KR102332268B1 - Customer Consultation Summary Apparatus and Method - Google Patents

Abstract

An exemplary embodiment of the present invention is designed to improve VoC analysis efficiency by reflecting the characteristics of 1) several topics are mixed, 2) the beginning and the end of a sentence are ambiguous, and 3) the subject word is not repeated. To a customer consultation summary apparatus and method, the customer consultation summary apparatus according to an aspect of the present invention comprises: a paragraph separation unit for separating the consultation transcript into paragraphs by conversation topic; a summary sentence output unit for extracting or generating a sentence similar to a counseling memo from each of the separated paragraphs and outputting it as a summary sentence; and a summary sentence correction unit for correcting the output summary sentence based on at least one of a preceding sentence and a trailing sentence for the output summary sentence and a preset key word.

Description

Customer Consultation Summary Apparatus and Method

The present invention relates to a customer consultation summary apparatus and method.

In general, various companies, such as financial/telecommunications companies that directly deal with general customers, are conducting improvement activities by centrally analyzing the content requested by customers (VoC: Voice of Customer) in order to solve the inconvenience of customers.

To this end, the contents of the consultation are recorded, converted into text that is easy to analyze, and stored on the server. At this time, in order not to miss the customer's complaints that are not revealed in the recording, the counselor directly summarizes the customer's complaints in the system and writes them down in a consulting memo.

In order to understand these VoCs at a glance, the company directly views the consultation memos and recorded data and organizes the VoCs by type. Since this type of VoC changes every time, it is an area that is difficult to automate, so it is necessary to go through the work of checking with the human eye. Hundreds of thousands of calls come in a day, but only a fraction of them can be sorted out because people have to check them.

Therefore, by introducing an automatic summary system for customer center counseling that extracts or generates sentences that contain the content requested by the customer from the customer center counseling contents, it is possible to efficiently check the recorded data that lacks counseling notes. In addition, it is possible to reduce the number of times the counselor writes memos, and it is possible to reduce human error by identifying customer complaints that the counselor has not yet written.

Many companies with call centers wanted to introduce a summary automation system to streamline VoC analysis. As a general conventional summary method, a method of extracting important sentences by calculating the frequency of keywords is used.

However, according to the existing summary method described above, it is possible to extract sentences that are frequent but not necessary for VoC analysis (ex: Hello, customer, your location. There is a problem.

That is, the counseling recording data has characteristics that 1) multiple topics are mixed, 2) the beginning and end of a sentence are ambiguous, and 3) the subject word is not repeated. Although the frequency is high, there was a problem that sentences that were not helpful for VoC analysis could be extracted.

Laid-open Patent Publication No. 10-2018-0073226 (2016.12.22.)

The present invention is to solve the above-mentioned problems of the prior art, and the purpose is to reflect the characteristics of the counseling recording data in which 1) several subjects are mixed, 2) the beginning and end of a sentence are ambiguous, and 3) the main word is not repeated To provide a customer consultation summary device and method to improve VoC analysis efficiency.

In order to achieve the above object, a customer counseling summary apparatus according to an aspect of the present invention includes: a paragraph separator for separating the counseling transcript into paragraphs for each topic of conversation; a summary sentence output unit for extracting or generating a sentence similar to a counseling memo from each of the separated paragraphs and outputting it as a summary sentence; and a summary sentence correction unit for correcting the output summary sentence based on at least one of a preceding sentence and a trailing sentence for the output summary sentence and a preset key word.

The paragraph dividing unit obtains the probability of belonging to each counseling type for each sentence of the counseling transcript, and if the sum of the probabilities of the counseling type of the same counseling type for N adjacent sentences is equal to or greater than a preset reference value, the counseling type is the first or the previous counseling type If different from , it may be recognized as a paragraph of a conversation topic corresponding to the corresponding counseling type and divided into a paragraph distinguished from other paragraphs.

The paragraph separator may obtain a counseling type classification probability for each sentence using a learned statistical model (classification model), and the statistical model may include at least one of an artificial intelligence network and a support vector machine (SVM).

The summary sentence correction unit may check whether a preceding sentence and a trailing sentence are connected to the output summary sentence, respectively, and connect the connected sentence to the output summary sentence to correct it.

The summary sentence correction unit calculates the probability of connectivity between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence) using a pre-learned sequential language model. and if the calculated connectivity probability is greater than or equal to a preset reference value, it can be confirmed that there is connectivity.

The summary sentence correction unit uses a pre-learned sequential language model to correspond to the probability of connectivity between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence) calculates a first score, gives a weight corresponding to the preset number of key words included in a sentence to be connected (that is, each of the preceding sentence and the following sentence), and the sum of the first score and the weight is If it exceeds the preset reference value, it can be confirmed that there is connectivity.

The key words may be extracted from the consultation memo, and the language model may include a Long Short-Term Memory Network (LSTM).

In order to achieve the above object, there is provided a customer counseling summary method according to another aspect of the present invention, comprising the steps of: (a) separating the counseling transcript into paragraphs by conversation topic; (b) extracting or generating a sentence similar to a counseling memo from each of the separated paragraphs and outputting it as a summary sentence; and (c) correcting the output summary sentence based on at least one of a preceding sentence and a trailing sentence for the output summary sentence, and a preset key word.

In step (a), the probability of belonging to each counseling type for each sentence of the counseling transcript is obtained, the sum of the probabilities of the counseling type of the same counseling type for N adjacent sentences is equal to or greater than a preset reference value, and the corresponding counseling type is the first or preceding If it is different from the counseling type (in a separate paragraph), the transcript of the counseling may be separated into a paragraph of a conversation topic corresponding to the corresponding counseling type.

In the step (a), a counseling type classification probability for each test sentence may be obtained using a learned statistical model, and the statistical model may include at least one of an artificial intelligence network and a support vector machine (SVM).

In the step (c), it is possible to check whether or not there is a connection between the preceding sentence and the trailing sentence with respect to the output summary sentence, respectively, and connect the connected sentence to the output summary sentence and correct it.

The step (c) uses a pre-learned sequential language model to determine the probability of connectivity between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence) If the calculated connectivity probability is greater than or equal to a preset reference value, it can be confirmed that there is connectivity.

The step (c) uses a pre-learned sequential language model to determine the probability of connectivity between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence). A corresponding first score is calculated, a weight corresponding to the preset number of key words included in a sentence to be connected (ie, each of the preceding sentence and the following sentence) is given a weight, and the sum of the first score and the weight is If it is more than this preset reference value, it can be confirmed that there is connectivity.

The key words may be extracted from the consultation memo, and the language model may include a Long Short-Term Memory Network (LSTM).

According to another aspect of the present invention in order to achieve the above object, there may be provided a computer-readable recording medium in which a program for executing the customer consultation summary method on a computer is recorded.

According to another aspect of the present invention in order to achieve the above object, an application stored in a computer-readable recording medium may be provided in order to execute the customer consultation summary method in combination with hardware.

According to another aspect of the present invention in order to achieve the above object, there may be provided a computer program stored in a computer-readable recording medium to execute the customer consultation summary method on a computer.

As described above, according to various aspects of the present invention, 1) several topics are mixed, 2) the beginning and the end of a sentence are ambiguous, and 3) the subject words are not repeated by reflecting the characteristics of the customer consultation recording data by conversation topic It has the effect of improving VoC analysis efficiency by creating a summary of the customer consultation transcript including one or more extracted or generated summary sentences for each paragraph separated by .

That is, according to one aspect of the present invention, by prior to paragraph separation by conversation topic and extracting summarized sentences for each topic, the existing disadvantage that duplicate sentences may appear due to the characteristics of the counseling recording data is prevented, and VoC analysis efficiency is improved. effect that can be maximized.

In addition, in the case of dialogue, the section in which the sentence is completed is not clear due to the slurring of words (for example, to do, did, etc.), the sentence separation is excessive due to interruption, and the In many cases, only a part of a conversation is extracted, and in particular, in the case of VoC, customers tend to slurp without clearly expressing their dissatisfaction. According to one aspect of the present invention, after extracting the core sentence, Since it can be corrected as a complete sentence in the meaning of dialogue, it has the effect of maximizing the VoC analysis efficiency.

1 is a block diagram of a customer consultation summary apparatus according to an exemplary embodiment of the present invention;
2 is an exemplary diagram for explaining a paragraph separation method according to an embodiment of the present invention;
3 is an exemplary diagram for explaining a summary sentence correction according to an embodiment of the present invention;
4 is an exemplary diagram of input (consultation content) and output (summary) according to an embodiment of the present invention;
5 is an exemplary diagram for quantitative evaluation according to an embodiment of the present invention;
Fig. 6 is a flowchart of a customer consultation summary method according to an exemplary embodiment of the present invention;

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In adding reference numerals to the components of each drawing, the same components are to have the same reference numerals as much as possible even though they are indicated in different drawings. In addition, when it is determined that a detailed description of a known configuration or function related to the embodiment of the present invention may obscure the gist of the present invention, the detailed description thereof will be omitted.

1 is a block diagram of a customer consultation summary apparatus according to an exemplary embodiment of the present invention, and as shown in the figure, a transcript text input unit 11, a paragraph separation unit 12, a consultation memo input unit 13, It may include a morpheme analysis unit 14 , a normalization unit 15 , a key word extraction unit 16 , a summary sentence output unit 17 , a summary sentence correction unit 18 , and a summary sentence output/evaluation unit 19 . .

The recorded text input unit 11 is for inputting text data that is recorded and converted from a voice file to a text file. have.

The paragraph separation unit 12 is for dividing the counseling transcript into paragraphs by one or more conversation topics.

In general, the existing document summary method (model) extracts or creates a main sentence based on one topic. In the case of counseling, only one topic of inquiry exists in one conversation (consultation content), but there are many cases in which two or more topics are mixed in one conversation. For example, when consulting with a telecommunication company customer, there may be cases where you start with a new product inquiry and sign up for a family reunification service by recommending a counselor on the way. In this case, two topics (new product and family reunification) are mixed in one consultation. . As another example, in the case of consultations including customer complaints, there are many cases in which a number of topics are mixed, such as a conversation about a rate plan in the beginning and then a rate complaint from the middle.

Therefore, since the complaint call is a particularly important item in the case of summarizing the counseling transcript data, by separating the counseling transcript text into one or more dialog topic-specific paragraphs through the paragraph separation unit 12 according to this embodiment, it is included in one counseling content. Make sure you don't miss out on specific inquiries as well as customer complaints.

The paragraph separation unit 12 divides the counseling transcript into sequential sentences to obtain a probability of belonging to each counseling type (or referred to as a counseling type probability) for each sentence, and the counseling type probability of the same counseling type for N adjacent sentences If the sum of s is greater than or equal to the preset reference value and the corresponding counseling type is the first or is different from the preceding (separated paragraph) counseling type, the paragraph may be divided into a paragraph of a conversation topic corresponding to the corresponding counseling type.

The counseling type and counseling type probability for each sentence of the counseling transcript as shown in FIG. Through this, the probability of belonging to type A of counseling 5%, probability of belonging to type B of 40%, probability of belonging to type C of counseling 30%, and probability of belonging to type D of 25% can be obtained. For reference, the consultation type may be divided into A for a fee-related consultation type, B for a call quality-related consultation type, and the like.

The classification model (statistical model) of FIG. 2(b) can be generated by learning a model that gives a probability of belonging to a counseling type for each test sentence using a model including at least one of an artificial intelligence network and a support vector machine (SVM). .

When the sum of the probabilities of the same counseling type for N adjacent sentences exceeds a threshold and is different from the previous counseling type, the paragraph separator 12 may recognize that the content of the paragraph is separated.

For example, when N=3, when the counseling type probabilities for sequential sentences 1, 2, and 3 are shown in Table 1 below, the sum of the counseling type probabilities of the same counseling type A for N(=3) sentences is 5%+0+0=5%, the sum of the probabilities of the counseling types of the same counseling type B is 40%+0+40=80%, and the sum of the probabilities of the counseling types of the same counseling type C is 30%+70% +40%=140%, and the sum of the probabilities of the counseling type of the same counseling type D is 25%+30%+20%=75%, so the reference value is 120% and the preceding paragraph indicates that the counseling type is of the conversation topic corresponding to A. Assuming that it is a paragraph, the current paragraph can be divided into a paragraph of a conversation topic corresponding to the corresponding counseling type C.

sentence Counseling Type Probability One A --> 5%
B --> 40%
C --> 30%
D --> 25% 2 C --> 70%
D --> 30% 3 B --> 40%
C --> 40%
D --> 20%

For reference, it is important to prevent a general sentence such as "Hello" from becoming a specific counseling type by learning a classification model (statistical model) by making the number of sentences for each counseling type similar.

The counseling memo input unit 13 is for inputting the counseling memo text summarized by the counselor directly, for example, the summary model of the summary sentence output unit 17 summarizes the counseling memo directly summarized by the counselor as the correct answer of the (question) By learning to summarize sentences similar to the counseling memo using the .

The morpheme analysis unit 14 is for separating main words, endings, and propositions, etc. through morpheme analysis on the sentences output through the paragraph separation unit 12 and the sentences input through the counseling memo input unit 13 . Since the morpheme analysis method of can be applied, a detailed description will be omitted.

The normalization unit 15 is to reduce the total number of characters by normalizing the words separated by the morpheme analysis unit 14 into a certain form, for example, while doing, doing. Did, but, etc. can be normalized in the same form as 'had', and since the existing normalization method can be applied, a detailed description will be omitted.

The key word extraction unit 16 extracts the key words that must be included in the correct answer sentence (consultation memo) from the consultation memo data output through the normalization unit 15, and corrects the summary sentence in the summary sentence correction unit 18 and the summary sentence output/evaluation unit 19 to be used for comparison/evaluation of the summary sentence, and since the existing method can be applied to the key word extraction method, a detailed description thereof will be omitted.

The summary sentence output unit 17 is for extracting or generating a sentence similar to a counseling memo from each separated paragraph of the counseling transcript outputted through the normalization unit 15 and outputting it as a summary sentence, and the paragraph separating unit 12 ) to create and print a summary sentence for each paragraph separated by .

As an example, the summary sentence output unit 17 may extract a sentence similar to the VoC counseling memo from the body (consultation transcript), or use a model trained to generate a sentence similar to the VoC counseling memo as another example, for example, , a method in which a person directly defines and uses a pattern, a method to determine a main sentence through a formula, or a method to extract/generate results by learning an artificial neural network, etc. can be applied, so detailed descriptions are omitted.

The summary sentence correction unit 18 is at least one of the leading and trailing sentences in the counseling transcript for the summary sentence output through the summary sentence output unit 17 and the key words set and output through the key word extraction unit 16 . This is to correct the summary sentence output through the summary sentence output unit 17 based on the above.

The summary sentence correction unit 18 is configured based on the summary sentence when the customer's dissatisfaction is long due to the nature of the counseling recording text data, the speech is interrupted due to the intervention of the counselor during speaking, or the main sentence is separated due to an error in STT (Speech To Text). Check the connection between the front and back sentences (ie, preceding and following sentences) with the summary sentence and attach them. The summary sentence correction unit 18 includes a module for checking connectivity, and this module learns, for example, an O for a sentence with a connection and an X for a sentence without a connection. can increase the probability. At this time, when the key word extracted from the key word extraction unit 16 is included in the sentence to be added (the preceding sentence or the following sentence), a weight is given to the corresponding sentence.

The summary sentence correction unit 18 checks the connectivity of the preceding sentence and the trailing sentence for the summary sentence output through the summary sentence output unit 17, respectively, and connects the connected sentence to the corresponding summary sentence. As an example, by using a pre-learned sequential language model, the probability of connectivity between the preceding sentences (ie, between the preceding sentence and the corresponding summary sentence, and between the corresponding summary sentence and the following sentence) is calculated, and the calculated connectivity probability is If it is more than the preset reference value, it is confirmed that there is a connection, or, as another example, using a pre-learned sequential language model, the connection between the preceding sentences (that is, between the preceding sentence and the corresponding summary sentence, and between the corresponding summary sentence and the following sentence) A first score corresponding to the probability is calculated, a weight corresponding to the number of key words included in a sentence to be connected (that is, each preceding sentence and a following sentence) is given, and the sum of the first score and the weight is a preset reference value If it is abnormal, it can be confirmed that there is connectivity.

3 is an exemplary view for explaining the summary sentence correction according to an embodiment of the present invention, as shown in the same figure, 'ⓐ [customer] Now look at U+ in the basement, there is a repeater box [ Counselor] Repeater [Counselor] Yes, ⓑ [Customer] But 00 Telecommunication companies and other phones all work well [Counselor] Um, Yes, ⓒ[Customer] This is only LG’, and output a summary sentence out of three sentences of ⓐⓑⓒ. If the summary sentence output from the part 17 is ⓑ, and the preceding and following sentences for the summary sentence ⓑ are respectively ⓐⓒ, the summary sentence correction unit 18 checks and corrects the connectivity between the preceding and following sentences as follows. can For example, the score calculated in response to the connectivity probability between the preceding sentences and the following sentences obtained using the pre-learned sequential language model, that is, the score calculated in response to 50% of the connectivity probability between the preceding sentence ⓐ and the summary sentence ⓑ. 5 points, the score calculated in response to 40% of the probability of connection between the summary sentence ⓑ and the trailing sentence ⓒ is 4 points and the corresponding weights are 0 and 3, respectively, the connectivity score between the preceding sentence ⓐ and the summary sentence ⓑ (that is, the sum of the corresponding connectivity probability score 5 and the weight score 0) is 5+0=5 point, and the connectivity score between the summary sentence ⓑ and the trailing sentence ⓒ (that is, the sum of the corresponding connectivity probability score 4 and weight score 3) is 4 + 3 = 7 points, so assuming that the preset connectivity standard score is 6 points, By confirming that there is a connection between the summary sentence ⓑ and the trailing sentence ⓒ, the summary sentence can be corrected with the sentence 'But SK and other phones all work well, but only LG'.

In this embodiment, the language model may include a Long Short-Term Memory Network (LSTM).

1 again, the summary sentence output/evaluation unit 19 is for finally outputting and evaluating a summary sentence including one or more summary sentences outputted for each paragraph through the summary sentence correction unit 18, for example, finally The summary sentence extraction/generation model of the summary sentence output unit 17 when the key words in the correct answer are included in the summary sentence in a number less than the standard by evaluating the degree of agreement between the output summary sentence and the key words of the counseling memo used as the correct answer can be used for re-learning.

4 is an exemplary view of the input and output of the device of FIG. 1 according to an embodiment of the present invention. As shown in the same figure, the input is the consultation content and the consultation memo, and the output for this is the consultation content based on the consultation memo. This is a summary extracted from As an example, for each key word (paragraph by subject), 1/10 of the paragraph may be extracted and summarized.

The evaluation (verification) method in the summary sentence output evaluation unit 19 is, for example, a qualitative method of O/X checking according to whether the final output summary matches the topic/contents thought by the person or has no relation at all, and the core The number of words may include a quantitative method calculated by 'the number of matches between the key word and the summary word / the number of key words'.

5 is an exemplary view of quantitative evaluation according to an embodiment of the present invention, for example, included in the counseling memo as a correct answer memo by the counselor from the counseling content of call 001 as illustrated in FIG. 5 (a). When there are 8 key words, as illustrated in (b) of FIG. 5 , summary sentence 1 matches 2 words out of 8 key words, and is evaluated as 2/8x100=25 points, and summary sentence 2 is evaluated as 8 key words out of 8 key words. Five words can be matched to be rated 5/8x100 = 62.5 points.

6 is a flowchart of a customer consultation summary method according to an exemplary embodiment of the present invention. Since it is applied to the customer consultation summary apparatus described with reference to FIGS. 1-5, it will be described in parallel with the operation of the corresponding apparatus.

First, through the transcription text input unit 11, the consultation transcript corresponding to the customer consultation content such as VoC (Voice of Customer) is input (S601), and the consultation memo text directly summarized by the counselor through the consultation memo input unit 13 When is input (S603), the paragraph separation unit 12 divides the counseling transcript inputted in step S601 into one or more paragraphs for each topic of conversation (S605).

Then, the morpheme analysis unit 14 receives the sentences of the paragraphs for each conversation topic output and inputted through the paragraph separation unit 12 in step S605 and the counseling memo sentences inputted through the counseling memo input unit 13 in step S603, respectively. Separately, morpheme analysis is performed to separate main words, endings, and propositions (S607).

Next, the normalization unit 15 normalizes the words of the sentences of the paragraphs for each conversation topic outputted through the morpheme analysis unit 14 and the words of the sentences of the counseling memo separately in a predetermined form to reduce the total number of characters (S609) .

Next, the key word extraction unit 16 extracts the key words that must be included in the correct answer sentence (consultation memo) from the consultation memo data output through the normalization unit 15 in step S609 (S611)

Next, the summary sentence output unit 17 receives each paragraph (paragraph by conversation topic) and counseling memo of the counseling transcript outputted through the normalization unit 15 in step S609, respectively, and inputted from the inputted conversation topic paragraph A sentence similar to the counseling memo is extracted or generated and output as a summary sentence, that is, a summary sentence is created and outputted for each paragraph separated by conversation topic through the paragraph separator 12 (S613).

Then, the summary sentence correction unit 18 is output through the summary sentence output unit 17 in step S613 and the leading and trailing sentences in the counseling transcript for the input summary sentence and the key word extraction unit 16 in step S611. The summary sentence is corrected based on at least one of the key words output and input, and since the detailed correction process is the same as described above, a corresponding description is omitted (S615).

Finally, the summary sentence output/evaluation unit 19 finally outputs and evaluates the summary sentences composed of one or more summary sentences output for each paragraph for each conversation topic through the summary sentence correction unit 18 in step S615, for example, finally The summary sentence output unit 17 extracts/generates the summary sentence output unit 17 when the key word in the correct answer is included in the summary sentence in a number less than the standard by evaluating the degree of matching between the summary sentence output as the output and the key words of the counseling memo used as the correct answer Re-learning may be performed through the model (S617).

Meanwhile, according to the above-described customer consultation summary method, it is possible to implement a computer-readable recording medium in which a program for executing the method in a computer is recorded.

On the other hand, according to the above-described customer consultation summary method, it is possible to implement an application stored in a computer-readable recording medium in order to execute the method in combination with hardware.

On the other hand, according to the above-described customer consultation summary method, it is possible to implement a computer program stored in a computer-readable recording medium in order to execute the method on a computer.

For example, as described above, the customer consultation summary method according to an exemplary embodiment of the present invention is a computer-readable recording medium including program instructions for performing various computer-implemented operations or an application stored in the recording medium. can be implemented. The computer-readable recording medium may include program instructions, local data files, local data structures, etc. alone or in combination. The recording medium may be specially designed and configured for the embodiment of the present invention, or may be known and used by a person skilled in the art of computer software. Examples of the computer-readable recording medium include hard disks, magnetic media such as floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floppy disks, and ROMs, RAMs, flash memories, and the like. Hardware devices specially configured to store and execute the same program instructions are included. Examples of program instructions may include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

The above description is merely illustrative of the technical spirit of the present invention, and various modifications and variations will be possible without departing from the essential characteristics of the present invention by those skilled in the art to which the present invention pertains. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical spirit of the present invention, but to explain, and the scope of the technical spirit of the present invention is not limited by these embodiments. The protection scope of the present invention should be construed by the following claims, and all technical ideas within the equivalent range should be construed as being included in the scope of the present invention.

11: Recorded text input unit
12: Paragraph division
13: Counseling memo input unit
14: morpheme analysis unit
15: normalization unit
16: Key word extraction unit
17: summary sentence output unit
18: Summary sentence correction department
19: Summary text output/evaluation unit

Claims (21)

a paragraph separation unit for dividing the counseling transcript into paragraphs by topic;
a summary sentence output unit for extracting or generating a sentence similar to a counseling memo from each of the separated paragraphs and outputting it as a summary sentence; and
and a summary sentence correction unit for correcting the output summary sentence based on at least one of a preceding sentence and a trailing sentence for the output summary sentence and a preset key word;
The paragraph dividing unit obtains the probability of belonging to each counseling type for each sentence of the counseling transcript, and if the sum of the probabilities of the counseling type of the same counseling type for N adjacent sentences is equal to or greater than a preset reference value, the counseling type is the first or the previous counseling type and, if different, customer consultation summary device, characterized in that the paragraphs of the conversation topic corresponding to the corresponding consultation type are separated. delete According to claim 1,
The paragraph separation unit is a customer counseling summary device, characterized in that by using the learned classification model to obtain the counseling type classification probability for each sentence. 4. The method of claim 3,
The classification model is a customer consultation summary device, characterized in that it includes at least one of an artificial intelligence network and a support vector machine (SVM). According to claim 1,
The summary sentence correction unit confirms the connectivity of the preceding sentence and the following sentence with respect to the output summary sentence, respectively, and connects the connected sentence to the output summary sentence to correct it. 6. The method of claim 5,
The summary sentence correction unit calculates the probability of connectivity between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence) using a pre-learned sequential language model. And when the calculated connectivity probability is greater than or equal to a preset reference value, customer consultation summary device, characterized in that it is confirmed that there is connectivity. 6. The method of claim 5,
The summary sentence corrector corresponds to the probability of connectivity between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence) using a pre-learned sequential language model. calculates a first score, gives a weight corresponding to the preset number of key words included in a sentence to be connected (that is, each of the preceding sentence and the following sentence), and the sum of the first score and the weight is Customer consultation summary device, characterized in that it is confirmed that there is connectivity when it is more than a preset reference value. 8. The method of claim 7,
The key words are customer consultation summary device, characterized in that extracted from the consultation memo. 8. The method of claim 6 or 7,
and the language model includes a Long Short-Term Memory Network (LSTM). A customer service summary method performed by a customer service summary device, the method comprising:
(a) separating the counseling transcript into paragraphs for each topic of conversation;
(b) extracting or generating a sentence similar to a counseling memo from each of the separated paragraphs and outputting it as a summary sentence; and
(c) correcting the output summary sentence based on at least one of a preceding sentence and a trailing sentence for the output summary sentence and a preset key word;
In the step (a), the probability of belonging to each counseling type for each sentence of the counseling transcript is obtained, and the sum of the probabilities of the counseling type of the same counseling type for N adjacent sentences is equal to or greater than a preset reference value, and the corresponding counseling type is the first or preceding When different from the consultation type, the customer consultation summary method, characterized in that separated into paragraphs of the conversation topic corresponding to the consultation type. delete 11. The method of claim 10,
The step (a) is a customer counseling summary method, characterized in that by using the learned classification model to obtain the counseling type classification probability for each test sentence. 13. The method of claim 12,
The classification model includes at least one of an artificial intelligence network and a support vector machine (SVM). 11. The method of claim 10,
In the step (c), the connection of the preceding sentence and the following sentence with respect to the output summary sentence is checked, respectively, and the connected sentence is connected to the output summary sentence and corrected. 15. The method of claim 14,
In step (c), the probability of connectivity between the preceding sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence) using a pre-learned sequential language model is determined. Customer consultation summary method, characterized in that the calculation and confirming that there is connectivity when the calculated connectivity probability is greater than or equal to a preset reference value. 15. The method of claim 14,
The step (c) uses a pre-learned sequential language model to determine the connectivity probability between preceding and following sentences (ie, between the preceding sentence and the output summary sentence, and between the output summary sentence and the following sentence). A corresponding first score is calculated, a weight corresponding to the preset number of key words included in a sentence to be connected (that is, each of the preceding sentence and the following sentence) is given, and the sum of the first score and the weight Customer consultation summary method, characterized in that it is confirmed that there is connectivity when it is more than this preset reference value. 17. The method of claim 16,
The key words are customer consultation summary method, characterized in that extracted from the consultation memo. 17. The method of claim 16,
wherein the language model comprises a Long Short-Term Memory Network (LSTM). A computer-readable recording medium recording a program for executing the customer consultation summary method of any one of claims 10 and 12 to 18 on a computer. An application stored in a computer-readable recording medium for executing the customer consultation summary method of any one of claims 10 and 12 to 18 in combination with hardware. A computer program stored in a computer-readable recording medium in order to execute the customer consultation summary method of any one of claims 10 and 12 to 18 on a computer.

Images