer from the input layer to the output layer, other than the output layer. Including the first element belonging to the above and the second element whose value is calculated based on the weights of the first element and the first element, and for the character information input to the input layer, to each layer other than the output layer. A computer is used to output character information converted into a type different from a predetermined type from the output layer by performing an operation based on the first element and the weight of the first element with each element to which it belongs as the first element. It is a model to make it work.

For example, the first element included in the model M1 corresponds to either node of the input layer or the intermediate layer. Further, the second element corresponds to the node of the next stage, which is the node to which the value is transmitted from the node corresponding to the first element. Further, the weight of the first element corresponds to a connection coefficient which is a weight considered for the value transmitted from the node corresponding to the first element to the node corresponding to the second element.

The model information storage unit 123 is not limited to the above, and may store various model information depending on the purpose.

(Control unit 130)
Returning to the description of FIG. 4, the control unit 130 is a controller, and is stored in a storage device inside the information processing device 100 by, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). Various programs (corresponding to an example of an information processing program such as a learning program and a generation program) are realized by executing the RAM as a work area. Further, the control unit 130 is a controller, and is realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). The control unit 130 has an input layer, an output layer, and an input layer to an output layer into which a predetermined type of character information is input by information processing according to a model M1 or the like stored in the model information storage unit 123. A first element which is any layer and belongs to a layer other than the output layer, and a second element whose value is calculated based on the weights of the first element and the first element are included and input to the input layer. The character information was converted into a type different from the predetermined type by performing an operation based on the first element and the weight of the first element, with each element belonging to each layer other than the output layer as the first element. Output character information from the output layer.

As shown in FIG. 4, the control unit 130 includes an acquisition unit 131, a learning unit 132, a generation unit 133, and a provision unit 134, and realizes or executes an information processing function or operation described below. .. The internal configuration of the control unit 130 is not limited to the configuration shown in FIG. 4, and may be any other configuration as long as it is configured to perform information processing described later. The control unit 130 calculates a value based on the first element and the weight of the first element by information processing according to the model M1 (each model data MDT11 to MDT14, etc.) stored in the storage unit 120. For character information of a predetermined type input to the input layer including two elements, an operation based on the first element and the weight of the first element is performed with each element belonging to each layer other than the output layer as the first element. By doing so, the character information converted into a type different from the predetermined type is output from the output layer.

The control unit 130 calculates a value based on the first element and the weight of the first element by information processing according to the model M1 (model data MDT11 to MTD14, etc.) stored in the storage unit 120. For input information input to the input layer including elements, input is performed by performing an operation based on the weights of the first element and the first element, with each element belonging to each layer other than the output layer as the first element. The character information in which the type of the converted character information is converted is output from the output layer. The model such as the model M1 may be expected to be used as a program module which is a part of the artificial intelligence software. A model such as model M1 may be a program.

(Acquisition unit 131)
The acquisition unit 131 acquires various types of information. The acquisition unit 131 acquires various information from an external information processing device such as the terminal device 10 or the information providing device 50. The acquisition unit 131 may acquire various information shown in the first type information storage unit 121, the second type information storage unit 122, the model information storage unit 123, and the like from an external information processing device.

The acquisition unit 131 acquires various information from the storage unit 120. The acquisition unit 131 acquires various types of information from the first type information storage unit 121, the second type information storage unit 122, the model information storage unit 123, and the like.

The acquisition unit 131 acquires the first character information which is the character information of the first type and the second character information which is the character information of the second type different from the first type. The acquisition unit 131 acquires selection information for selecting the type of character information output by the conversion model. The acquisition unit 131 acquires selection information for selecting a learning type, which is a type corresponding to the learning character information input to the conversion model.

The acquisition unit 131 has a conversion model that converts the first character information, which is the character information of the first type, into the second character information, which is the character information of the second type different from the first type, and the target character of the first type. Get information and. The acquisition unit 131 acquires a conversion model that converts the second character information into the first character information and other target character information of the second type. The acquisition unit 131 acquires selection information for selecting the type of character information output by the conversion model. The acquisition unit 131 acquires the target character information to be input to the conversion model and the selection information for selecting the other type, which is one of the first type and the second type and is different from the type of the target character information.

In the example of FIG. 1, the acquisition unit 131 acquires the first character information as the learning data LD1. The acquisition unit 131 adds the first character information included in the learning data LD1 as learning data. The acquisition unit 131 adds the learning data LD1 to the first type information storage unit 121. The acquisition unit 131 acquires the second character information as the learning data LD2. The acquisition unit 131 adds the second character information included in the learning data LD2 as learning data. The acquisition unit 131 adds the learning data LD2 to the second type information storage unit 122.

In the example of FIG. 2, the acquisition unit 131 acquires character information for which conversion character information is to be generated from the information providing device 50 used by the information provider CP1. The acquisition unit 131 acquires the target character information INF21, which is the character information of the first type, from the information providing device 50. The acquisition unit 131 acquires the conversion model. The acquisition unit 131 acquires the model M1 from the model information storage unit 123.

(Learning Department 132)
The learning unit 132 learns various information. The learning unit 132 generates various models by learning processing. For example, the learning unit 132 performs learning processing based on the information stored in the first type information storage unit 121, the second type information storage unit 122, and the like. The learning unit 132 generates various models based on the information stored in the first type information storage unit 121, the second type information storage unit 122, and the like.

For example, the learning unit 132 uses the learning data stored in the first type information storage unit 121 and the second type information storage unit 122 to generate a model as shown in the model information storage unit 123. For example, the learning unit 132 generates a model used to generate character information converted from character information of a predetermined type to a type different from the predetermined type based on the learning data acquired by the acquisition unit 131. .. The learning unit 132 generates a model and stores the generated model in the model information storage unit 123. Specifically, the learning unit 132 is a layer that is any of an input layer into which a predetermined type of character information is input, an output layer, and a layer from the input layer to the output layer, and belongs to a layer other than the output layer. Each element belonging to each layer other than the output layer with respect to the character information input to the input layer, including one element and the second element whose value is calculated based on the first element and the weight of the first element. By performing an operation based on the first element and the weight of the first element, a model is generated in which the character information is converted from a predetermined type to another type and the character information is output from the output layer. ..

The learning unit 132 uses the first character information as input information, learning using the first character information as correct answer information, second character information as input information, and the second character information as correct answer information. Based on the learning, a conversion model is generated that outputs the converted character information in which the target character information is converted to the second type in response to the input of the target character information of the first type. The learning unit 132 has a coding unit that outputs intermediate information indicating the characteristics of the first character information in response to input of the first character information, and converted character information corresponding to the intermediate information using the intermediate information as input information. Generates a conversion model that includes a decoding unit that outputs.

The learning unit 132 has a coding unit that outputs intermediate information indicating the characteristics of the second character information in response to input of the second character information, and a first character corresponding to the intermediate information using the intermediate information as input information. It is a conversion model including a decoding unit that outputs information, and outputs other converted character information in which other target character information is converted to the first type in response to input of other target character information of the second type. Generate a transformation model to do. The learning unit 132 generates a conversion model that outputs character information of the type corresponding to the selection information. The learning unit 132 generates a conversion model by performing a learning process so as to output the learning character information based on the learning type in response to the input of the learning character information.

The learning unit 132 has a first encoder that outputs intermediate information indicating the characteristics of the first character information in response to the input of the first character information, and the learning unit 132 has the second character information in response to the input of the second character information. A conversion model including a coding unit having a second encoder that outputs intermediate information indicating the characteristics is generated. The learning unit 132 generates a conversion model including a decoding unit having a first decoder that outputs the first character information corresponding to the intermediate information and a second decoder that outputs the second character information corresponding to the intermediate information. ..

In the example of FIG. 1, the learning unit 132 learns the first character information FCH11 to FCH14 and the like in the first type information storage unit 121 and the second character information SCH11 to SCH14 and the like in the second type information storage unit 122 ( A transformation model is generated by learning as teacher data). The learning unit 132 generates a conversion model by a learning process using the first character information. The learning unit 132 generates the model M1 by performing learning using the first character information FCH11 to FCH14 and the like in the first type information storage unit 121 as learning data (teacher data). The learning unit 132 receives the input information FIN1 which is the first character information as an input, and performs learning to output the output information FOUT1 which is the first character information via the second partial model M11 and the second partial model M13.

In the first learning FLR1, the learning unit 132 performs a learning process so that the first character information input to the model M1 is output from the model M1. When the first character information FCH 11 is input to the model M1, the learning unit 132 performs a learning process so that the first character information FCH 11 is output from the model M1. When the first character information FCH 12 is input to the model M1, the learning unit 132 performs a learning process so that the first character information FCH 12 is output from the model M1.

In the first learning FLR1, the learning unit 132 performs learning processing using the second partial model M11 and the second partial model M13 as an autoencoder of the first character information. The learning unit 132 performs learning processing by a method such as backpropagation (error back propagation method) so that the character information output by the model M1 approaches the character information input to the model M1. The learning unit 132 corrects the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, so that the first character information is transmitted to the second part model M11 and the second part model M13. Learn the characteristics they have. The learning unit 132 learns the model M1 by processing such as backpropagation that corrects a parameter (connection coefficient) so that an error between the output and the input in the model M1 is reduced. The learning unit 132 generates the model M1 by performing processing such as backpropagation so as to minimize a predetermined loss function.

The learning unit 132 generates a conversion model by a learning process using the second character information. The learning unit 132 generates the model M1 by performing learning using the second character information SCH11 to SCH14 and the like in the second type information storage unit 122 as learning data (teacher data). The learning unit 132 receives the input information SIN1 which is the second character information as an input, and performs learning to output the output information SOUT1 which is the second character information via the second partial model M12 and the second partial model M14.

In the second learning SLR1, the learning unit 132 performs a learning process so that the second character information input to the model M1 is output from the model M1. When the second character information SCH11 is input to the model M1, the learning unit 132 performs a learning process so that the second character information SCH11 is output from the model M1. When the second character information SCH12 is input to the model M1, the learning unit 132 performs a learning process so that the second character information SCH12 is output from the model M1.

In the second learning SLR1, the learning unit 132 performs learning processing using the second partial model M12 and the second partial model M14 as an autoencoder of the second character information. The learning unit 132 performs learning processing by a method such as backpropagation so that the character information output by the model M1 approaches the character information input to the model M1. The learning unit 132 corrects the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, so that the second character information is transmitted to the second part model M12 and the second part model M14. Learn the characteristics they have.

In the first learning FLR1, the learning unit 132 performs learning processing on the second partial model M11 and the second partial model M13, and in the second learning SLR 1, the second partial model M12 and the second partial model M14 are used. Perform learning processing as a target. The learning unit 132 learns about each of the second partial models M11 to M14 included in the model M1.

(Generation unit 133)
The generation unit 133 generates various information. The generation unit 133 generates various character information by the generation process. For example, the generation unit 133 performs a generation process based on the information stored in the model information storage unit 123 and the like. The learning unit 132 generates character information using the model stored in the model information storage unit 123.

The generation unit 133 generates the converted character information in which the target character information is converted into the second type by inputting the target character information into the conversion model. By inputting other target character information into the conversion model, the generation unit 133 generates other converted character information in which the other target character information is converted into the first type. The generation unit 133 generates character information of the type corresponding to the selection information by inputting the selection information into the conversion model. The generation unit 133 generates the converted character information in which the target character information is converted into the other type by inputting the target character information and the selection information into the conversion model.

In the example of FIG. 2, the generation unit 133 uses the model M1 to generate the converted character information INF22 obtained by converting the type of the target character information INF21. The generation unit 133 generates the conversion character information INF22 of the second type by inputting the target character information INF21 of the first type into the model M1. The generation unit 133 generates the converted character information INF22 by the processing as shown in the processing group PS1. Since the target character information INF21 is the first type, the generation unit 133 sets the selection information SSL1 for selecting the second type. Since the target character information INF21 is the first type, the generation unit 133 sets specific information for selecting the first type.

The generation unit 133 inputs the target character information INF21 into the model M1. The generation unit 133 receives the input information TIN1 which is the target character information INF21 as an input, and outputs the output information TOUT1 which is the conversion character information INF22 of the second type via the second partial model M11 and the second partial model M14. .. The generation unit 133 inputs the target character information INF21 to the second partial model M11. The generation unit 133 inputs the intermediate information INT1 output by the second partial model M11 to the second partial model M14. The generation unit 133 causes the second partial model M14 to output the converted character information INF22. The generation unit 133 generates the converted character information INF22 in which the type of the target character information INF21 of the first type is converted into the second type.

(Providing section 134)
The providing unit 134 provides various information. For example, the providing unit 134 provides various information to the terminal device 10 and the information providing device 50. The providing unit 134 transmits various information to the terminal device 10 and the information providing device 50. The providing unit 134 distributes various information to the terminal device 10 and the information providing device 50. The providing unit 134 provides various information based on various information acquired by the acquiring unit 131. The providing unit 134 provides various information based on the various information learned by the learning unit 132. The providing unit 134 provides various information based on the model learned by the learning unit 132. The providing unit 134 provides various information based on various information generated by the generating unit 133. The providing unit 134 provides various information based on the converted character information generated by the generating unit 133.

In the example of FIG. 2, the providing unit 134 provides the converted character information INF 22 to the information providing device 50. The providing unit 134 transmits the converted character information INF 22 to the information providing device 50. The providing unit 134 provides a content distribution service using the converted character information. The providing unit 134 provides an advertisement distribution service using the converted character information.

[4. Information processing (learning process) flow]
Next, the procedure of the learning process in the information processing by the information processing system 1 according to the embodiment will be described with reference to FIG. FIG. 8 is a flowchart showing an example of information processing according to the embodiment.

As shown in FIG. 8, the information processing apparatus 100 acquires the first character information of the first type (step S101). For example, the information processing apparatus 100 acquires the first character information as learning data from the first type information storage unit 121.

After that, the information processing apparatus 100 learns the conversion model using the first character information as input information and the first character information as correct answer information (step S102). For example, the information processing apparatus 100 learns a conversion model using the first character information stored in the first type information storage unit 121.

Further, the information processing apparatus 100 acquires the second character information of the second type (step S103). For example, the information processing apparatus 100 acquires the second character information as learning data from the second type information storage unit 122.

After that, the information processing apparatus 100 learns the conversion model using the second character information as input information and the second character information as correct answer information (step S102). In the example of FIG. 1, the information processing apparatus 100 learns a conversion model using the second character information stored in the first type information storage unit 121.

[5. Information processing (generation processing) flow]
Next, the procedure of the generation process (conversion process) in the information processing by the information processing system 1 according to the embodiment will be described with reference to FIG. 9. FIG. 9 is a flowchart showing an example of information processing according to the embodiment.

As shown in FIG. 9, the information processing apparatus 100 acquires a conversion model (step S201). Further, the information processing apparatus 100 acquires the target character information of the first type (step S202). Then, the information processing apparatus 100 inputs the target character information into the conversion model to generate the converted character information in which the target character information is converted into the second type (step S203). The information processing apparatus 100 provides a service based on the converted character information (step S204).

[6. Modification example]
The information processing system 1 according to the above-described embodiment may be implemented in various different embodiments other than the above-described embodiment. Hereinafter, another embodiment of the above information processing system 1 will be described. The same points as those in the above-described embodiment will be appropriately described by adding the same reference numerals. For example, the information processing system 1 according to the embodiment and the information processing system (not shown) in the modified example have the same configuration except for the difference between the information processing device 100 and the information processing device 100A. Hereinafter, information processing by the information processing apparatus 100A will be described.

[6-1. Information processing (learning process)]
First, the estimation process in the information processing system according to the modified example will be described with reference to FIG. FIG. 10 is a diagram showing an example of information processing according to a modified example. Specifically, FIG. 10 describes an example of learning processing among the information processing according to the embodiment. FIG. 1 shows a case where the information processing apparatus 100A generates a conversion model for converting the type of character information of the first type into the second type based on the first character information and the second character information.

First, the information processing apparatus 100A acquires the first character information as the learning data LD1 (step S51). Then, the information processing apparatus 100A adds the first character information included in the learning data LD1 acquired in step S51 as learning data (step S52). Specifically, the information processing apparatus 100A adds the learning data LD1 to the first type information storage unit 121.

The first type information storage unit 121 shown in FIG. 10 includes items such as "character information ID" and "first character information".

The "character information ID" in the first type information storage unit 121 of FIG. 10 indicates identification information for identifying character information. The "first character information" in the first type information storage unit 121 of FIG. 10 indicates character information identified by the character information ID.

In the example shown in FIG. 10, the character information identified by the character information ID "FID11" indicates that it is the first character information FCH11, and the character information identified by the character information ID "FID12" is the first character information. Indicates that it is FCH12. Further, the character information identified by the character information ID "FID13" indicates that it is the first character information FCH13, and the character information identified by the character information ID "FID14" is the first character information FCH14. show. The first character information FCH11 to FCH14, etc. is any character information as long as it is general news not for children (hereinafter, also referred to as "normal news") or character information distributed by ordinary content distribution. May be. For example, the first character information FCH11 to FCH14 and the like may be ordinary news about politics and economy, ordinary news about sports and entertainment, and the like.

Further, the information processing apparatus 100A acquires the second character information as the learning data LD2 (step S53). Then, the information processing apparatus 100A adds the second character information included in the learning data LD2 acquired in step S53 as learning data (step S54). Specifically, the information processing apparatus 100A adds the learning data LD2 to the second type information storage unit 122.

The second type information storage unit 122 shown in FIG. 10 includes items such as "character information ID" and "second character information".

The "character information ID" in the second type information storage unit 122 of FIG. 10 indicates identification information for identifying character information. The "second character information" in the second type information storage unit 122 of FIG. 10 indicates character information identified by the character information ID.

In the example shown in FIG. 10, the character information identified by the character information ID "SID11" indicates that it is the second character information SCH11, and the character information identified by the character information ID "SID12" is the second character information. Indicates that it is SCH12. Further, the character information identified by the character information ID "SID13" indicates that it is the second character information SCH13, and the character information identified by the character information ID "SID14" is the second character information SCH14. show. The second character information SCH11 to SCH14 and the like may be any character information as long as it is character information such as news created for children (hereinafter, also referred to as "news for children"). For example, the second character information SCH11 to SCH14 and the like may be children's news created for children, children's news about sports and entertainment, and the like.

Then, the information processing apparatus 100A generates a model by learning processing based on the learning data. For example, the information processing apparatus 100A uses learning data (teacher data) for the first character information FCH11 to FCH14 and the like in the first type information storage unit 121 and the second character information SCH11 to SCH14 and the like in the second type information storage unit 122. As a result, a transformation model is generated by training. It is assumed that there is no correspondence between the first character information FCH11 to FCH14 and the like and the second character information SCH11 to SCH14 and the like in the second type information storage unit 122. For example, the first character information FCH11 to FCH14 and the like and the second character information SCH11 to SCH14 and the like in the second type information storage unit 122 are assumed to be news with different contents.

First, a conversion model (learner) generated by the information processing apparatus 100A will be briefly described. In the example of FIG. 10, the model M2 in the model information storage unit 123A is a model of the architecture as shown in the structural information ARC2. In the example of FIG. 10, the model M2, which is a conversion model generated by the information processing apparatus 100A, is a model including a plurality of partial models M21, M22, and the like.

As described above, when "partial model M * (* is an arbitrary numerical value)" is described, it means that the partial model is a model identified by the partial model ID "M *". For example, when the description is described as "partial model M21", the partial model is a model identified by the partial model ID "M21".

The partial model M21 is a coding unit, and corresponds to an encoder that extracts the features of the input information by compressing the dimensional amount of the input information. As shown in FIG. 10, the partial model M21 corresponds to a model in which the input information IN2 is input and the intermediate information INT2 is output. The partial model M21 corresponds to a model in which first character information and second character information are input as input information IN2 and intermediate information INT2 is output. That is, the partial model M21 is an encoder that inputs the first character information and the second character information. The partial model M21 extracts the features of the input first character information and the second character information by compressing the dimensional amount of the input first character information and the second character information. The partial model M21 outputs, as the intermediate information INT2 in FIG. 10, intermediate information indicating the characteristics of the first character information and the second character information in which the dimension amount of the first character information is compressed. For example, the intermediate information INT2 in FIG. 10 is an intermediate output (encoded vector or the like) showing the characteristics of the input information, and may be, for example, a distributed representation or a real value vector. For example, the intermediate information INT2 includes information regarding the meaning (concept) of the character information input to the model M2.

The partial model M22 is a decoding unit, and corresponds to a decoder (decoder) that increases the dimensional amount of the feature extracted by the encoder and outputs character information of a predetermined type. That is, the partial model M22 corresponds to a decoder that takes the intermediate information output by the encoder as an input, increases the dimensional amount of the input intermediate information, and outputs the second type of character information. As shown in FIG. 10, the partial model M22 corresponds to a model in which the intermediate information INT2 is input and the output information OUT2 is output. That is, the partial model M22 is a decoder that inputs intermediate information. The partial model M22 generates the second character information corresponding to the input intermediate information by increasing the dimension amount of the input intermediate information. The partial model M22 outputs the second character information corresponding to the input intermediate information as the output information OUT2 in FIG.

Further, the model M2 receives the type information (control vector) indicating the type of the character information input at the time of learning. In the model M2, the type information TL2 for selecting the type of the character information to be input is input. That is, the model M2 has a function of specifying the type of the input character information according to the type information TL2 and performing learning according to the type. For example, the model M2 has a function of accepting type information and causing the partial model M22 to perform learning according to the type information. For example, the model M2 causes the partial model M22 to perform learning processing according to the type information TL2 by a predetermined switching function (switching function).

For example, in the partial model M22, when the type information is information indicating "first type", the partial model M22 is made to perform learning corresponding to the first type. For example, the model M2 causes the partial model M22 to perform learning processing such as reinforcement learning when the type information TL2 is the first type. For example, the partial model M22 determines that intermediate information is input to the second partial model M24 when the type information is information indicating "second type". For example, the model M2 causes the partial model M22 to perform learning processing as an autoencoder so as to decode the input information when the type information TL2 is the second type. The above is an example, and the information processing apparatus 100A may realize the determination of the model to be input by any function as long as the input to the desired model is possible.

Further, in the example of FIG. 10, the information processing apparatus 100A performs learning processing using the model M5 as shown in the structural information ARC2.

The model M5 is a decoding unit other than the model M2, and is a decoder (corresponding to a decoder) that increases the dimensional amount of the feature extracted by the encoder of the model M2 and outputs a predetermined type of character information. The model M5 corresponds to a decoder that takes the intermediate information output by the encoder of the model M2 as an input, increases the dimensional amount of the input intermediate information, and outputs the character information input to the encoder of the model M2. As shown in 10, the model M5 corresponds to a model that outputs the learning information OUT5 by using the intermediate information INT2 output by the encoder of the model M2 as an input. That is, the model M5 is a decoder that inputs the intermediate information. The model M5 generates (decodes) the character information that is the basis of the input intermediate information by increasing the dimension amount of the input intermediate information. The model M5 is the learning information in FIG. As OUT5, the character information corresponding to the input intermediate information is output. The model M5 may be included in the model M2.

As the encoder and decoder, not only a neural network that simply changes the number of dimensions of the input information amount, but also various neural networks such as RNN such as GRU and LSTM can be adopted.

The information processing apparatus 100A generates a conversion model by learning processing using the first character information (step S55). The information processing apparatus 100A generates a model M2 by learning using the first character information FCH11 to FCH14 and the like in the first type information storage unit 121 as learning data (teacher data). In FIG. 10, as shown in the first learning FLR2, the type information FTL2 for selecting the first type is set. As a result, in the first learning FLR2, the model M2 is controlled so as to perform learning corresponding to the first type for the partial model M22. Then, the information processing apparatus 100A receives the input information FIN2 which is the first character information as an input, and learns to output the output information FOUT2 which is the second character information corresponding to the input information FIN2.

In the first learning FLR2, the information processing apparatus 100A performs learning processing so that the second character information obtained by converting the first character information input to the model M2 into the second type is output from the model M2. For example, when the first character information FCH 11 is input to the model M2, the information processing apparatus 100A performs a learning process so that the second type of character information corresponding to the first character information FCH 11 is output from the model M2. For example, when the first character information FCH 12 is input to the model M2, the information processing apparatus 100A performs a learning process so that the second type of character information corresponding to the first character information FCH 12 is output from the model M2.

For example, the information processing apparatus 100A performs learning processing in the first learning FLR2 so that the number of terms for children (hereinafter referred to as “children's terms”) included in the output information FOUT2 increases. For example, children's terms may be words that are characteristic of articles for children. For example, the information processing apparatus 100A may calculate the score of the output information FOUT2 by using the list information of the children's terms used in the article for children, and learn so that the score becomes high. For example, the information processing apparatus 100A may calculate the score so that the higher the ratio of the child terms included in the output information FOUT2, the higher the score, and perform learning so that the score of the output information FOUT2 becomes higher. For example, the information processing apparatus 100A may perform learning processing by a method such as backpropagation (error back propagation method) so that the score of the output information FOUT2 becomes high. As a result, the information processing apparatus 100A can perform learning processing on the model M2. The information processing apparatus 100A is not limited to the above, and learning may be performed by appropriately using various information.

For example, the information processing apparatus 100A may use information output by a classification model (hereinafter, also referred to as “classification model MX”) for classifying news articles and articles for children. For example, the information processing apparatus 100A may use the score output by the classification model MX. The classification model MX may be any model as long as it is a model for classifying news articles and articles for children, and may be in various forms such as SVM (Support Vector Machine). Here, it is assumed that the classification model MX is a model trained to output a high score when an article for children is input and to output a low score when a normal news article is input. In this case, the information processing apparatus 100A may input the output information FOUT2 into the classification model MX and perform learning so that the score output by the classification model MX becomes high.

Further, in the information processing apparatus 100A, when the first type is the article text and the second type is the summary, reinforcement learning may be performed using the number of characters as an evaluation standard. For example, in the information processing apparatus 100A, when the first type is the article text and the second type is the summary, the information processing apparatus 100A learns so that the number of characters in the output information FOUT2 is smaller than a predetermined threshold value (for example, "20"). May be done. For example, in the information processing apparatus 100A, when the first type is the article text and the second type is the summary, the score is calculated so that the smaller the number of characters in the output information FOUT2, the higher the score, and the score of the output information FOUT2. You may study so that the value becomes high.

Further, the information processing apparatus 100A performs learning processing using the model M5 in the first learning FLR2. The information processing apparatus 100A performs learning processing so that the first character information input to the model M2 is output from the model M5. For example, when the first character information FCH 11 is input to the model M2, the information processing apparatus 100A performs a learning process so that the first character information FCH 11 is output from the model M5. For example, when the first character information FCH 12 is input to the model M2, the information processing apparatus 100A performs a learning process so that the first character information FCH 12 is output from the model M5.

For example, in the first learning FLR2, the information processing apparatus 100A performs learning processing using the partial model M21 and the model M5 of the model M2 as the autoencoder of the first character information. For example, the information processing apparatus 100A performs learning processing by a method such as backpropagation so that the character information output by the model M5 approaches the character information input to the model M2. For example, the information processing apparatus 100A modifies the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, so that the partial model M21 and the model M5 of the model M2 have the first character information. Learn the characteristics of. For example, the information processing apparatus 100A performs learning processing by a method such as backpropagation (error back propagation method) so that the character information output by the model M5 approaches the character information input to the model M2. For example, the information processing apparatus 100A modifies the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, so that the partial model M21 and the model M5 of the model M2 have the first character information. Learn the characteristics of. In this way, the information processing apparatus 100A learns by processing such as backpropagation that corrects parameters (connection coefficients) so that the error between the output in the model M5 and the input in the model M2 is reduced, so that the model M2 and the model M2 and the model M2 can be reduced. Generate model M5. For example, the information processing apparatus 100A generates the model M2 and the model M5 by performing processing such as backpropagation so as to minimize a predetermined loss function. As a result, the information processing apparatus 100A can perform learning processing on the model M2 and the model M5 using only the first character information.

Further, the information processing apparatus 100A generates a conversion model by a learning process using the second character information (step S56). The information processing apparatus 100A generates a model M2 by learning using the second character information SCH11 to SCH14 in the second type information storage unit 122 as learning data (teacher data). In FIG. 10, as shown in the second learning SLR2, the type information STL2 for selecting the second type is set. As a result, in the second learning SLR2, the model M2 is controlled so as to perform learning corresponding to the second type for the partial model M22. Then, the information processing apparatus 100A learns to output the output information SOUT2 which is the second character information by inputting the input information SIN2 which is the second character information.

In the second learning SLR2, the information processing apparatus 100A performs learning processing so that the second character information input to the model M2 is output from the model M2. For example, when the second character information SCH11 is input to the model M2, the information processing apparatus 100A performs a learning process so that the second character information SCH11 is output from the model M2. For example, when the second character information SCH12 is input to the model M2, the information processing apparatus 100A performs a learning process so that the second character information SCH12 is output from the model M2.

For example, in the second learning SLR2, the information processing apparatus 100A performs learning processing using the partial model M21 and the partial model M22 as an autoencoder of the second character information. For example, the information processing apparatus 100A performs learning processing by a method such as backpropagation so that the character information output by the model M2 approaches the character information input to the model M2. For example, the information processing apparatus 100A has the second character information in the partial model M21 and the partial model M22 by modifying the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes. Learn the features. As a result, the information processing apparatus 100A can perform learning processing on the model M2.

Further, the information processing apparatus 100A performs learning processing using the model M5 in the second learning SLR2. The information processing apparatus 100A performs learning processing so that the second character information input to the model M2 is output from the model M5. For example, when the second character information SCH11 is input to the model M2, the information processing apparatus 100A performs a learning process so that the second character information SCH11 is output from the model M2. For example, when the second character information SCH12 is input to the model M2, the information processing apparatus 100A performs a learning process so that the second character information SCH12 is output from the model M2.

For example, in the second learning SLR2, the information processing apparatus 100A performs learning processing using the partial model M21 and the model M5 of the model M2 as an autoencoder of the second character information. For example, the information processing apparatus 100A performs learning processing by a method such as backpropagation so that the character information output by the model M5 approaches the character information input to the model M2. For example, the information processing apparatus 100A modifies the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, so that the partial model M21 and the model M5 of the model M2 have the second character information. Learn the characteristics of. For example, the information processing apparatus 100A performs learning processing by a method such as backpropagation (error back propagation method) so that the character information output by the model M5 approaches the character information input to the model M2. For example, the information processing apparatus 100A modifies the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes, so that the partial model M21 and the model M5 of the model M2 have the second character information. Learn the characteristics of. In this way, the information processing apparatus 100A learns by processing such as backpropagation that corrects parameters (connection coefficients) so that the error between the output in the model M5 and the input in the model M2 is reduced, so that the model M2 and the model M2 and the model M2 can be reduced. Generate model M5. For example, the information processing apparatus 100A generates the model M2 and the model M5 by performing processing such as backpropagation so as to minimize a predetermined loss function. As a result, the information processing apparatus 100A can perform learning processing on the model M2 and the model M5 using only the second character information.

As described above, the information processing apparatus 100A performs the learning process so as to convert the input character information of the first type into the second type and output it in the first learning FLR2, and in the second learning SLR2, the information processing apparatus 100A performs the learning process. The learning process is performed so as to output the input second type character information. As a result, the information processing apparatus 100A can learn so that the model M2 outputs the second type of character information.

As described above, the information processing apparatus 100A associates the first character information with the second character information by performing the learning process using the first character information and the learning process using the second character information. Even if there is no learning data (teacher data), the learning process can be performed appropriately. That is, it is difficult for the information processing apparatus 100A to appropriately provide a service related to character information conversion when the combination (teacher data) of the first type character information and the second type character information is insufficient. It is possible to solve such problems. Then, the information processing apparatus 100A can appropriately convert the type of character information and perform various services by converting the type of the first character information into the second type using the learned model.

[6-2. Information processing (provided processing)]
An example of information processing according to a modified example will be described with reference to FIG. FIG. 11 is a diagram showing an example of information processing according to a modified example. Specifically, FIG. 11 describes an example of the provision process among the information processing related to the modified example. FIG. 11 shows a case where the information processing apparatus 100A provides the converted character information obtained by converting the type of the character information acquired from the information providing apparatus 50 to the information providing apparatus 50.

First, the information processing device 100A acquires character information to be generated as converted character information from the information providing device 50 used by the information provider CP1 (step S61). In the example of FIG. 11, the information processing apparatus 100A acquires the target character information INF 51, which is the character information of the first type, from the information providing apparatus 50.

Then, the information processing apparatus 100A acquires the conversion model (step S62). In the example of FIG. 11, the information processing apparatus 100A acquires the model M2 from the model information storage unit 123A. The information processing apparatus 100A does not acquire the model M5, which is necessary only at the time of learning. Then, the information processing apparatus 100A uses the model M2 to generate the converted character information INF 52 in which the type of the target character information INF 51 is converted into the second type. Specifically, the information processing apparatus 100A generates the converted character information INF 52 of the second type by inputting the target character information INF 51 of the first type into the model M2.

In the example of FIG. 11, the information processing apparatus 100A generates the converted character information INF 52 by the processing as shown in the processing group PS2. The information processing apparatus 100A inputs the target character information INF51 into the model M2 (step S63). The information processing apparatus 100A inputs the input information TIN2 which is the target character information INF51 into the model M2, and outputs the output information TOUT2 which is the second type conversion character information INF52 from the model M2. Specifically, the information processing apparatus 100A inputs the target character information INF51 into the partial model M21. Then, the information processing apparatus 100A inputs the intermediate information INT2 output by the partial model M21 to the partial model M22. Then, the information processing apparatus 100A causes the partial model M22 to output the converted character information INF52 (step S64). As a result, the information processing apparatus 100A generates the converted character information INF 52 in which the type of the target character information INF 51 of the first type is converted into the second type.

Then, the information processing apparatus 100A provides the converted character information INF 52 to the information providing apparatus 50 (step S65). In the example of FIG. 11, the information processing apparatus 100A transmits the converted character information INF 52 to the information providing apparatus 50.

As described above, the information processing apparatus 100A appropriately provides the service related to the character information conversion by providing the converted character information INF 52 in which the type of the target character information INF 51 is converted into the second type by using the model M2. be able to.

[6-3. Information processing device configuration]
Next, the configuration of the information processing apparatus 100A according to the modified example will be described with reference to FIG. FIG. 12 is a diagram showing a configuration example of the information processing apparatus 100A according to the modified example. As shown in FIG. 12, the information processing apparatus 100A includes a communication unit 110, a storage unit 120A, and a control unit 130A. Regarding the information processing apparatus 100A, the same points as those of the information processing apparatus 100 according to the embodiment will be omitted as appropriate.

(Memory unit 120A)
The storage unit 120A is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk. As shown in FIG. 11, the storage unit 120A according to the modified example has a first-type information storage unit 121, a second-type information storage unit 122, and a model information storage unit 123A.

(Model information storage unit 123A)
The model information storage unit 123A according to the modified example stores information about the model. For example, the model information storage unit 123A stores model information (model data) generated by information processing. FIG. 13 is a diagram showing an example of a model information storage unit according to a modified example. The model information storage unit 123A shown in FIG. 13 includes items such as "model ID", "partial model ID", "use", and "model data".

The "model ID" indicates identification information for identifying the model. For example, the model identified by the model ID "M2" corresponds to the model M2 shown in the example of FIG. The "partial model ID" indicates identification information for identifying a partial model of the corresponding model. For example, the partial model identified by the partial model ID "M21" corresponds to the partial model M21 shown in the example of FIG. "Use" indicates the use of the corresponding partial model. Further, "model data" indicates data of the corresponding partial model. For example, "model data" includes information including nodes in each layer, functions adopted by each node, connection relationships of the nodes, and connection coefficients set for connections between the nodes.

For example, in the example shown in FIG. 13, it is shown that the model M2 includes the partial models M21, M22, and the like. Further, for example, the model (partial model M21) identified by the partial model ID "M21" has a purpose of "character information ⇒ intermediate information" and is used to generate intermediate information from the input character information. show. That is, it is shown that the partial model M21 is used for extracting features from the input character information. Further, it is shown that the model data of the partial model M21 is the model data MDT21.

Further, for example, the model (partial model M22) identified by the partial model ID "M22" has a purpose of "intermediate information ⇒ second character information" and is used to generate second character information from the intermediate information. show. That is, it is shown that the partial model M22 is used to generate the second character information from the intermediate information which is the information indicating the characteristics of the extracted character information. Further, it is shown that the model data of the partial model M22 is the model data MDT22.

The model M2 (each model data MDT21, MDT22, etc.) is an input layer into which a predetermined type of character information is input, an output layer, and any layer from the input layer to the output layer, which is a layer other than the output layer. A first element belonging to the above and a second element whose value is calculated based on the weights of the first element and the first element are included, and the character information input to the input layer is applied to each layer other than the output layer. A computer so as to output the second type of character information corresponding to the input character information from the output layer by performing an operation based on the first element and the weight of the first element with each element to which it belongs as the first element. It is a model to make it work.

For example, the first element included in the model M2 corresponds to either node of the input layer or the intermediate layer. Further, the second element corresponds to the node of the next stage, which is the node to which the value is transmitted from the node corresponding to the first element. Further, the weight of the first element corresponds to a connection coefficient which is a weight considered for the value transmitted from the node corresponding to the first element to the node corresponding to the second element.

For example, the model identified by the model ID "M5" corresponds to the model M5 shown in the example of FIG. FIG. 10 shows that model M5 does not have a partial model. The model M5 shows that the purpose is "intermediate information ⇒ character information decoding", and it is used to decode the character information that is the generation base of the intermediate information from the intermediate information. That is, it is shown that the model M5 is used to decode the character information input to the partial model M21 from the intermediate information by using the intermediate information generated by the partial model M21 of the model M2 as an input. Further, it is shown that the model data of the model M5 is the model data MDT5.

The model information storage unit 123A is not limited to the above, and may store various model information depending on the purpose.

(Control unit 130A)
Returning to the description of FIG. 4, the control unit 130A is a controller, and is stored in a storage device inside the information processing device 100A by, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). Various programs (corresponding to an example of an information processing program such as a learning program and a generation program) are realized by executing the RAM as a work area. Further, the control unit 130A is a controller, and is realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). The control unit 130A has an input layer, an output layer, and an input layer to an output layer into which the first character information of the first type is input by information processing according to the model M2 or the like stored in the model information storage unit 123A. The input layer includes a first element which is any of the above layers and belongs to a layer other than the output layer, and a second element whose value is calculated based on the weights of the first element and the first element. With respect to the input first character information, each element belonging to each layer other than the output layer is set as the first element, and the first character information is second by performing an operation based on the weights of the first element and the first element. The character information converted to the type is output from the output layer.

As shown in FIG. 4, the control unit 130A has an acquisition unit 131, a learning unit 132A, a generation unit 133A, and a provision unit 134, and realizes or executes an information processing function or operation described below. .. The internal configuration of the control unit 130A is not limited to the configuration shown in FIG. 4, and may be any other configuration as long as it is configured to perform information processing described later. The second control unit 130A calculates a value based on the first element and the weight of the first element by information processing according to the model M2 (model data MDT21, MTD22, etc.) stored in the storage unit 120A. For the first character information of the first type input to the input layer including the element, each element belonging to each layer other than the output layer is set as the first element, and the weight of the first element and the first element is set. By performing the calculation based on this, the character information obtained by converting the first character information into the second type is output from the output layer.

The second control unit 130A calculates a value based on the first element and the weight of the first element by information processing according to the model M2 (model data MDT21, MTD22, etc.) stored in the storage unit 120A. By performing an operation based on the first element and the weight of the first element with each element belonging to each layer other than the output layer as the first element for the character information including the element and input to the input layer. The character information in which the type of the input character information is converted to the second type is output from the output layer. The model such as model M2 may be expected to be used as a program module which is a part of artificial intelligence software. A model such as model M2 may be a program.

(Acquisition unit 131)
The acquisition unit 131 according to the modified example performs the same processing as the acquisition unit 131 according to the embodiment. The acquisition unit 131 acquires various information in the same manner as the acquisition unit 131 according to the embodiment. For example, the acquisition unit 131 acquires the learning character information to be input to the conversion model.

In the example of FIG. 10, the acquisition unit 131 acquires the first character information as the learning data LD1. The acquisition unit 131 adds the first character information included in the learning data LD1 as learning data. The acquisition unit 131 adds the learning data LD1 to the first type information storage unit 121. The acquisition unit 131 acquires the second character information as the learning data LD2. The acquisition unit 131 adds the second character information included in the learning data LD2 as learning data. The acquisition unit 131 adds the learning data LD2 to the second type information storage unit 122.

In the example of FIG. 11, the acquisition unit 131 acquires character information for which conversion character information is to be generated from the information providing device 50 used by the information provider CP1. The acquisition unit 131 acquires the target character information INF 51, which is the character information of the first type, from the information providing device 50. The acquisition unit 131 acquires the conversion model. The acquisition unit 131 acquires the model M2 from the model information storage unit 123A.

(Learning Department 132A)
The learning unit 132A according to the modified example performs the same processing as the learning unit 132 according to the embodiment. The learning unit 132A learns various information in the same manner as the learning unit 132 according to the embodiment.

The learning unit 132A uses the first character information as input information, learning using the first character information as correct answer information, second character information as input information, and the second character information as correct answer information. Based on the learning, a conversion model is generated that outputs the converted character information in which the target character information is converted to the second type in response to the input of the target character information of the first type. The learning unit 132A has a coding unit that outputs intermediate information indicating the characteristics of the first character information in response to input of the first character information, and converted character information corresponding to the intermediate information using the intermediate information as input information. Generates a conversion model that includes a decoding unit that outputs.

The learning unit 132A generates a conversion model including a coding unit that outputs intermediate information indicating the characteristics of the second character information in response to the input of the second character information. When the second character information is input, the learning unit 132A generates a conversion model trained to output the second character information. When the first character information is input, the learning unit 132A generates a conversion model trained so that the first character information is converted into the second type. The learning unit 132A generates a conversion model by learning using another decoding unit that inputs intermediate information output by the decoding unit. In response to the input of the learning character information, the learning unit 132A inputs the intermediate information output by the decoding unit to another decoding unit, and performs learning processing so that the other decoding unit outputs the learning character information. By doing so, a transformation model is generated. When the learning character information is the first type, the learning unit 132A converts the learning character information into the second type, and when the learning character information is the second type, the conversion model converts the learning character information. A transformation model is generated by performing learning processing so that it is output.

The learning unit 132A generates a conversion model by a learning process using the first character information. The learning unit 132A generates the model M2 by performing learning using the first character information FCH11 to FCH14 and the like in the first type information storage unit 121 as learning data (teacher data). As shown in the first learning FLR2, the learning unit 132A is set with the type information FTL2 for selecting the first type. The learning unit 132A receives the input information FIN2 which is the first character information as an input, and performs learning to output the output information FOUT2 which is the second character information corresponding to the input information FIN2.

In the first learning FLR2, the learning unit 132A performs a learning process so that the second character information obtained by converting the first character information input to the model M2 into the second type is output from the model M2. When the first character information FCH 11 is input to the model M2, the learning unit 132A performs a learning process so that the second type character information corresponding to the first character information FCH 11 is output from the model M2. When the first character information FCH 12 is input to the model M2, the learning unit 132A performs a learning process so that the second type character information corresponding to the first character information FCH 12 is output from the model M2.

In the first learning FLR2, the learning unit 132A performs learning processing so that the number of terms (children's terms) for children included in the output information FOUT2 increases. The learning unit 132A may calculate the score of the output information FOUT2 by using the list information of the children's terms used in the article for children, and learn so that the score becomes high. The learning unit 132A may calculate the score so that the higher the ratio of the child terms contained in the output information FOUT2, the higher the score, and the learning unit 132A may perform learning so that the score of the output information FOUT2 becomes higher. The learning unit 132A performs learning processing by a method such as backpropagation (error back propagation method) so that the score of the output information FOUT2 becomes high.

The learning unit 132A performs learning processing using the model M5 in the first learning FLR2. The learning unit 132A performs learning processing so that the first character information input to the model M2 is output from the model M5. When the first character information FCH 11 is input to the model M2, the learning unit 132A performs a learning process so that the first character information FCH 11 is output from the model M5. When the first character information FCH12 is input to the model M2, the learning unit 132A performs a learning process so that the first character information FCH12 is output from the model M5.

In the first learning FLR2, the learning unit 132A performs learning processing using the partial model M21 and the model M5 of the model M2 as the autoencoder of the first character information. The learning unit 132A performs learning processing by a method such as backpropagation so that the character information output by the model M5 approaches the character information input to the model M2. The learning unit 132A performs learning processing on the model M2 and the model M5.

The learning unit 132A generates a conversion model by a learning process using the second character information. The learning unit 132A generates the model M2 by performing learning using the second character information SCH11 to SCH14 and the like in the second type information storage unit 122 as learning data (teacher data). The learning unit 132A learns to output the output information SOUT2 which is the second character information by inputting the input information SIN2 which is the second character information.

In the second learning SLR2, the learning unit 132A performs a learning process so that the second character information input to the model M2 is output from the model M2. When the second character information SCH11 is input to the model M2, the learning unit 132A performs a learning process so that the second character information SCH11 is output from the model M2. When the second character information SCH12 is input to the model M2, the learning unit 132A performs a learning process so that the second character information SCH12 is output from the model M2.

In the second learning SLR2, the learning unit 132A performs learning processing using the partial model M21 and the partial model M22 as an autoencoder of the second character information. The learning unit 132A performs learning processing by a method such as backpropagation so that the character information output by the model M2 approaches the character information input to the model M2. The learning unit 132A learns the characteristics of the second character information in the partial model M21 and the partial model M22 by modifying the value of the weight (that is, the connection coefficient) considered when the value is transmitted between the nodes. Let me.

The learning unit 132A performs learning processing using the model M5 in the second learning SLR2. The information processing apparatus 100A performs learning processing so that the second character information input to the model M2 is output from the model M5. When the second character information SCH11 is input to the model M2, the learning unit 132A performs a learning process so that the second character information SCH11 is output from the model M2. When the second character information SCH12 is input to the model M2, the learning unit 132A performs a learning process so that the second character information SCH12 is output from the model M2.

In the second learning SLR2, the learning unit 132A performs learning processing using the partial model M21 and the model M5 of the model M2 as the autoencoder of the second character information. The learning unit 132A performs learning processing by a method such as backpropagation so that the character information output by the model M5 approaches the character information input to the model M2.

(Generation unit 133A)
The generation unit 133A performs the same processing as the generation unit 133 according to the embodiment. The generation unit 133A provides various information in the same manner as the generation unit 133 according to the embodiment.

In the example of FIG. 11, the generation unit 133A uses the model M2 to generate the converted character information INF 52 obtained by converting the type of the target character information INF 51 into the second type. The generation unit 133A generates the conversion character information INF 52 of the second type by inputting the target character information INF 51 of the first type into the model M2.

The generation unit 133A generates the converted character information INF52 by the processing as shown in the processing group PS2. The generation unit 133A inputs the target character information INF51 into the model M2. The generation unit 133A inputs the input information TIN2 which is the target character information INF51 into the model M2, and outputs the output information TOUT2 which is the second type conversion character information INF52 from the model M2. The generation unit 133A inputs the target character information INF51 into the partial model M21. The generation unit 133A inputs the intermediate information INT2 output by the partial model M21 to the partial model M22. The generation unit 133A causes the partial model M22 to output the converted character information INF52. The generation unit 133A generates the converted character information INF 52 in which the type of the target character information INF 51 of the first type is converted into the second type.

(Providing section 134)
The providing unit 134 according to the modified example performs the same processing as the providing unit 134 according to the embodiment. The providing unit 134 provides various information in the same manner as the providing unit 134 according to the embodiment.

In the example of FIG. 11, the providing unit 134 provides the converted character information INF 52 to the information providing device 50. In the providing unit 134, the information processing device 100A transmits the converted character information INF 52 to the information providing device 50.

[7. effect〕
As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example have an acquisition unit 131 and learning units 132 and 132A. The acquisition unit 131 acquires the first character information which is the character information of the first type and the second character information which is the character information of the second type different from the first type. Further, the learning units 132 and 132A use the first character information as input information, the first character information as correct answer information, the second character information as input information, and the second character information as input information. Based on the learning as correct answer information, a conversion model is generated that outputs the converted character information in which the target character information is converted to the second type in response to the input of the target character information of the first type.

As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example are learned to use the first character information as input information and the first character information as correct answer information, and the second. Based on learning that uses the character information as input information and the second character information as correct answer information, the target character information is converted to the second type in response to the input of the target character information of the first type. By generating a conversion model that outputs character information, it is possible to appropriately provide services related to character information conversion.

Further, in the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example, the learning units 132 and 132A respond to the input of the first character information and provide intermediate information indicating the characteristics of the first character information. A conversion model including a coding unit for outputting the intermediate information and a decoding unit for outputting the converted character information corresponding to the intermediate information is generated.

As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example are encoded to output intermediate information indicating the characteristics of the first character information in response to the input of the first character information. By generating a conversion model including a unit and a decoding unit that outputs the converted character information corresponding to the intermediate information by using the intermediate information as input information, it is possible to appropriately provide a service related to character information conversion. ..

Further, in the information processing apparatus 100 according to the embodiment, the learning unit 132 inputs an intermediate information and a coding unit that outputs intermediate information indicating the characteristics of the second character information in response to the input of the second character information. As information, it is a conversion model including a decoding unit that outputs the first character information corresponding to the intermediate information, and the other target character information is the first in response to the input of the other target character information of the second type. Generate a conversion model that outputs other conversion character information converted to type.

As described above, the information processing apparatus 100 according to the embodiment has a coding unit that outputs intermediate information indicating the characteristics of the second character information in response to the input of the second character information, and the intermediate information as input information. It is a conversion model including a decoding unit that outputs the first character information corresponding to the intermediate information, and the other target character information is converted to the first type in response to the input of the other target character information of the second type. By generating a conversion model that outputs other converted character information, it is possible to appropriately provide a service related to character information conversion.

Further, in the information processing apparatus 100 according to the embodiment, the acquisition unit 131 acquires selection information for selecting the type of character information output by the conversion model. The learning unit 132 generates a conversion model that outputs character information of the type corresponding to the selection information.

As described above, the information processing apparatus 100 according to the embodiment acquires the selection information for selecting the type of character information output by the conversion model, and generates the conversion model for outputting the character information of the type corresponding to the selection information. Thereby, it is possible to appropriately provide a service related to character information conversion.

Further, in the information processing apparatus 100 according to the embodiment, the acquisition unit 131 acquires selection information for selecting a learning type, which is a type corresponding to the learning character information input to the conversion model. The learning unit 132 generates a conversion model by performing a learning process so as to output the learning character information based on the learning type in response to the input of the learning character information.

As described above, the information processing apparatus 100 according to the embodiment acquires the selection information for selecting the learning type which is the type corresponding to the learning character information input to the conversion model, and learns according to the input of the learning character information. By performing the learning process so as to output the learning character information based on the type, it is possible to appropriately provide the service related to the character information conversion by generating the conversion model.

Further, in the information processing apparatus 100 according to the embodiment, the learning unit 132 has a first encoder that outputs intermediate information indicating the characteristics of the first character information in response to an input of the first character information, and a second character information. In response to the input of, a conversion model including a coding unit having a second encoder that outputs intermediate information indicating the characteristics of the second character information is generated.

As described above, the information processing apparatus 100 according to the embodiment receives the input of the first character information, the first encoder that outputs the intermediate information indicating the characteristics of the first character information, and the input of the second character information. Accordingly, by generating a conversion model including a coding unit having a second encoder that outputs intermediate information indicating the characteristics of the second character information, it is possible to appropriately provide a service related to character information conversion. can.

Further, in the information processing apparatus 100 according to the embodiment, the learning unit 132 includes a first decoder that outputs the first character information corresponding to the intermediate information and a second decoder that outputs the second character information corresponding to the intermediate information. Generates a conversion model that includes a decoding unit with.

As described above, the information processing apparatus 100 according to the embodiment has a first decoder that outputs the first character information corresponding to the intermediate information and a second decoder that outputs the second character information corresponding to the intermediate information. By generating a conversion model including a conversion part, it is possible to appropriately provide a service related to character information conversion.

Further, in the information processing apparatus 100A according to the modification, the learning unit 132A generates a conversion model including a coding unit that outputs intermediate information indicating the characteristics of the second character information in response to the input of the second character information. do.

As described above, the information processing apparatus 100A according to the modified example generates a conversion model including a coding unit that outputs intermediate information indicating the characteristics of the second character information in response to the input of the second character information. , It is possible to appropriately provide services related to character information conversion.

Further, in the information processing apparatus 100A according to the modified example, when the second character information is input, the learning unit 132A generates a conversion model trained to output the second character information.

As described above, the information processing apparatus 100A according to the modified example relates to the character information conversion by generating a conversion model trained to output the second character information when the second character information is input. The service can be appropriately provided.

Further, in the information processing apparatus 100A according to the modification, the learning unit 132A generates a conversion model learned so that when the first character information is input, the first character information is converted into the second type. do.

As described above, the information processing apparatus 100A according to the modified example generates a conversion model trained so that when the first character information is input, the first character information is converted into the second type. , It is possible to appropriately provide services related to character information conversion.

Further, in the information processing apparatus 100A according to the modified example, the learning unit 132A generates a conversion model by learning using another decoding unit that inputs intermediate information output by the decoding unit.

As described above, the information processing apparatus 100A according to the modified example provides a service related to character information conversion by generating a conversion model by learning using another decoding unit that inputs intermediate information output by the decoding unit. It can be appropriately provided.

Further, in the information processing apparatus 100A according to the modified example, the acquisition unit 131 acquires the learning character information to be input to the conversion model. In response to the input of the learning character information, the learning unit 132A inputs the intermediate information output by the decoding unit to another decoding unit, and performs learning processing so that the other decoding unit outputs the learning character information. By doing so, a transformation model is generated.

As described above, the information processing apparatus 100A according to the modified example acquires the learning character information to be input to the conversion model, and in response to the input of the learning character information, the intermediate information output by the decoding unit is output to another decoding unit. By generating a conversion model by inputting to and performing a learning process so that another decoding unit outputs the learning character information, it is possible to appropriately provide a service related to character information conversion.

Further, in the information processing apparatus 100A according to the modification, when the learning character information is the first type, the conversion model converts the learning character information into the second type, and the learning character information is the second type. If, the conversion model is generated by performing the learning process so that the conversion model outputs the learning character information.

As described above, in the information processing apparatus 100A according to the modification, when the learning character information is the first type, the conversion model converts the learning character information into the second type, and the learning character information is the second type. By performing the learning process so that the conversion model outputs the learning character information, the conversion model can be generated, so that the service related to the character information conversion can be appropriately provided.

Further, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example have generation units 133 and 133A. The acquisition unit 131 has a conversion model that converts the first character information, which is the character information of the first type, into the second character information, which is the character information of the second type different from the first type, and the target character of the first type. Get information and. The generation unit 133, 133A generates the converted character information in which the target character information is converted into the second type by inputting the target character information into the conversion model.

As described above, in the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example, the first character information which is the character information of the first type is the character information of the second type different from the first type. The conversion model to be converted into the second character information and the target character information of the first type are acquired, and the target character information is input to the conversion model. As a result, the information processing devices 100 and 100A can generate the converted character information in which the target character information is converted into the second type, so that the service related to the character information conversion can be appropriately provided.

Further, in the information processing apparatus 100 according to the embodiment, the acquisition unit 131 acquires a conversion model for converting the second character information into the first character information and other target character information of the second type. By inputting other target character information into the conversion model, the generation unit 133 generates other converted character information in which the other target character information is converted into the first type.

As described above, the information processing apparatus 100 according to the embodiment acquires the conversion model for converting the second character information into the first character information and the other target character information of the second type. The generation unit 133 inputs other target character information into the conversion model. As a result, the information processing apparatus 100 can generate other converted character information in which other target character information is converted into the first type, so that it is possible to appropriately provide a service related to character information conversion. ..

Further, in the information processing apparatus 100 according to the embodiment, the acquisition unit 131 acquires selection information for selecting the type of character information output by the conversion model. The generation unit 133 generates character information of the type corresponding to the selection information by inputting the selection information into the conversion model.

As described above, the information processing apparatus 100 according to the embodiment acquires the selection information for selecting the type of the character information output by the conversion model, and inputs the selection information to the conversion model. As a result, the information processing apparatus 100 can generate character information of the type corresponding to the selection information, so that it is possible to appropriately provide a service related to character information conversion.

Further, in the information processing apparatus 100 according to the embodiment, the acquisition unit 131 is classified into either the first type or the second type of the target character information to be input to the conversion model, and is different from the type of the target character information. Acquires selection information for selecting the other type. The generation unit 133 generates the converted character information in which the target character information is converted into the other type by inputting the target character information and the selection information into the conversion model.

As described above, the information processing apparatus 100 according to the embodiment is classified into one of the target character information to be input to the conversion model and the first type or the second type, and the other type different from the type of the target character information. The selection information to be selected is acquired, and the target character information and the selection information are input to the conversion model. As a result, the information processing apparatus 100 can generate converted character information in which the target character information is converted into the other type, so that it is possible to appropriately provide a service related to character information conversion.

Further, in the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example, the first type and the second type are types in the same language.

As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example perform processing for the first type and the second type, which are the types in the same language, to obtain character information. It is possible to appropriately provide services related to conversion.

Further, in the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example, the first type is a news article and the second type is an article for children.

As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example are processed by processing the first type of news articles and the second type of articles for children. , It is possible to appropriately provide services related to character information conversion.

Further, in the information processing device 100 according to the embodiment and the information processing device 100A according to the modified example, the first type is character information including characters having the first number of characters or more, and the second type is the character information having the first number of characters or less. Character information including characters less than the second number of characters.

As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example have the first type to which the character information including the first character number or more belongs and the second character number less than the first character number. By performing processing for the second type to which the character information including the character of the above belongs, it is possible to appropriately provide the service related to the character information conversion.

Further, in the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example, the first type is the text and the second type is the summary.

As described above, the information processing apparatus 100 according to the embodiment and the information processing apparatus 100A according to the modified example process the first type which is the text of the article and the second type which is the summary of the article and the like. By doing so, it is possible to appropriately provide a service related to character information conversion.

[8. Hardware configuration]
The information processing apparatus 100 according to the above-described embodiment is realized by, for example, a computer 1000 having a configuration as shown in FIG. FIG. 14 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing device. The computer 1000 includes a CPU 1100, a RAM 1200, a ROM (Read Only Memory) 1300, an HDD (Hard Disk Drive) 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F). ) Has 1700.

The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400, and controls each part. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started, a program depending on the hardware of the computer 1000, and the like.

The HDD 1400 stores a program executed by the CPU 1100, data used by such a program, and the like. The communication interface 1500 receives data from another device via the network N and sends it to the CPU 1100, and transmits the data generated by the CPU 1100 to the other device via the network N.

The CPU 1100 controls an output device such as a display or a printer, and an input device such as a keyboard or a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. Further, the CPU 1100 outputs the generated data to the output device via the input / output interface 1600.

The media interface 1700 reads a program or data stored in the recording medium 1800 and provides the program or data to the CPU 1100 via the RAM 1200. The CPU 1100 loads the program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable Disk), a magneto-optical recording medium such as MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. And so on.

For example, when the computer 1000 functions as the information processing devices 100 and 100A according to the embodiment, the CPU 1100 of the computer 1000 is a program or data loaded on the RAM 1200 (for example, model M1 (each model data MDT11 to MDT14, etc.)). By executing the above, the functions of the control units 130 and 130A are realized. The CPU 1100 of the computer 1000 reads and executes these programs or data (for example, models M1 (for example, each model data MDT11 to MDT14)) from the recording medium 1800, but as another example, it is executed from another device via the network N. You may obtain these programs.

Although some of the embodiments and modifications of the present application have been described in detail with reference to the drawings, these are examples and various based on the knowledge of those skilled in the art, including the embodiments described in the disclosure line of the invention. It is possible to carry out the present invention in another form in which the above is modified or improved.

[9. others〕
Further, among the processes described in the above-described embodiments and modifications, all or part of the processes described as being automatically performed can be manually performed, or are described as being manually performed. It is also possible to automatically perform all or part of the performed processing by a known method. In addition, information including processing procedures, specific names, various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each figure is not limited to the information shown in the figure.

Further, each component of each of the illustrated devices is a functional concept, and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of distribution / integration of each device is not limited to the one shown in the figure, and all or part of them may be functionally or physically distributed / physically in any unit according to various loads and usage conditions. Can be integrated and configured.

Further, the above-described embodiments and modifications can be appropriately combined as long as the processing contents do not contradict each other.

Further, the above-mentioned "section, module, unit" can be read as "means" or "circuit". For example, the acquisition unit can be read as an acquisition means or an acquisition circuit.

1 Information processing system 100, 100A Information processing device 121 First type information storage unit 122 Second type information storage unit 123, 123A Model information storage unit 130, 130A Control unit 131 Acquisition unit 132, 132A Learning unit 133, 133A Generation unit 134 Provider 10 Terminal device 50 Information provider N Network

Claims (17)

An acquisition unit for acquiring the first character information which is the character information of the first type and the second character information which is the character information of the second type different from the first type.
When the first character information is input, intermediate information indicating the characteristics of the first character information is output, and when the second character information is input, the intermediate information indicating the characteristics of the second character information is output. A conversion model including a coding unit to be output and a decoding unit to output converted character information which is character information corresponding to a specified designated type by inputting the intermediate information is input with the first character information. At the time of learning processing using the first character information as the first correct answer information, the designated type is set as the first type, and learning is performed so that the output of the conversion model becomes the first correct answer information, and the second character information. In the learning process in which the second character information is used as the second correct answer information, the designated type is set as the second type and the output of the conversion model is learned to be the second correct answer information. , The learning part to generate,
An information processing device characterized by being equipped with. The acquisition unit
Acquire the selection information for selecting the specified type , and
The information processing apparatus according to claim 1 , wherein the decoding unit of the conversion model receives the selection information as information indicating the designated type . The acquisition unit
When the learning character information which is the character information to be input to the conversion model is the first type, the first type is selected, and the second type where the learning character information is the second type is selected. Get information,
The learning unit
The information processing apparatus according to claim 2 , wherein the conversion model is generated by inputting the learning character information and performing a learning process so as to output the learning character information. The learning unit
The first encoder that outputs the intermediate information indicating the characteristics of the first character information in response to the input of the first character information, and the characteristics of the second character information in response to the input of the second character information. The information processing apparatus according to any one of claims 1 to 3 , wherein the conversion model including the coding unit including the second encoder for outputting the intermediate information shown is generated. The learning unit
Generate the conversion model including the decoding unit having the first decoder that outputs the first character information corresponding to the intermediate information and the second decoder that outputs the second character information corresponding to the intermediate information. The information processing apparatus according to any one of claims 1 to 4 , wherein the information processing apparatus is to be used. An acquisition unit for acquiring the first character information which is the character information of the first type and the second character information which is the character information of the second type different from the first type.
When the first character information is input, intermediate information indicating the characteristics of the first character information is output, and when the second character information is input, the intermediate information indicating the characteristics of the second character information is output. The first character information is a conversion model including a coding unit to be output and a decoding unit to output converted character information which is character information corresponding to a specified designated type by inputting the intermediate information. In the learning process in which the first character information is used as the first correct answer information, the designated type is set as the first type, and the intermediate information output by the coding unit is input as another decoding unit. Therefore, the output of the coding unit other than the conversion model is learned to be the first correct answer information, the second character information is input, and the second character information is the second correct answer information. At the time of learning processing, the designated type is set as the second type, and the learning unit generated by learning so that the output of the conversion model becomes the second correct answer information .
An information processing device characterized by being equipped with. The acquisition unit
Acquire the type information indicating the specified type, and
The conversion model accepts the type information as information indicating the designated type.
The information processing apparatus according to claim 6. The learning unit
When the first character information is input, the conversion model trained so that the converted character information to which the first character information is converted includes many terms corresponding to the second type is generated. The information processing apparatus according to claim 6 or 7 . The acquisition unit
Acquire the learning character information to be input to the conversion model,
The learning unit
In response to the input of the learning character information, the intermediate information output by the coding unit is input to the other decoding unit, and the learning process is performed so that the other decoding unit outputs the learning character information. The information processing apparatus according to claim 6 or 7 , wherein the conversion model is generated by performing the above. The information processing apparatus according to any one of claims 1 to 9 , wherein the first type and the second type are types in the same language. The information processing apparatus according to any one of claims 1 to 10 , wherein the first type is a news article, and the second type is an article for children. The first type is character information including characters having a first number of characters or more, and the second type is character information including characters having a number of characters less than or equal to the first number of characters. The information processing apparatus according to any one of 1 to 11 . The information processing apparatus according to any one of claims 1 to 12 , wherein the first type is a text and the second type is a summary. It is an information processing method executed by a computer.
An acquisition process for acquiring first character information, which is character information of the first type, and second character information, which is character information of the second type different from the first type.
When the first character information is input, intermediate information indicating the characteristics of the first character information is output, and when the second character information is input, the intermediate information indicating the characteristics of the second character information is output. A conversion model including a coding unit to be output and a decoding unit to output converted character information which is character information corresponding to a specified designated type by inputting the intermediate information is input with the first character information. At the time of learning processing using the first character information as the first correct answer information, the designated type is set as the first type, and learning is performed so that the output of the conversion model becomes the first correct answer information, and the second character information. In the learning process in which the second character information is used as the second correct answer information, the designated type is set as the second type and the output of the conversion model is learned to be the second correct answer information. , The learning process to generate,
An information processing method characterized by including. An acquisition procedure for acquiring the first character information which is the character information of the first type and the second character information which is the character information of the second type different from the first type.
When the first character information is input, intermediate information indicating the characteristics of the first character information is output, and when the second character information is input, the intermediate information indicating the characteristics of the second character information is output. A conversion model including a coding unit to be output and a decoding unit to output converted character information which is character information corresponding to a specified designated type by inputting the intermediate information is input with the first character information. At the time of learning processing using the first character information as the first correct answer information, the designated type is set as the first type, and learning is performed so that the output of the conversion model becomes the first correct answer information, and the second character information. In the learning process in which the second character information is used as the second correct answer information, the designated type is set as the second type and the output of the conversion model is learned to be the second correct answer information. , The learning procedure to generate,
An information processing program characterized by having a computer execute. It is an information processing method executed by a computer.
An acquisition process for acquiring first character information, which is character information of the first type, and second character information, which is character information of the second type different from the first type.
When the first character information is input, intermediate information indicating the characteristics of the first character information is output, and when the second character information is input, the intermediate information indicating the characteristics of the second character information is output. A conversion model including a coding unit to be output and a decoding unit to output converted character information which is character information corresponding to a specified designated type by inputting the intermediate information is input with the first character information. In the learning process in which the first character information is the first correct answer information, the designated type is the first type, and the intermediate information output by the coding unit is input as another decoding unit. At the time of learning processing in which the output of the coding unit other than the conversion model is learned to be the first correct answer information, the second character information is input, and the second character information is the second correct answer information. Is a learning process generated by learning so that the output of the conversion model becomes the second correct answer information, with the designated type as the second type.
An information processing method characterized by including. An acquisition procedure for acquiring the first character information which is the character information of the first type and the second character information which is the character information of the second type different from the first type.
When the first character information is input, intermediate information indicating the characteristics of the first character information is output, and when the second character information is input, the intermediate information indicating the characteristics of the second character information is output. A conversion model including a coding unit to be output and a decoding unit to output converted character information which is character information corresponding to a specified designated type by inputting the intermediate information is input with the first character information. In the learning process in which the first character information is the first correct answer information, the designated type is the first type, and the intermediate information output by the coding unit is input as another decoding unit. At the time of learning processing in which the output of the coding unit other than the conversion model is learned to be the first correct answer information, the second character information is input, and the second character information is the second correct answer information. Is a learning procedure generated by learning so that the output of the conversion model becomes the second correct answer information with the designated type as the second type.
An information processing program characterized by having a computer execute.

Images