KR20200086986A - Dialog analysis server, method and computer program for providing recommendation sentence - Google Patents

Abstract

Provided is a dialog analysis server, which comprises: a dictionary generation unit which generates a plurality of word dictionaries including at least one word; a replacement use weight calculation unit which calculates a replacement use weight between the plurality of word dictionaries based on dialog history data; a frequency weight calculation unit which calculates a frequency weight for the plurality of word dictionaries based on the frequency of use of a plurality of verbal units included in the dialog history data; a reference score calculation unit which calculates reference scores for the plurality of word dictionaries based on the calculated replacement use weight and the calculated frequency weight; and a recommendation unit which recommends a recommendation sentence for an input sentence input from a user based on the calculated reference score.

Description

DIALOG ANALYSIS SERVER, METHOD AND COMPUTER PROGRAM FOR PROVIDING RECOMMENDATION SENTENCE}

The present invention relates to a server, method, and computer program that provides recommended sentences.

Conventional conversation platforms have been designed as simple combinations of individually developed components for speech recognition and speech synthesis. Therefore, a third party (3 ^rd Party) with no prior knowledge of this dialogue platform saga had difficulties in providing the recommended sentence by dialogue modeling tool.

Korean Registered Patent Publication No. 10-1806151 (Registration on December 1, 2017)

The present invention is to solve the problems of the prior art described above, calculates reference scores for a plurality of word dictionaries based on conversation history data, and recommends sentences for input sentences received from a user based on the calculated reference scores I would like to recommend However, the technical problems to be achieved by the present embodiment are not limited to the technical problems as described above, and other technical problems may exist.

As a technical means for achieving the above-described technical problem, the conversation analysis server according to the first aspect of the present invention includes a dictionary generator for generating a plurality of word dictionaries including at least one word; A replacement usage weight calculating unit for calculating replacement usage weights between the plurality of word dictionaries based on conversation history data; A frequency weight calculator configured to calculate frequency weights for the plurality of word dictionaries based on frequency of use for a plurality of word units included in the conversation history data; A reference score calculating unit calculating reference scores for the plurality of word dictionaries based on the calculated replacement usage weight and the calculated frequency weight; And a recommendation unit recommending a recommendation sentence for an input sentence received from a user based on the calculated reference score.

A method of providing a recommendation sentence in a conversation analysis server according to a second aspect of the present invention includes generating a plurality of word dictionaries including at least one word; Calculating replacement usage weights between the plurality of word dictionaries based on conversation history data; Calculating frequency weights for the plurality of word dictionaries based on frequency of use for a plurality of verbal units included in the conversation history data; Calculating reference scores for the plurality of word dictionaries based on the calculated replacement usage weights and the calculated frequency weights; And recommending a recommendation sentence for the input sentence received from the user based on the calculated reference score.

A computer program stored in a medium comprising a sequence of instructions for providing a recommended sentence according to the third aspect of the present invention, the computer program, when executed by a computing device, comprises a plurality of word dictionaries including at least one word. Generating, calculating replacement usage weights between the plurality of word dictionaries based on conversation history data, and calculating frequency weights for the plurality of word dictionaries based on frequency of use for a plurality of verbs included in the conversation history data Calculating, calculating reference scores for the plurality of word dictionaries based on the calculated replacement usage weights and the calculated frequency weights, and recommending sentences for input sentences received from a user based on the calculated reference scores It may contain a sequence of recommended commands.

The above-described problem solving means are merely exemplary and should not be construed as limiting the invention. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and detailed description of the invention.

According to any one of the above-described problem solving means of the present invention, reference scores for a plurality of word dictionaries are calculated based on conversation history data, and recommendation sentences for input sentences received from a user based on the calculated reference scores Can recommend

1 is a block diagram of a system for providing a recommendation sentence according to an embodiment of the present invention.
2 is a block diagram of the conversation analysis server shown in FIG. 1 according to an embodiment of the present invention.
3 is a diagram for explaining a method of providing a recommendation sentence, according to an embodiment of the present invention.
4 is a flowchart illustrating a method for providing a recommendation sentence according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art to which the present invention pertains may easily practice. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and like reference numerals are assigned to similar parts throughout the specification.

Throughout the specification, when a part is "connected" to another part, this includes not only "directly connected" but also "electrically connected" with other elements in between. . Also, when a part “includes” a certain component, this means that other components may be further included instead of excluding other components, unless otherwise specified.

In the present specification, the term “unit” includes a unit realized by hardware, a unit realized by software, and a unit realized by using both. Further, one unit may be realized by using two or more hardware, and two or more units may be realized by one hardware.

Some of the operations or functions described in this specification as being performed by a terminal or device may be performed instead on a server connected to the corresponding terminal or device. Similarly, some of the operations or functions described as being performed by the server may be performed in a terminal or device connected to the corresponding server.

Hereinafter, specific contents for carrying out the present invention will be described with reference to the accompanying drawings or process flow charts.

1 is a block diagram of a system for providing a recommendation sentence according to an embodiment of the present invention.

Referring to FIG. 1, a system for providing a recommended sentence may include a conversation analysis server 100 and a voice input/output device 110. However, the system for providing a recommended sentence of FIG. 1 is only an embodiment of the present invention, and thus the present invention is not limitedly interpreted through FIG. 1, and may be configured differently from FIG. 1 according to various embodiments of the present invention. .

Generally, each component of the recommendation sentence providing system of FIG. 1 is connected through a network (not shown). The network means a connection structure capable of exchanging information between nodes such as terminals and servers, and a local area network (LAN), a wide area network (WAN), and the Internet (WWW: World) Wide Web), wired and wireless data communication networks, telephone networks, and wired and wireless television communication networks. Examples of wireless data communication networks include 3G, 4G, 5G, 3rd Generation Partnership Project (3GPP), Long Term Evolution (LTE), World Interoperability for Microwave Access (WIMAX), Wi-Fi, Bluetooth communication, infrared communication, ultrasound Communication, Visible Light Communication (VLC), LiFi, and the like are included, but are not limited thereto.

The conversation analysis server 100 interprets the intention of the user's input sentence (speech sentence) input from the voice input/output device 110, generates an answer in accordance with the interpreted intention, and may provide it to the voice input/output device 110. have.

The conversation analysis server 100 may store conversation history data including an input sentence, an analysis result for the input sentence, and an answer to the input sentence in the usage history DB.

The conversation analysis server 100 may generate a plurality of word dictionaries including at least one word based on the conversation history data. Specifically, the conversation analysis server 100 may generate a plurality of word dictionaries including a representative word and similar words related to the representative word.

The conversation analysis server 100 may learn whether the verbs corresponding to the verbs of each sentence included in the conversation history data can be used interchangeably. For example, the conversation analysis server 100 may learn whether'execute' and'replay' can be used interchangeably in each sentence of'Kim Kun-mo playing music' and'Kim Kun-mo playing music'.

The conversation analysis server 100 may calculate a replacement usage weight between a plurality of word dictionaries based on the conversation history data. For example, the conversation analysis server 100 may calculate a replacement usage weight between the first word dictionary to which'playback' belongs and the second word dictionary to which'playback' belongs.

The conversation analysis server 100 may calculate frequency weights for a plurality of word dictionaries based on the frequency of use of a plurality of verbs in each sentence included in the conversation history data.

The conversation analysis server 100 may calculate reference scores for a plurality of word dictionaries based on the replacement usage weight calculated for each word dictionary and the calculated frequency weight.

The conversation analysis server 100 may inversely recommend the recommendation sentence for the input sentence received from the user based on the calculated reference score.

Hereinafter, the operation of each component of the recommendation sentence providing system of FIG. 1 will be described in more detail.

2 is a block diagram of the conversation analysis server 100 shown in FIG. 1 according to an embodiment of the present invention.

Referring to FIG. 2, the dialogue analysis server 100 includes a dictionary generating unit 200, a replacement weight calculation unit 210, a frequency weight calculation unit 220, a reference score calculation unit 230, and a recommendation unit 240 And a threshold update unit 250. However, the conversation analysis server 100 illustrated in FIG. 2 is only one implementation example of the present invention, and various modifications are possible based on the components illustrated in FIG. 2. Hereinafter, it will be described with reference to FIG. 3 together with FIG. 2.

The dictionary generator 200 may generate a plurality of word dictionaries including at least one word. Specifically, the dictionary generator 200 may generate each of a plurality of word dictionaries including a representative word and similar words related to the representative word. Referring to FIG. 3, the dictionary generator 200 generates a first word dictionary 301 in which the representative word is'play' and includes similar words related to'play' ('twist','play', etc.) can do. Also, the dictionary generator 200 may generate a second word dictionary 303 in which the representative word is'execution' and similar words related to'execution' ('action','twist', etc.).

The replacement usage weight calculator 210 may calculate replacement usage weights among a plurality of word dictionaries based on the conversation history data. Referring to FIG. 3, when there is a user's utterance sentence in the conversation history data such as'playing Kim Gun-mo music' and'playing Kim Gun-mo music','playing' and'playing' corresponding to the usage part of each sentence Can be used interchangeably. In this case, the replacement usage weight calculator 210 may calculate the replacement usage weight between the first word dictionary 301 to which the'play solution' belongs and the second word dictionary 303 to which the'execution solution' belongs.

Specifically, the replacement usage weight calculating unit 210 classifies a plurality of sentences included in the conversation history data according to the intention of each sentence to be interpreted, and the replacement use weights between the word dictionaries corresponding to the verbs included in the classified sentences Can be calculated. For example, the replacement usage weight calculating unit 210 is a sentence used with the same intention of each sentence of'playing Kim Gun-mo music' and'playing Kim Gun-mo music', so that'playing' corresponding to the usage part of each sentence And the replacement word weight between the first word dictionary 301 and the second word dictionary 303 to which'execution' belongs.

In addition, the replacement usage weight calculating unit 210 may sum the frequency of the usage of the replacement by the usage unit for the same sentence, except for the usage unit included in the classified sentence among the classified sentences. For example, when looking at the sentences of'playing Kim Gun-mo music' and'playing Kim Gun-mo music' classified with the same intention,'Kim Gun-mo music' corresponding to the body part is the same,'playing music' and ' 'Do it', you can see the difference. At this time, the replacement usage weight calculator 210 calculates the replacement usage weight between the first word dictionary 301 and the second word dictionary 303 by summing up the frequency of replacement usage between the'play solution' and the'run solution'. can do.

The replacement usage weight calculator 210 may calculate replacement usage weights among the plurality of word dictionaries by summing up the replacement usage weights calculated for each word included in the plurality of word dictionaries. For example, the replacement usage weight calculating unit 210 sets the replacement usage weight used for replacement between each of a plurality of words included in the first word dictionary 301 and each of a plurality of words included in the second word dictionary 303. By adding up, it is possible to calculate the replacement usage weight between the first word dictionary 301 and the second word dictionary 303.

The frequency weight calculation unit 220 may calculate the frequency weights for the plurality of word dictionaries based on the frequency of use for the plurality of verbs included in the conversation history data. Here, the frequency of use for the plurality of verbal units included in the conversation history data may be calculated by dividing the number of times each verbal unit appears in the conversation history data by the frequency of the entire verbal units. For example, assuming that the entire verbatim word was used 1000 times in the conversation history data, the first word ('say') was used 200 times, and the second word ('open the order') was used once. The frequency of use for one word is 0.2 (=200/1000), and the frequency of use for the second word can be calculated as 0.001 (=1/1000).

The frequency weight calculation unit 220 calculates the frequency of use for the plurality of verbs included in the conversation history data, classifies the plurality of verbs by the word dictionary corresponding to each of the plurality of verbs, and calculates each classified word dictionary The frequency weights for a plurality of word dictionaries can be calculated by summing up the frequency of use for the used words.

The frequency weight calculation unit 220 sums up the frequency weights (the frequency of use for the verbatim unit of the conversation history data corresponding to each word) calculated for each word included in the plurality of word dictionaries for the plurality of word dictionaries. The frequency weight can be calculated. For example, referring to FIG. 3, the frequency weight calculation unit 220 is a frequency weight for the first word dictionary 301 as a summation of the frequency weights of each of a plurality of words included in the first word dictionary 301. Can be calculated.

The reference score calculator 230 may calculate reference scores for the plurality of word dictionaries based on the replacement usage weight calculated for each word dictionary and the calculated frequency weight. Here, the reference score can be calculated through [Equation 1].

[Equation 1]

Reference score = replacement usage weight X-log (frequency weight)/number of sentences classified with the same intention

The reference score calculator 230 may calculate the reference score to be inversely proportional to the calculated frequency weight. For example, a word dictionary containing a word with a low frequency weight increases the reference score, and a word dictionary containing a word with a high frequency weight decreases the reference score. Referring to FIG. 3, numerical values displayed on both sides of edges connecting between a plurality of word dictionaries 301, 303, 305, and 307 are reference scores. The reference score in which the first word dictionary 301 can be used crosswise as the second word dictionary 303 is 0.4, and the reference score in which the second word dictionary 303 can be used crosswise as the first word dictionary 301 is 0.2. to be.

The recommender 240 may recommend a recommendation sentence for an input sentence received from a user based on the calculated reference score. For example, the recommender 240 may recommend the recommendation sentences in the order in which the calculated reference score value is largest.

The recommendation unit 240 may extract the terminology included in the input sentence, and recommend the recommended sentence based on the reference score for the word dictionary corresponding to the extracted terminology. For example, referring to FIG. 3, when the input sentence input from the user is'play a song' and'play a song', the recommendation unit 240 plays'play a song' corresponding to the pronoun A recommendation sentence may be recommended based on the reference scores of the first word dictionary 301 to which the year belongs and the second word dictionary 303 to which the'execution solution' belongs to the pronoun of the'song execution solution'. .

The recommendation unit 240 may recommend the recommendation sentence based on the reference scores of other word dictionaries except for the word dictionary corresponding to the usage statement included in the input sentence. For example, the recommendation unit 240 excludes the first word dictionary 301 and the second word dictionary 303 to which the verb included in the input sentence belongs, and the third word dictionary 305 ) And the recommendation sentence based on the reference score for the fourth word dictionary 307. At this time, in the case of the third word dictionary 305, the recommendation unit 240 may include a reference score (0.1) between the third word dictionary 305 and the first word dictionary 301 and a third word dictionary 305 and a third The reference score (0.1=(0.1+0.1)/2) for the third word dictionary 305 may be calculated by summing the reference scores (0.1) between the two word dictionaries (303) and dividing by the reference dictionary number (2). In the case of the fourth word dictionary 307, the recommendation unit 240 may have a reference score (0.2) between the fourth word dictionary 307 and the first word dictionary 301 and a fourth word dictionary 307 and a second word dictionary The reference scores (0.25 = (0.3+0.2)/2) for the fourth word dictionary 307 may be calculated by summing the reference scores (0.3) between (303) and dividing by the reference dictionary number (2). At this time, the recommender 240 uses the fourth word dictionary 307 because the reference score (0.25) for the fourth word dictionary 307 is higher than the reference score (0.1) for the third word dictionary 305. By creating a recommendation sentence such as'Please sing a song', you can recommend it.

The threshold update unit 250 may update the threshold for the use of the word dictionary included in the recommended sentence according to whether the recommended recommended sentence is selected by the user. For example, if the user does not select a recommendation sentence having a threshold of 0.1 for use of the third word dictionary 305 and a reference score of 0.3 for the third word dictionary 305, the threshold update unit 250 May update the threshold value 0.13 for use of the third word dictionary 305 by adding a constant value (0.03=0.3/10) for the third word dictionary 305 to the current threshold value 0.1. Thereafter, when the user selects a recommendation sentence having a reference score of 0.2 for the third word dictionary 305, the threshold update unit 250 sets a constant value (0.02) for the third word dictionary 305 to the current threshold value 0.13. =0.2/10) to subtract the threshold 0.11 for use of the third word dictionary 305 again.

On the other hand, when the user selects a recommendation sentence recommended to the user, the recommendation unit 240 recommends the recommendation sentence based on a reference score for other word dictionaries except for the word dictionary to which the glossary of the recommendation sentence belongs when recommending the next sentence Can recommend For example, referring to FIG. 3, when the user selects a recommendation sentence generated using the fourth word dictionary 307, the recommender 240 excludes the fourth word dictionary 307 when recommending the next sentence. A recommendation sentence may be recommended based on reference scores for each of the first word dictionary 301, the second word dictionary 303, and the third word dictionary 305.

The recommendation unit 240 may determine whether to generate a recommendation sentence using each word dictionary based on whether the reference score for each word dictionary exceeds a threshold. That is, the recommender 240 may generate a recommendation sentence through a word dictionary whose reference score for the word dictionary is greater than or equal to a threshold.

For example, referring to FIG. 3, when the user selects a recommendation sentence generated using the fourth word dictionary 307, the first word dictionary 301 excluding the fourth word dictionary 307 when recommending the next sentence ), the second word dictionary 303 and the third word dictionary 305 among the third word dictionaries 305 that are not selected by the user may be selected as dictionary candidates for generating a recommended sentence. The recommendation unit 240 includes a reference score (0.1) between the third word dictionary 305 and the first word dictionary 301 and a reference score (0.1) between the third word dictionary 305 and the second word dictionary 303. The reference score (0.1=(0.1+) for the third word dictionary 305 is divided by adding the reference score (0.1) between the third word dictionary 305 and the fourth word dictionary 307 and dividing by the reference dictionary number (3). 0.1+0.1)/3) can be calculated. At this time, since the reference score for the third word dictionary 305 is less than the threshold, a recommendation sentence using the third word dictionary 305 is not generated.

On the other hand, if a person skilled in the art, the dictionary generation unit 200, the replacement weight calculation unit 210, the frequency weight calculation unit 220, the reference score calculation unit 230, the recommendation unit 240 and the threshold update unit 250, respectively It will be fully understood that this may be implemented separately, or one or more of these may be implemented integrally.

4 is a flowchart illustrating a method for providing a recommendation sentence according to an embodiment of the present invention.

Referring to FIG. 4, in step S401, the conversation analysis server 100 may generate a plurality of word dictionaries including at least one word.

In step S403, the conversation analysis server 100 may calculate the replacement usage weight between the plurality of word dictionaries based on the conversation history data.

In step S405, the conversation analysis server 100 may calculate the frequency weights for the plurality of word dictionaries based on the frequency of use for the plurality of verbs included in the conversation history data.

In step S407, the conversation analysis server 100 may calculate reference scores for the plurality of word dictionaries based on the calculated replacement use weight and the calculated frequency weight.

In step S409, the conversation analysis server 100 may recommend the recommended sentence for the input sentence received from the user based on the calculated reference score.

Although not illustrated in FIG. 4, in step S403, the dialogue analysis server 100 may calculate the replacement usage weights among the plurality of word dictionaries by summing the replacement usage weights calculated for each word included in the plurality of word dictionaries. .

Although not shown in FIG. 4, in step S405, the dialogue analysis server 100 may calculate the frequency weights for the plurality of word dictionaries by adding up the frequency weights calculated for each word included in the plurality of word dictionaries.

In the above description, steps S401 to S409 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present invention. In addition, some steps may be omitted if necessary, and the order between the steps may be changed.

One embodiment of the present invention may also be implemented in the form of a recording medium including instructions executable by a computer, such as program modules, being executed by a computer. Computer readable media can be any available media that can be accessed by a computer, and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium may include any computer storage medium. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.

The above description of the present invention is for illustration only, and those skilled in the art to which the present invention pertains can understand that the present invention can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present invention is indicated by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and equivalent concepts should be interpreted to be included in the scope of the present invention. .

100: chat analysis server
110: voice input and output device
200: dictionary generation unit
210: replacement use weight calculator
220: frequency weight calculation unit
230: reference score calculator
240: recommendation part
250: threshold update unit

Claims (18)

In the conversation analysis server,
A dictionary generating unit generating a plurality of word dictionaries including at least one word;
A replacement usage weight calculating unit for calculating replacement usage weights between the plurality of word dictionaries based on conversation history data;
A frequency weight calculator configured to calculate frequency weights for the plurality of word dictionaries based on frequency of use for a plurality of verbal units included in the conversation history data;
A reference score calculating unit calculating reference scores for the plurality of word dictionaries based on the calculated replacement use weights and the calculated frequency weights; And
A recommendation unit recommending a recommendation sentence for an input sentence received from a user based on the calculated reference score
Containing, the conversation analysis server.
According to claim 1,
Each word dictionary includes a representative word and similar words related to the representative word, the conversation analysis server.
According to claim 2,
The replacement usage weight calculator calculates replacement usage weights between the plurality of word dictionaries by adding replacement usage weights calculated for each word included in the plurality of word dictionaries,
The frequency weight calculation unit calculates a frequency weight for the plurality of word dictionaries by summing the frequency weights calculated for each word included in the plurality of word dictionaries.
The method of claim 3,
The replacement usage weight calculating unit classifies a plurality of sentences included in the conversation history data according to the intention of each sentence to be interpreted, and calculates replacement usage weights between word dictionaries corresponding to the verbs included in the classified sentences. Phosphorus, conversation analysis server.
The method of claim 4,
The replacement usage weight calculating unit sums up the frequency of replacement usage of the usage unit with respect to the sentence having the same body unit except for the usage unit included in the classified sentence among the classified sentences.
The method of claim 3,
The frequency weight calculation unit calculates the frequency of use for a plurality of verbs included in the conversation history data, classifies the plurality of verbs by word dictionary corresponding to each of the plurality of verbs, and by the classified word dictionary Conversation analysis server to calculate the frequency weight for the plurality of word dictionaries by summing the frequency of use for the calculated verb.
According to claim 1,
The reference score calculation unit is to calculate the reference score to be inversely proportional to the calculated frequency weight, conversation analysis server.
According to claim 1,
The recommendation unit extracts the usage statement included in the input sentence, and recommends the recommendation sentence based on a reference score for a word dictionary corresponding to the extracted usage statement.
According to claim 1,
And a threshold update unit that updates a threshold for the use of the word dictionary included in the recommended sentence according to whether the recommended recommended sentence is selected by the user.
The method of claim 9,
The recommendation sentence, the reference score is generated through a dictionary of words above the threshold, the conversation analysis server.
In the method of providing a recommended sentence in the conversation analysis server,
Generating a plurality of word dictionaries including at least one word;
Calculating replacement usage weights between the plurality of word dictionaries based on conversation history data;
Calculating frequency weights for the plurality of word dictionaries based on frequency of use for a plurality of verbal units included in the conversation history data;
Calculating reference scores for the plurality of word dictionaries based on the calculated replacement usage weights and the calculated frequency weights; And
Recommending a recommendation sentence for an input sentence received from a user based on the calculated reference score
That includes, how to provide a recommended sentence.
The method of claim 11,
The step of calculating the replacement use weight is
And calculating replacement usage weights between the plurality of word dictionaries by adding replacement usage weights calculated for each word included in the plurality of word dictionaries,
The step of calculating the frequency weight is
And calculating frequency weights for the plurality of word dictionaries by summing up frequency weights calculated for each word included in the plurality of word dictionaries.
The method of claim 12,
The step of calculating the replacement use weight is
Comprising the steps of classifying a plurality of sentences included in the conversation history data according to the intention of each sentence to be interpreted, and calculating a replacement usage weight between word dictionaries corresponding to the verbs included in the classified sentences. How to provide recommended sentences.
The method of claim 13,
The step of calculating the replacement use weight is
And a step of summing the frequency with which the saying part is used for the same sentence among the classified sentences, except for the saying part included in the classified sentence.
The method of claim 12,
The step of calculating the frequency weight is
Calculate the frequency of use for a plurality of verbs included in the conversation history data, classify the plurality of verbs by word dictionary corresponding to each of the plurality of verbs, and calculate the verbs by each classified word dictionary And calculating frequency weights for the plurality of word dictionaries by summing the frequency of use for.
The method of claim 11,
The step of calculating the reference score
And calculating the reference score to be inversely proportional to the calculated frequency weight.
The method of claim 11,
The step of recommending the recommended sentence is
And extracting a verbal phrase included in the input sentence, and recommending the recommended sentence based on a reference score for a word dictionary corresponding to the extracted verbal phrase.
A computer program stored on a medium comprising a sequence of instructions providing a recommendation sentence,
When the computer program is executed by a computing device,
Create a plurality of word dictionaries containing at least one word,
Based on the conversation history data, the replacement usage weight between the plurality of word dictionaries is calculated,
Frequency weights for the plurality of word dictionaries are calculated based on a frequency of use for a plurality of verbal units included in the conversation history data,
A reference score for the plurality of word dictionaries is calculated based on the calculated replacement usage weight and the calculated frequency weight,
And a sequence of instructions for recommending a recommendation sentence for an input sentence received from a user based on the calculated reference score.

Images