JP7170299B2 - Information processing system, information processing method and program - Google Patents

Description

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

2. Description of the Related Art Conventionally, techniques for generating sentences using artificial intelligence (AI) are known (for example, Patent Literature 1, etc.).

In recent years, AI technology for writing novels has been attracting attention as a novel written by AI has passed the selection process for a prestigious award. In this technique, for example, a method is used in which the structure of a novel and various attributes appearing in the novel are manually set in advance, and the generation of sentences is automated by AI. For example, a method is known in which logs of games played by a plurality of players are recorded to create scenarios, and scenarios that humans find interesting are selected and written as novels by humans.

Further, there is known a technique for generating a word list of a plurality of words from a vector indicating the association probability of associated colors (hereinafter referred to as "color vector") (for example, Patent Documents 2 and 3). etc).

JP 2017-102328 A Patent No. 5618150 Patent No. 5344756

However, with the conventional technology, there is a problem that sentences cannot be generated from an image representing a sound source or a product concept, or an intuitive image represented by colors or an adjective pair evaluation scale.

One aspect of the present invention has been made in view of such problems, and generates sentences, etc., from images that indicate sound sources or product concepts, etc., and intuitive images that are indicated by colors, adjectives, evaluation scales, etc. for the purpose.

In order to solve the above problems, the information processing system
an acquisition unit for acquiring an adjective pair rating scale, which is a rating scale entered by a user for an image, color or adjective pair;
a list generation unit that generates at least one of a word list and a phonetic symbol word list based on the image, the color, or the adjective pair evaluation scale acquired by the acquisition unit;
Text generation for selecting a plurality of words or phonetic symbolic words from at least one of the word list and the phonetic symbolic word list according to a predetermined condition, and generating output text including the selected words or phonetic symbolic words. including the part.

A sentence or the like can be generated from an image representing a sound source or a product concept, or an intuitive image represented by a color or an adjective paired with an evaluation scale or the like.

1 is a diagram illustrating an example of the overall configuration of an information processing system according to a first embodiment; FIG. It is a block diagram which shows the hardware structural example of an information processing apparatus. 4 is a flowchart showing an example of overall processing in the first embodiment; FIG. 11 is a diagram (part 1) showing an example of an input screen; FIG. 4 is a diagram showing an example of data used to generate a word list; FIG. FIG. 4 is a diagram showing an example of data used for generating a phonetic symbol word list; FIG. 11 is a diagram (part 2) showing an example of an input screen; It is a figure which shows the whole structural example of the information processing system in 2nd Embodiment. 9 is a flowchart showing an example of overall processing in the second embodiment; FIG. 11 is a flowchart showing an example of overall processing in the third embodiment; FIG.

Specific examples of preferred embodiments of the present invention will be described below.

<First Embodiment>
<Overall configuration example>
FIG. 1 is a diagram showing an example of the overall configuration of an information processing system according to the first embodiment. As illustrated, the information processing system 1 has one or more information processing devices. Specifically, the information processing system 1 includes a server 2, which is an example of an information processing device, and a PC (Personal Computer) 3, which is an example of an information processing device.

As illustrated, the server 2 and the PC 3 are connected by a network or the like, and can exchange data with each other.

Note that the server 2 and the PC 3 may be integrated, that is, the information processing system 1 may be configured by one information processing device. On the other hand, the information processing system 1 may have three or more information processing devices. The illustrated overall configuration will be described below as an example.

<Hardware configuration example>
FIG. 2 is a block diagram showing a hardware configuration example of an information processing apparatus. For example, the server 2 and PC3 have the same hardware configuration. Hereinafter, the server 2 will be described as an example, and the description of the PC 3 will be omitted.

The server 2 has a CPU (Central Processing Unit) 2H01, a storage device 2H02, an input device 2H03, an output device 2H04, and an interface 2H05. Each hardware resource is connected by a bus and exchanges data with each other.

CPU2H01 is an example of an arithmetic device and a control device. The storage device 2H02 is a main storage device or the like. Also, the storage device 2H02 may be an auxiliary storage device such as an SSD (Solid State Drive) or a hard disk. The input device 2H03 is a device for inputting an operation from the user UR. For example, the input device 2H03 is a keyboard, mouse, or combination thereof. The output device 2H04 is a device that displays processing results and the like to the user UR. For example, the output device 2H04 is a display or the like. The interface 2H05 is a device for inputting/outputting data with an external device. For example, the interface 2H05 is a connector, communication device, or the like.

Note that the hardware configuration is not limited to the illustrated configuration. For example, the hardware configuration may be a configuration in which an arithmetic device, a control device, or the like is further inside or outside.

<Overall processing example>
FIG. 3 is a flow chart showing an example of overall processing in the first embodiment.

<Example of Acquisition of Adjective Versus Rating Scale>
In step S01, the information processing system 1 acquires the adjective pair evaluation scale ADJ by the operation of the user UR on the input screen. For example, the information processing system 1 first displays the following input screen on the screen of the PC 3 .

FIG. 4 is a diagram (part 1) showing an example of an input screen.

For example, using the input screen PNL1, the user UR can input a rating scale for a given adjective pair PAD using a UI (User Interface) such as a scroll bar. Specifically, the illustrated adjective pair PAD is an example of a pair of adjectives indicating opposite meanings of brightness, such as "bright" and "dark." For example, as described in Japanese Patent No. 5678836, integer values of several stages are set in advance for the adjective pair PAD, and when the user UR inputs the integer value through the UI, the information processing system 1 , an adjective pair rating scale ADJ can be obtained for each adjective pair.

Note that adjective pairs other than the adjective pairs shown on the input screen PNL1 may be used. Also, the adjective pair to be used may be selected by a UI such as a check box indicated by the input screen PNL1. Furthermore, the adjective pair evaluation scale ADJ is not limited to an integer value, and a certain value after the decimal point may be used.

<Calculation example of the first color vector>
In step S02, the information processing system 1 calculates a vector (hereinafter referred to as "first color vector") indicating the association probability of each color associated with the obtained adjective pair evaluation scale ADJ. Note that the color may be a single color or a color in which a plurality of colors are mixed.

Specifically, in an experiment with a total of 120 subjects, when 20 people associate the color “red”, the association probability of “red” is “20 people / 120 people ≒ 0.17 ”. And the color vector is like "color vector = (association probability of black, association probability of brown, association probability of red, ..., association probability of blue, association probability of green, association probability of white)". show.

Therefore, in the case of calculating a color vector corresponding to an adjective paired with a rating scale, an experiment in which a subject responds to a color associated with an adjective pair provides data that associates the adjective paired with the rating scale with the color vector. prepared in advance. With such data, the information processing system 1 can calculate the first color vector from the adjective versus the rating scale ADJ.

<Generation example of word list, etc.>
In step S03, the information processing system 1 selects words W1 stored in advance in the database DBS based on the calculated first color vector, and generates a word list LI1.

In the database DBS, a word W1 and a color vector corresponding to the word W1 (hereinafter referred to as a "second color vector") are linked and stored in advance through experiments or the like.

FIG. 5 is a diagram showing an example of data used to generate a word list. For example, the database DBS stores a color table T1 and an association probability table T2 as shown.

In the color table T1, candidate colors Ci (data indicating) are associated with color IDs that can uniquely identify the candidate colors Ci. Note that the suffix i employs, for example, an integer having a value from "0" to "C-1".

In the association probability table T2, association probabilities p(ci|wj) are associated with combinations of candidate colors Ci and words wj. The association probability p(ci|wj) is the association probability of the candidate color Ci associated with the subject when the word wj is given. Therefore, the second color vector is a value calculated by dividing the number of responses of the color associated with the subject by the total number of responses when a predetermined word is shown in advance in the experiment.

As described above, the information processing system 1 calculates the degree of similarity between the first color vector calculated according to the adjective pair evaluation scale ADJ and the second color vector corresponding to the word W1, and generates a word list. It is desirable to generate LI1. Also, the degree of similarity is calculated using, for example, a cosine scale.

The information processing system 1 may similarly generate the phonetic symbol word list. In this case, in the database DBS, the phonetic symbolic word and the second color vector corresponding to the phonetic symbolic word are linked and stored by an experiment conducted in advance.

FIG. 6 is a diagram showing an example of data used to generate a phonetic symbol word list. For example, the database DBS stores an association probability table T3 as shown. Similar to the association probability table T2, association probabilities p(ci|wj) are associated with combinations of phonetic symbol words and candidate colors Ci.

<Example of selecting a plurality of words or phonetic symbols from a word list or a phonetic symbolic word list according to a predetermined condition to generate an output text>
At step S04, the information processing system 1 generates the output text DOUT based on the word list LI1 and the like generated at step S03.

The predetermined condition is, for example, a word selection method based on similarity, and is set in the information processing system 1 in advance. Specifically, a predetermined condition is set in the information processing system 1, such as selecting words with a high degree of similarity with a high probability. Alternatively, the predetermined condition may be, for example, to select only words with a degree of similarity equal to or greater than a predetermined value.

The output text DOUT is desirably generated by the information processing system 1 acquiring the input text DIN and converting the words included in the input text DIN, for example, as follows.

FIG. 7 is a diagram (part 2) showing an example of the input screen. In this example, it is assumed that an input text DIN acquired by the information processing system 1 is first displayed on the input screen PNL2. For example, the input text DIN is input in text data format or the like. Note that the input text DIN may be input by the user UR directly inputting characters and the like on the screen as shown.

Also, in this example, when the image input button BIMG is pressed, the information processing system 1 can acquire the image IMG and the like. Note that the image IMG is not limited to being acquired by the information processing system 1 by operating the image input button BIMG or the like, and may be acquired by the information processing system 1 by, for example, a drag-and-drop operation. The image IMG obtained in this way is data representing the image of the text generated as the output text DOUT.

Specifically, in product development and the like, when considering a sentence indicating an image of a product to be developed, the user UR inputs to the information processing system 1 an image IMG indicating a product concept or a scene in which the product is used.

Next, the information processing system 1 analyzes the acquired image IMG. For example, the information processing system 1 extracts the colors used in the image IMG, or calculates and ranks the areas of the colors used in the image IMG. In this way, the information processing system 1 can calculate the color vector from the acquired image IMG. Note that the image IMG may be subjected to processing such as removal of the background color. Alternatively, the information processing system 1 may recognize the subject appearing in the image IMG by image recognition or the like, and specify the type of the subject.

Further, as shown in the figure, when the input text DIN is acquired, the information processing system 1 sets the rate at which the words included in the input text DIN are converted (hereinafter referred to as "first conversion rate P1"). Input is preferred. The illustrated example is an example in which a first conversion rate P1 is input by a conversion rate input bar SRB1. When the first conversion rate P1 is set high, the input text DIN is converted at a high rate. That is, when the first conversion rate P1 is high, the information processing system 1 increases the percentage of words to be converted among the words included in the input text DIN. In the illustrated example, "100 percent" is input as the first conversion rate P1 through the conversion rate input bar SRB1, so the information processing system 1 is set to convert all words included in the input text DIN. In this way, when the first conversion rate P1 is input to the information processing system 1, the user UR can set the rate for how much the input text DIN is to be converted.

Further, as shown in the figure, the information processing system 1 converts the words included in the input text DIN into phonetic symbolic words such as onomatopoeia or converts them into words W1 (hereinafter referred to as "second conversion rate P2"). ) should be entered. In the illustrated example, like the first conversion rate P1, the second conversion rate P2 is an example input by the onomatopoeia rate input bar SRB2. In the illustrated example, the onomatopoeia rate input bar SRB2 is used to input the onomatopoeia rate of "35%", which is an example of the second conversion rate P2. , converted to onomatopoeia.

Furthermore, the number of characters CNM and the number of times the same word can be used in conversion may be specified in the selection interface SEL or the like. Mainly, when the input text DIN does not exist and the number of characters CNM is designated, the information processing system 1 may be configured to designate the number of characters of the entire output text DOUT by the number of characters CNM. Similarly, when the input text DIN does not exist, the information processing system 1 selects words or phonetic symbol words with the same number of characters as the specified number of characters CNM or words or phonetic symbol words with the specified number of characters CNM or more, They may be combined to produce the output text DOUT.

Note that when the input text DIN contains a plurality of the same words, all of the words may be converted into the same word or into different words. Specifically, when the input text DIN contains two or more words "A", the first "A" is "B" and the second "A" is "C". and so on, or both the first "A" and the second "A" may be converted to "B". When converting into different words, for example, the information processing system 1 randomly selects and converts a plurality of words with high similarities among the words in the word list LI1.

The information processing system 1 may also morphologically analyze the input text DIN. By morphologically analyzing the input text DIN, the information processing system 1 can identify the part of speech of each word included in the input text DIN.

Next, the information processing system 1 desirably converts a word included in the input text DIN from the analysis result of the morphological analysis if the part of speech is a verb, adjective, noun, adverb, or adnominal. It is assumed that the part of speech to be converted can be selected by the user UR through the selection interface SEL or the like. Specifically, the information processing system 1 sets the part of speech selected by the user UR with a check box UI or the like on the selection interface SEL so as to be the conversion target. Note that the method of setting the part of speech to be converted is not limited to the method shown in the figure.

On the other hand, particles are preferably excluded from conversion targets. Specifically, even though the first conversion rate P1 is "100 percent" as in the illustrated example, the case particle "wo" is not converted and is maintained in the output text DOUT. On the other hand, in the illustrated setting, the adverb "also", the adjective "big", the noun "computer", and the verb "bought" are displayed in the selection interface SEL. Because it is a part of speech selected by , it is subject to conversion. Then, the information processing system 1 converts the word to be converted into another word of the same part of speech to generate the output text DOUT. In this way, the information processing system 1 preferably analyzes and converts the parts of speech of the words included in the input text DIN.

The information processing system 1 may also analyze the semantic relationship between words included in the output text DOUT using a thesaurus. Then, the information processing system 1 may select words from the word list such that the semantic relationships are established. For example, when selecting a predicate word, the information processing system 1 selects a predicate that makes the output text DOUT a sentence, depending on the type of subject. Specifically, when the subject is a word that refers to a plant, the information processing system 1 selects a word that refers to a plant from the word list instead of a term that refers to an animal. In this way, the information processing system 1 may select a word from the word list or a phonetic symbolic word from the phonetic symbolic word list so that the semantic relationship between the subject and the predicate is appropriate.

Furthermore, the information processing system 1 may calculate the color vector VCT of the output text DOUT. For example, when the output text DOUT contains a plurality of words, the information processing system 1 normalizes the result of multiplying the color vector of each word, the frequency of appearance of each word, and the degree of influence of each word. to compute the color vector VCT of the output text DOUT.

The frequency is the ratio of the number of occurrences of the target word to the total number of words included in the output text DOUT.

The degree of influence is the degree of influence given to associating a color with a word. Also, the degree of influence indicates the strength of the connection with the color. Therefore, the degree of influence is the proportion of subjects who answered that they were likely to associate a color with a certain word. Specifically, in an experiment conducted in advance with a total of 20 subjects, when 10 subjects answered that they associated "rain" with a text containing "rain", the degree of influence of "rain" was It becomes "10÷20=0.5".

In an example in which the output text DOUT includes three words “sea”, “sun” and “summer”, the information processing system 1 first determines the “colors” of “sea”, “sun” and “summer”. Calculate "vector x frequency x impact". Next, the information processing system 1 adds the calculation result of “sea”, the calculation result of “sun”, and the calculation result of “summer”, divides by “3” which is the total number of words, and normalizes the result. , compute the color vector VCT of the output text DOUT.

Then, the information processing system 1 calculates the degree of similarity between the first color vector calculated from the adjective pair evaluation scale or the like and the color vector VCT of the output text DOUT. In order to increase the degree of similarity calculated in this way, the information processing system 1 may select words included in the word list or phonetic symbolic words included in the phonetic symbolic word list to generate the output text DOUT. good.

<Example of functional configuration>
As illustrated in FIG. 1, the information processing system 1 has a functional configuration including an acquisition unit F1, a list generation unit F2, and a text generation unit F3. In addition, as illustrated, the information processing system 1 further includes a text acquisition section F4, a designation section F5, a first conversion rate input section F6, a second conversion rate input section F7, and a calculation section F8. A configuration is desirable. That is, in the functional configuration shown in FIG. 1, the information processing system 1 includes the text acquisition unit F4, the designation unit F5, the first conversion rate input unit F6, the second conversion rate input unit F7, and the calculation unit F8 as essential functional configurations. do not do. The illustrated functional configuration will be described below as an example.

The obtaining unit F1 performs an obtaining procedure for obtaining the adjective-pair rating scale ADJ, the adjective-pair rating scale ADJ from the color CLR, the image IMG, or the like. For example, the acquisition unit F1 is realized by the input device 2H03 or the like.

The list generation unit F2 performs a list generation procedure for generating the word list LI1 based on the adjective pair evaluation scale ADJ acquired by the acquisition unit F1. Note that the list generation unit F2 may generate a phonetic symbol word list. For example, the list generator F2 is realized by the CPU2H01 and the like.

The text generation unit F3 selects a plurality of words satisfying a predetermined condition from the word list LI1 generated by the list generation unit F2. Then, the text generation unit F3 performs a text generation procedure for generating the output text DOUT using the selected words. For example, the text generator F3 is realized by the CPU2H01 or the like.

The text acquisition unit F4 performs a text acquisition procedure for acquiring the input text DIN. Thus, when the text acquisition unit F4 acquires the input text DIN, the text generation unit F3 converts the words included in the input text DIN into the words indicated by the word list generated by the list generation unit F2 and outputs the words. Generate the text DOUT. For example, the text acquisition unit F4 is realized by the input device 2H03 or the like.

The designation unit F5 performs a designation procedure for designating the number of characters CNM. Thus, when the number of characters CNM is specified by the specification unit F5, the text generation unit F3 converts the words included in the input text DIN to the number of characters CNM to generate the output text DOUT. For example, the designation unit F5 is realized by the input device 2H03 or the like.

The first conversion rate input unit F6 performs a first conversion rate input procedure for inputting the first conversion rate P1. As described above, when the first conversion rate P1 is input by the first conversion rate input unit F6, the text generation unit F3 selects the words indicated by the first conversion rate P1 among the words included in the input text DIN. Generate the output text DOUT such that the percentage words are transformed. For example, the first conversion rate input unit F6 is realized by the input device 2H03 or the like.

The second conversion rate input unit F7 performs a second conversion rate input procedure for inputting a second conversion rate P2. In this way, when the second conversion rate P2 is input by the second conversion rate input unit F7, the text generation unit F3 converts the words to be converted into the ratio indicated by the second conversion rate P2. The output text DOUT is generated by converting into phonetic symbols and words as follows. For example, the second conversion rate input unit F7 is realized by the input device 2H03 or the like.

A calculator F8 calculates the color vector VCT of the output text DOUT. Then, the text generation unit F3 increases the similarity between the color vector VCT of the output text DOUT calculated by the calculation unit F8 and the color vector calculated based on the adjective pair evaluation scale ADJ acquired by the acquisition unit F1. to generate the output text DOUT. For example, the calculation unit F8 is implemented by the CPU2H01 or the like.

<Second embodiment>
In the second embodiment, the information processing system 1 generates an output text DOUT containing phonetic symbols.

<Overall configuration example>
FIG. 8 is a diagram showing an example of the overall configuration of an information processing system according to the second embodiment. Compared with the first embodiment, the list generation unit F2 is different in calculating the adjective pair evaluation scale from the color CLR or the image IMG acquired by the acquisition unit F1. Then, the list generation unit F2 selects the phonetic symbol word W2 and generates the phonetic symbol word list LI2 based on the calculated adjective pair evaluation scale. In the following, different configurations will be mainly described, and overlapping descriptions will be omitted.

Further, if the color CLR is previously associated with the adjective-pair rating scale in an experiment, the adjective-pair rating scale (hereinafter referred to as "first adjective-pair rating scale") obtained from the color CLR or the image IMG acquired by the acquisition unit F1. can be calculated.

For the phonetic symbol word W2, the adjective pair evaluation scale can be calculated by the method described in Japanese Patent No. 5354425, for example. Specifically, first, if the phoneme morphology database is available, the phonetic symbol word W2 can be analyzed to generate phoneme notation information. Next, if a quantitative image table is prepared in advance for each phoneme form indicated by the phoneme notation information, each adjective pair evaluation scale (hereinafter referred to as "second adjective pair evaluation scale") can be specified. Based on the respective adjective pair rating scale for each phonological form thus identified, a second adjective pair rating scale corresponding to phonetic symbol word W2 can be pre-computed. Therefore, phonetic symbol word W2 can be associated with a second adjective pair rating scale in database DBS.

Then, it is preferable that the list generator F2 calculates the degree of similarity between the first adjective-pair rating scale and the second adjective-pair rating scale. The degree of similarity is calculated using, for example, a cosine scale.

<Overall processing example>
FIG. 9 is a flow chart showing an example of overall processing in the second embodiment.

<Example of acquiring an image or color>
In step S11, the information processing system 1 acquires the image IMG or the color CLR by the operation of the user UR on the input screen. For example, the image IMG is acquired by the input screen PNL2 or the like, as in the first embodiment. Also, the color CLR is acquired by an operation screen for specifying a color, an operation for inputting RGB values, or the like.

<Calculation example of adjective versus rating scale according to image or color>
In step S12, the information processing system 1 calculates a first adjective pair evaluation scale according to the acquired image IMG or color CLR.

<Generation example of phonetic symbol word list>
In step S13, the information processing system 1 selects a plurality of phonetic symbol words W2 and generates a phonetic symbol word list LI2 based on the calculated first adjective pair evaluation scale.

<Example of generating output text by selecting multiple data from the phonetic symbol word list according to predetermined conditions>
In step S14, the information processing system 1 selects a phonetic symbolic word from the phonetic symbolic word list according to a predetermined condition set in advance to generate an output text. Note that the predetermined condition is, for example, the same as in the first embodiment.

<Modification>
If there is an unknown word, the degree of influence, color vector, etc. of the unknown word may be estimated. For example, the unknown word influence is a value calculated by multiplying the primitive word influence by the similarity. On the other hand, the unknown word color vector is a value calculated by multiplying the primitive word color vector by the similarity.

Primitive words are words associated with colors.

Unknown words are words other than primitive words that appear in the corpus.

Specifically, first, assume that the word "swimming" is an unknown word and the primitive words are "sea" and "summer". It is also assumed that the degree of similarity between “swimming” and “sea” is “0.7” and the degree of similarity between “swimming” and “summer” is “0.9”. The degree of similarity between primitive words and unknown words is cosine similarity or the like based on latent semantic analysis (LSA).

In this example, the influence of "swimming" is the value obtained by multiplying the influence of "sea" by the similarity of "0.7", and the influence of "summer" by the similarity of "0.9". multiplied by , and divided by "2", which is the total number of primitive words related to "swimming".

The color vector of "swimming" is obtained by multiplying the color vector of "sea" by the similarity of "0.7" and the color vector of "summer" by the similarity of "0.9". is added and divided by "2".

<Third Embodiment>
In the third embodiment, the information processing system 1 generates the output text DOUT based on a recurrent neural network (hereinafter referred to as "RNN"). An example in which RNN processing is applied to the overall processing shown in FIG. 9 will be described below.

<Overall processing example>
FIG. 10 is a flowchart showing an example of overall processing in the third embodiment. Compared with FIG. 9, the illustrated process differs in that step S14 is replaced with step S21. In the following, the points different from the second embodiment will be mainly described.

<Example of output text generated by RNN>
In step S21, the information processing system 1 selects a plurality of pieces of data from the phonetic symbol word list or the like by RNN and generates an output text.

Specifically, first, in step S13, the words "red", "apple" and "Christmas" are extracted and listed. In this case, the information processing system 1 uses the RNN to generate the output text "Eating red apples for Christmas."

Using the RNN, the information processing system 1 can learn how to use specific words in the learning data. Therefore, for example, when generating lyrics, the information processing system 1 learns how to use words unique to the lyrics, You can predict the probability of appearance.

In addition, the use of RNN enables the information processing system 1 to learn in consideration of past states. Therefore, compared to the case of using a so-called normal neural network that performs feed forward processing, the use of RNN allows the information processing system 1 to predict the meaning of words in a time-series flow. , can be classified. In particular, the information processing system 1 can predict the words with high accuracy when the next words are different in terms of context.

Additionally, RNNs allow for variable length inputs and outputs.

Note that application of the RNN is not limited to the overall processing. That is, various known machine learning methods that can be used for creating sentences other than RNN can be applied to the overall processing.

<Summary>
As described above, for example, when the image IMG, the color CLR, or the adjective pair rating scale ADJ can be obtained, the information processing system 1 can obtain the word W1 or the sound symbol that matches the image indicated by the image IMG, the color CLR, or the adjective pair rating scale ADJ. Word W2 can be identified. Thus, the information processing system 1 generates the output text DOUT using the specified word W1 or phonetic symbol word W2.

Also, the output text DOUT may be lyrics or the like. In such a case, an image IMG or the like representing the image of the song to be composed is input to the information processing system 1 .

In this way, the information processing system 1 can generate the output text DOUT indicating the advertisement copy or the like from the image IMG, the color CLR, or the adjective pair evaluation scale ADJ, from an abstract image such as the scene of use of the product or the concept of the product.

The information processing system 1 may generate and output sentences such as novels or comics. Also, the information processing system 1 may be applied to a dialogue system such as a chat bot.

<Other embodiments>
Note that the overall processing is not limited to the illustrated order. For example, part or all of each process may be performed in parallel or in an order different from the order shown. Also, each process may be executed by a plurality of information processing apparatuses redundantly, distributed, in parallel, virtualized, or in combination thereof.

Note that the information processing method according to the present invention may be implemented by a program. That is, the program is a computer program for causing a computer, such as an information processing system including one or more information processing devices, to execute each process.

Therefore, when the information processing method is executed based on the program, the arithmetic device and control device of the computer perform calculation and control based on the program in order to execute each process. In addition, a storage device included in the computer stores data used for processing based on a program in order to execute each processing.

Also, the program can be recorded on a computer-readable recording medium and distributed. Note that the recording medium is a medium such as a magnetic tape, flash memory, optical disk, magneto-optical disk, or magnetic disk. Also, the recording medium may be an auxiliary storage device or the like. Additionally, the program can be distributed over telecommunications lines.

Although specific examples of preferred embodiments of the present invention have been described in detail above, the present invention is not limited to the embodiments and the like described in the drawings. That is, various modifications or changes are possible within the broad scope of the invention described in the claims.

1 Information Processing System DIN Input Text DOUT Output Text UR User IMG Image CLR Color ADJ Adjective Pair Evaluation Scale W1 Word W2 Phonetic Symbol Word LI1 Word List LI2 Phonetic Symbol Word List

Claims (27)

an acquisition unit that acquires an adjective pair rating scale that is a rating scale input by a user for an adjective pair;
a list generation unit that generates at least one of a word list and a phonetic symbol word list based on a color vector indicating the association probability of each color associated with the adjective pair evaluation scale acquired by the acquisition unit;
a text acquisition unit for acquiring input text;
selecting a plurality of words or phonetic symbolic words from at least one of the word list and the phonetic symbolic word list according to a predetermined condition;words contained in the input text,said plurality of selected words or phonetic symbolsconvert toa text generator that generates output text;
Information processing system including; The list generation unit calculates a first color vector indicating a probability that a color is associated according to the adjective pair evaluation scale acquired by the acquisition unit, and based on the first color vector, the word 2. The information processing system according to claim 1, wherein the word list containing or the phonetic symbolic word list containing phonetic symbolic words is generated. The list generating unit generates the first color vector calculated based on the adjective pair evaluation scale acquired by the acquiring unit, and the word in the word list or the phonetic symbolic word in the phonetic symbolic word list. 3. The information processing system according to claim 2, wherein said word list or said phonetic symbol word list is generated by calculating a degree of similarity with a second color vector indicating the probability of color association. an acquisition unit that acquires an image or color;
a list generation unit that generates a phonetic symbolic word list including phonetic symbolic words based on the adjective pair evaluation scale corresponding to the image or the color acquired by the acquisition unit;
a text acquisition unit for acquiring input text;
selecting a plurality of phonetic symbolic words from the list of phonetic symbolic words according to a predetermined condition;words contained in the input text,said plurality of selected phonetic symbolsconvert toa text generator that generates output text;
Information processing system including; The list generation unit determines the degree of similarity between a first adjective pair evaluation scale calculated from the image or the color acquired by the acquisition unit and a second adjective pair evaluation scale obtained by evaluating the phonetic symbol word. 5. The information processing system according to claim 4, wherein the phonetic symbol word list is generated by calculation. 6. The information processing system according to any one of claims 1 to 5, further comprising a first conversion rate input unit for inputting a first conversion rate indicating a rate at which words included in said input text are to be converted. 7. The information according to claim 6, further comprising a second conversion rate input unit for inputting a second conversion rate indicating a rate at which words included in the input text are converted into phonetic symbols or words. processing system. The text generation unit performs morphological analysis on the input text, and includes the words in the input text if the part of speech of the words in the input text is a verb, adjective, noun, adverb, or adnominal. 4. The information processing system according to any one of claims 1 to 3 , wherein the word is converted into the word or the phonetic symbol word of the same part of speech as that of the word. 4. The text generator according to any one of claims 1 to 3 , wherein the text generation unit morphologically analyzes the input text and does not convert the words included in the input text if the part of speech of the words included in the input text is a particle. Information processing system as described. an acquisition unit that acquires an adjective pair rating scale that is a rating scale input by a user for an adjective pair;
a list generation unit that generates at least one of a word list and a phonetic symbol word list based on a color vector indicating the association probability of each color associated with the adjective pair evaluation scale acquired by the acquisition unit;
a text generation unit that generates output text by inputting the word list or the phonetic symbol word list into a machine learning model . an acquisition unit that acquires an image or color;
a list generation unit that generates a phonetic symbolic word list including phonetic symbolic words based on the adjective pair evaluation scale corresponding to the image or the color acquired by the acquisition unit;
and a text generator that generates output text by inputting the phonetic symbol word list into a machine learning model . 12. The information processing system according to claim 10, wherein the machine learning model is trained to receive one or more words as input and output a sentence including one or more words selected from the said words. 13. The information processing system according to claim 12, wherein said machine learning model is a recursive neural network. 4. The text generation unit analyzes semantic relationships between words in sentences included in the output text , and selects words included in the word list according to the semantic relationships. 11. The information processing system according to any one of 10 . further including a specification part specifying the number of characters,
11. The information processing system according to any one of claims 1 to 3 and 10 , wherein the text generation unit selects words included in the word list so as to have the number of characters. further comprising a calculator for calculating a color vector of the output text;
The text generation unit is configured to increase the similarity between the color vector calculated based on the adjective pair evaluation scale acquired by the acquisition unit and the color vector of the output text so that the color vector included in the word list is high. 11. The information processing system according to any one of claims 1 to 3 and 10 , wherein the output text is generated by selecting words that are wherein the text generation unit analyzes semantic relationships between words in sentences included in the output text, and selects phonetic symbolic words included in the phonetic symbolic word list according to the semantic relationships. Item 14. The information processing system according to any one of Items 1 to 5 and 10 to 13 . further including a specification part specifying the number of characters,
14. The information processing system according to any one of claims 1 to 5 and 10 to 13 , wherein the text generation unit selects phonetic symbol words included in the phonetic symbol word list so as to have the number of characters. further comprising a calculator for calculating a color vector of the output text;
The text generation unit generates the phonetic symbol word so that the color vector calculated based on the adjective pair evaluation scale acquired by the acquisition unit and the color vector of the output text have a high degree of similarity. 14. An information processing system according to any one of claims 1 to 5 and 10 to 13 , wherein phonetic symbol words contained in a list are selected to generate said output text. An information processing method performed by an information processing system,
an acquisition procedure in which an information processing system acquires an adjective pair rating scale that is a rating scale input by a user for an adjective pair;
a list generation procedure in which the information processing system generates at least one of a word list and a phonetic symbol word list based on a color vector indicating the association probability of each color associated with the adjective pair evaluation scale acquired in the acquisition procedure; ,
a text acquisition procedure for an information processing system to acquire an input text;
an information processing system that selects a plurality of words or phonetic symbolic words from at least one of the word list and the phonetic symbolic word list according to a predetermined condition;words contained in the input text,said plurality of selected words or phonetic symbolsconvert toa text generation procedure that produces output text;
Information processing method including. An information processing method performed by an information processing system,
an acquisition procedure for the information processing system to acquire an image or color;
a list generation procedure for the information processing system to generate a phonetic symbolic word list including phonetic symbolic words based on the adjective pair evaluation scale corresponding to the image or the color acquired by the acquisition procedure;
a text acquisition procedure for an information processing system to acquire an input text;
an information processing system that selects a plurality of phonetic symbolic words from the phonetic symbolic word list according to a predetermined condition;words contained in the input text,said plurality of selected phonetic symbolsconvert toa text generation procedure that produces output text;
Information processing methods, including An information processing method performed by an information processing system,
an acquisition procedure in which an information processing system acquires an adjective pair rating scale that is a rating scale input by a user for an adjective pair;
a list generation procedure in which the information processing system generates at least one of a word list and a phonetic symbol word list based on a color vector indicating the association probability of each color associated with the adjective pair evaluation scale acquired in the acquisition procedure; ,
an information processing method comprising: an information processing system generating an output text by inputting the word list or the phonetic symbol word list into a machine learning model . An information processing method performed by an information processing system,
an acquisition procedure for the information processing system to acquire an image or color;
a list generation procedure in which the information processing system generates a phonetic symbolic word list including phonetic symbolic words based on the adjective pair evaluation scale corresponding to the image or the color acquired by the acquisition procedure;
and a text generation procedure in which the information processing system inputs the phonetic symbol word list to a machine learning model to generate an output text. A program for causing a computer to execute an information processing method,
a text acquisition procedure in which the computer acquires the input text;
an acquisition step by which a computer acquires an adjective pair rating scale that is a user-entered rating scale for an adjective pair;
a list generation procedure in which a computer generates at least one of a word list and a phonetic symbol word list based on a color vector indicating the association probability of each color associated with the adjective pair evaluation scale acquired in the acquisition procedure;
A computer selects a plurality of words or phonetic symbols from at least one of the word list and the phonetic symbolic word list according to a predetermined condition;words contained in the input text,said plurality of selected words or phonetic symbolsconvert toa text generation procedure that produces output text;
program to run the A program for causing a computer to execute an information processing method,
a text acquisition procedure in which the computer acquires the input text;
a computer acquires an image or color;
a list generation procedure for generating a phonetic symbolic word list including phonetic symbolic words based on the adjective pair evaluation scale corresponding to the image or the color acquired by the acquisition procedure;
a computer selecting a plurality of phonetic symbolic words from the phonetic symbolic word list according to a predetermined condition;words contained in the input text,said plurality of selected phonetic symbol wordsconvert toa text generation procedure that produces output text;
program to run the A program for causing a computer to execute an information processing method,
an acquisition step by which a computer acquires an adjective pair rating scale that is a user-entered rating scale for an adjective pair;
a list generation procedure in which a computer generates at least one of a word list and a phonetic symbol word list based on a color vector indicating the association probability of each color associated with the adjective pair evaluation scale acquired in the acquisition procedure;
A program for causing a computer to execute a text generation procedure for generating output text by inputting said word list or said phonetic symbol word list into a machine learning model. A program for causing a computer to execute an information processing method,
a computer acquires an image or color;
a list generation procedure for generating a phonetic symbolic word list including phonetic symbolic words based on the adjective pair evaluation scale corresponding to the image or the color acquired by the acquisition procedure;
A program for causing a computer to execute a text generation procedure for generating output text by inputting the phonetic symbol word list into a machine learning model .

Images