KR20180055680A - Method of providing health care guide using chat-bot having user intension analysis function and apparatus for the same - Google Patents

Abstract

A method for extracting user intention using a chat-bot and recommending a health care instruction sentence based on the extracted user intention, and a computer device and a recording medium therefor are provided. The method includes the steps of: learning a document intention qualification extraction neural network model using examples of a health care instruction document and user intention information; and recommending a health care instruction by receiving the user intention through the chat-bot. The health care instruction suited to the user′s intention economically can be provided by extracting a meaning according to the intention of a dialogue using the chat-bot automated by learning based on the health care instruction and various intention information and providing the health care instruction suitable for a user situation.

Description

TECHNICAL FIELD [0001] The present invention relates to a method for providing healthcare instructions using a chat bot having a user's intention extracting function, and a device for the same. [0002]

The present invention relates to a technology for extracting an intention within a conversation of a user through a chat robot (shortly referred to as a " chat-bot ") and recommending a sentence of a health care service recommendation instruction.

Chat has long been used by users on the Internet to exchange messages with devices such as computers. Since the development of mobile phones and mobile devices has overcome the space limitations of Internet access, chat has become available wherever there is a device that can connect to the Internet.

Generally speaking, chat is a premise of a conversation between a person and a person. There is also a desire to be able to chat even when there is no other person (person) to chat with. Chat bot has been developed as a means to satisfy this demand. Chat bot has the function of recommending and expressing the appropriate answer according to the input words, so that it can play the role of chat partner on behalf of the person.

But the chatter bots developed so far are at the level of responding with the key keywords of simple questions and prebuilt answer sentences. Therefore, it is difficult for conventional chat bots to extract consecutive conversation intentions and to give accurate answers according to the user 's intention.

On the other hand, there is an Internet health counseling system that connects the health counselor with the medical staff using the Internet. In general, the Internet health counseling system can save time and money by reducing the trouble for the health consultant to go to the hospital. However, there is a disadvantage that professional medical staffs must wait for real time consultation.

The present invention provides a healthcare guidance providing method using a chat bot that can analyze and extract a user's intention from a continuous conversation with a user and provide an optimal healthcare guidance to a user based on the intention of the user, .

It is another object of the present invention to provide a computer readable recording medium on which a computer program for executing the above healthcare guidance providing method on a computer is recorded.

The object of the present invention is not limited to the above-described embodiments, but may be variously modified without departing from the spirit and scope of the present invention.

According to embodiments of the present invention for achieving the above object, there is provided a method for providing healthcare guidance using a chat bot having a user dialogue intention extracting function. The method includes: learning a document intentionally extracted neural network model based on an example of a healthcare guidance document and user intent information; And recommending health care instructions by inputting user intent via the chat bot.

According to one embodiment, the 'learning step' includes the steps of: dividing the natural language instruction document and the user's intention information into an eighth unit text; Extracting each of the divided unitary texts as input qualities based on a previously learned word-level neural network model; Learning a document intentionally extracted neural network model based on the extracted input qualities; And storing the learned guide document qualities in a guidance document database (DB).

According to an exemplary embodiment, learning the document intentionally extracted neural network model may include learning a document intentionally extracted neural network model using the extracted input qualities and target intention similarity.

According to an embodiment, the step of storing the instruction in the instruction document DB may include storing the instruction document qualities of the extracted natural language instruction document and the text of the natural language instruction document together.

According to one embodiment, the 'recommending step' may include: receiving the message of the user by the chat bot; Extracting a message of an input user as an input quality; Recognizing the user intention using the extracted input quality and extracting the user intention; Searching the retrieved user's intention qualities for a healthcare guidance document based on the similarity analysis; And recommending the guided document to the user.

According to one embodiment, the user's message may include at least one of a dialogue expressed in voice and a sentence input in the form of text.

According to an embodiment, the step of extracting with the user's intentionality may include extracting the intent of the user using the learned intentional attribute extraction neural network model based on the extracted input attribute.

According to an embodiment, the similarity analysis may be performed by comparing the user's intention qualities with previously generated guide document qualities existing in the guide document DB storing the guide document qualities learned in the learning step, And a guidance document with the most similar document document qualities.

According to another aspect of the present invention, there is provided a method for providing a health care instruction using a chat bot having a user dialogue intention extraction function according to any one of the above methods, A computer-readable recording medium is provided.

According to embodiments of the present invention, a computer device for a healthcare guidance service is provided. The computer device for the healthcare guidance service includes a data store, a processor, and a memory. The data storage unit stores a computer program and data for executing a healthcare guidance providing method using a chat bot having a user dialogue intention extraction function according to any one of the above methods. The processor reads the computer program from the data storage device and executes the computer program. The memory provides a data processing workspace for the processor. The processor analyzes the conversation between the counselor and the counselor by executing the computer program, extracts the conversation intention, and provides the appropriate health care instructions.

According to one embodiment, the computer device may be implemented as a cloud server device.

According to the present invention, an automated chat bot can be realized by learning based on healthcare instructions and various intention information. It is possible to extract the meaning of the user's utterance according to the intention of the user from the conversation with the user using the chat bot, and to provide the health care instruction suited to the user's situation. Through this, it is possible to economically provide healthcare guidelines that are tailored to the user's intent, thereby providing customized health care services.

FIG. 1 is a flowchart illustrating a method of extracting a user's intention using a chat bot and using the extracted intention for healthcare sentence recommendation according to an embodiment of the present disclosure.
FIG. 2 is a detailed flowchart of 'learning using a healthcare guidance document and an example of user's intention information' according to an embodiment of the present invention.
FIG. 3 is a block diagram illustrating a system structure in which a step of recommending a health care instruction by inputting a user intention through a chat bot is operated according to an embodiment of the present invention.
FIG. 4 is a detailed flowchart illustrating a step of recommending a health care instruction by inputting a user intention through a chat bot according to an embodiment of the present invention.
FIG. 5 is a block diagram illustrating a system structure in which a 'step of recommending a health care instruction by inputting a user intention through a chat bot' operates according to an embodiment of the present invention.
FIG. 6 illustrates a Korean sentence language model learning process based on a morphological embedding model according to an embodiment of the present invention.
FIG. 7 is a diagram exemplifying the concept when the present invention is applied to psychiatric consultation.

The following detailed description of the invention refers to the accompanying drawings, which illustrate, by way of illustration, specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different, but need not be mutually exclusive. For example, certain features, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in connection with an embodiment. It is also to be understood that the position or arrangement of the individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is to be limited only by the appended claims, along with the full scope of equivalents to which such claims are entitled, if properly explained. In the drawings, like reference numerals refer to the same or similar functions throughout the several views.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries should be construed as meaning consistent with meaning in the context of the relevant art and are not to be construed as ideal or overly formal in meaning unless expressly defined in the present application .

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 illustrates a method of extracting a user's intention using a chat bot and using the extracted intention in healthcare sentence recommendation according to an embodiment of the present disclosure.

Referring to FIG. 1, a method of extracting an intention of a user using a chat bot and using the extracted intention for healthcare sentence recommendation is largely classified into a health guidance document and an example of user intention information, (S100), and recommending a health care instruction by inputting a user intention through a chat bot (S200).

Step S100 of learning a document intentionally extracted neural network model using an example of a healthcare guidance document and user intention information is a process of learning an inference model using a predefined document of a health care instruction and intention information data. A Healthcare Guide document might be, for example, a document that contains healthcare information such as "If you get a cold, you should have a ginger tea." Ginger tea is made by pouring ginger and honey and pouring hot water. Intentional information can be, for example, natural language sentences such as "I feel like my body is cold," "I feel dizzy and feverish," and may contain documents related to healthcare documents.

Step S200 of recommending a health care instruction by inputting a user intention through a chat bot is a process of recommending a health care instruction by inputting a user's intention using a learned model and a chat bot. After inputting the user chat text through the user interface, the health care guidance document is searched through the inference process and recommended to the user. For example, when a user enters the sentence "What to do when the fever is cold and your body is cold," you can use a model that was learned earlier: "If you get a cold, you may want to take a ginger tea. Please pour hot water and make it. "

FIG. 2 illustrates a concrete procedure of 'learning using a healthcare guidance document and an example of user intention information (S100)'. FIG. 3 illustrates a system structure in which a step of recommending a health care instruction (S110) is operated by receiving a user intention through a chat bot according to an embodiment of the present invention.

Referring to FIG. 2, a step 100 of learning using a healthcare instruction document and an example of user's intention information includes a step (S200) of dividing a natural language instruction document and user's intention information into a word unit text (S200) A step S210 of extracting an eardrum unit text as an input qualification, a step S220 of learning a document intentionality extraction neural network model based on the extracted input qualities, and a step of storing the learned guide document qualities in a database Step S230.

The step of learning using the example of the healthcare guidance document and the user's intention information (100) may be learned by further utilizing the target intention similarity.

A step S200 of extracting the input text such as the natural language instruction document and the user's intention information as the eccentricity unit text will be described in more detail with reference to FIG. The input text can be divided into one or more phrases based on division conditions such as spacing, space, and the like. For example, if you divide the input text "Good for a ginger tea when you get a cold" on a space-based basis, one or more segmented phrases can be "cold", "hung", "when" If you eat, you can become "and" good ". Also, for example, if you split the input text "I am sick and feverish" based on spacing, one or more of the phrases may be "throat", "scolding", "heat", " Quot ;, "same ".

Step S210 of extracting the segmented unit-of-speech text as an input character may represent each of the unit-of-speech quality unit texts extracted through the word segmentation in step S200 as input qualities using a previously learned word-level neural network model. In the above example, the words "cold", "caught", "when", "ginger tea", "eat" and "good"

The learning of the neural network based on the extracted input qualities (S220) learns the document intentionally-extracted neural network model 310. That is, the document intentionality extraction neural network model 310 is learned using the extracted input qualities and the target intention similarity 340. In this step, the neural network is learned by using the target document similarity (330) derived from the neural network and the target similarity (340) of the user's intention characteristic (350).

Step S230 of storing the learned guide document qualities in the DB is performed by inferring the guide document qualities 330 of all guide documents after the learning of the document intentionally extracted neural network model, And storing them together.

A method of learning the guidance document and the user's intention statement (information) shown in FIG. 2 and storing the result of learning in the guidance document DB 320 will be described in more detail by way of example.

Korean morphological embedding techniques can be used to represent Korean sentences as real numbers. The morphological embedding method is useful for learning the embedding model for each morpheme in accordance with the characteristics of the Korean morpheme. In English, words basically use the stem independently or in combination with the affix to modify the verb. On the other hand, in Korean, only vocabulary morphemes corresponding to stem are meaningful, and words are composed of a combination of grammatical morphemes corresponding to the mother or the survey. Therefore, morphological embedding technology extracts only vocabulary morphemes from the morpheme level. An example sentence: "These models are a thin two-layer neural network, analyzing the morpheme of" from network "to" net ", for example, combining the lexical morpheme" net " It can be analyzed as a phrase. Among them, only the vocabulary term "network" can be extracted and studied.

In addition, the types of morphemes, the number of shoulders (meaning codes), and grammatical morphemes corresponding to the endings or surrogates combined with the vocabulary morpheme can be studied together as learning qualities. If these qualities are not used as learning qualities of a vocabulary morpheme, morphemes appearing at similar positions within sentences can be learned as similar vectors. In this case, the sentence is different or the opposite word is learned as a morpheme with a similar vector, which may affect the performance when extracting the qualification or generating the sentence.

In an exemplary embodiment, the language model learning method of Korean morphological embedding may be as follows. For example, sentences in Korean Wikipedia documents are analyzed by morpheme analyzer, and corpus is generated by tagging the type, the shoulder number (semantic code) of each morpheme, and combined endings and studies as learning qualities . Then, words with similar context information and antonyms are refined using language knowledge. Then, we can learn the sentence of corpus by using Deep Belief Network (DBN) model which is one kind of deep neural network. When learning the sentence of a corpus, the activation function can use ReLU and apply Dropout.

Since the sentence is a sequential arrangement of words, it is desirable to learn while considering the context information input from the front of the sentence. To do this, we can use a model that can receive input from the hidden layer of the neural network and process the input data and the previously learned context unit. In the exemplary embodiment, since the sentence is a sequential list of words, it may be advantageous to use a recurrent neural network (RNN) model that is a suitable model for learning time series data. Deep neural networks in the RNN family include Long Short Term Memory (LSTM), which solves gradient vanishing problems, and gated recurrent units (GRUs), which simplify them.

In an exemplary embodiment, a method of learning the language model including the semantic information of the sentence using the morphological embedding result of the vocabulary morpheme and encoding the Korean sentence can be used. In this regard, FIG. 6 illustrates a Korean sentence language model learning process based on a morphological embedding model in an exemplary embodiment of the present invention.

Referring to FIG. 6, in order to learn a Korean sentence language model based on a morphological embedding model, a Korean corpus to be modeled is first selected. The corpus of this corpus is analyzed by morpheme analyzer, and the results of word resolution are annotated according to morphological representation method in embedding morpheme with tags of learning qualities. The shoulder number and type of morpheme can be annotated with morphemes as morphemes. For example, you can use Wikipedia's dump data (: database_download), which is a Korean web encyclopedia. In Korean Wikipedia dump data (as of June 2016), you can construct a corpus by parsing only the title and text of the text. According to the inventors' actual practice, 2,619,773 sentences were parsed when only numbers and symbols, English and Korean sentences were refined. And the morpheme analysis result was analyzed as 77,172,818 morphemes.

The morphological vectors of the corresponding morpheme are retrieved from the result of performing morphological embedding with the precedence learning. The loaded morpheme vector is learned by input and output to the RNN model. Where the input layer of the RNN model is the morpheme embedding vector of the current morpheme and the output layer is the morpheme embedding vector of the next morpheme. Therefore, when learning the first morpheme, as shown in FIG. 6, the morphological embedding vector corresponding to "Islam 00NNG" is inputted and the RNN is learned by outputting a morpheme embedding vector corresponding to "00MAJ". In an exemplary embodiment of such learning, for example, a 100-dimensional morphological embedding vector may be used. In this way, it learns the Korean Wikipedia corpus sentences annotated with morpheme qualities. For the experimental learning using the GRU model, for example, 128 sentence lengths were used. This is the average length of the scholastic quiz corpus prepared for the evaluation corpus. Finally, the parameters of the GRU in / out / forget gate are stored and stored in the learning model.

Next, FIG. 4 is a detailed flowchart of 'step S110 of recommending a health care instruction by inputting a user intention through a chat bot' according to an exemplary embodiment of the present invention. FIG. 5 is a block diagram illustrating a system structure in which 'recommending a health care instruction by inputting a user intention through a chat bot (S110)' according to an exemplary embodiment of the present invention operates.

Referring to FIGS. 4 and 5, in step S110, a step of recommending a health care instruction by inputting a user's intention through a chat bot is performed by a chat bot, And receiving an input (S300). In step S300 of receiving the user's conversation in the chat bot, the user's message input 510 can be input in real time through the user interface. The user's message can be a dialogue expressed in voice, a sentence input in the form of text, and the like. The chat bot may include user interface means capable of interfacing such input.

The step of recommending the health care instructions by inputting the intention of the user through the chat bot (S110) may include extracting the natural language user input (510) as input quality (S310). In step S310 of extracting the natural language user input to the input feature 520, for example, when a text such as "a fever occurs and a body feels frustrated" is inputted, the text is classified according to a segmentation condition such as space, You can divide the text into units of text, for example, "text", "text", "body", "text", and "text". The input feature 520 can be extracted using the already-learned word unit neural network model for the generated eulogy unit text. The method of extracting the input qualities has been described in detail in the foregoing, and the contents thereof can be referred to.

In addition, 'recommending health guidance by inputting user's intention through a chat bot' (S110) is a step of recognizing the user intention based on the document intention characteristic extraction neural network model 530 using the extracted input feature, And extracting the attribute 540 (S320). In step S320, the user intention attribute 540 can be extracted using the learned intentional attribute extraction neural network model 530 based on the input attribute 520 extracted in the previous step S310.

The step of recommending the health care instructions by inputting the intention of the user through the chat bot (S110) may further include a step (S330) of searching the healthcare guidance document through the similarity analysis of the user's intention (540) And recommending the guidance document to the user (S340). For this, a similarity analysis model 550 can be used. The similarity analysis model 550 performs a step S330 of searching the health guidance document 570 through the similarity analysis of the user's intention characteristic 540. The similarity analysis model 550 compares the current user intention qualities 540 such as [1.0, 0, 0, 0.5] with pre-generated guide document qualities existing in the guidance document DB 560, A guidance document 570 having the most similar document document qualities (e.g., [0.8, 0, 0, 0.3]) can be found. The guidance document 570 thus searched can be delivered to the user through the user interface in real time and recommended.

Meanwhile, the method of providing health care instructions according to the embodiments of the present invention described above can be implemented by a computer program. The computer program may be implemented using hardware resources of a chat bot having a user dialogue intention extraction function. That is, the chat bot basically has a storage means such as a hard disk or a nonvolatile memory device, a computing device such as a CPU or a microprocessor that performs computation processing for executing a program, and a user for interactive communication with the user And an interface means (an input / output device such as a speaker and a monitor and a drive device for driving the input / output device). A computer program embodying the method of providing healthcare instructions may be stored in such storage means and the computing device may execute the program to provide a healthcare guidance providing service if necessary.

Meanwhile, the method of providing healthcare guidance according to the present invention can be applied to health consultation between a specialist and a patient. FIG. 7 is a diagram illustrating a concept of applying the healthcare guidance providing method according to the embodiments of the present invention to psychiatric consultation.

Referring to FIG. 7, the method for providing healthcare instructions according to the present invention can be implemented to be serviced in the cloud server system 700. The computer program implementing the healthcare guidance providing method according to the present invention can be loaded and executed in the cloud server 710. [ Psychiatrists (eg, psychiatrists), counselors (users) A and B can access the cloud server 710 through a communication network using their computer devices 720, 730 and 740. Various hardware (CPU, memory, storage means, communication means, etc.) and software resources (computer program, various data, database system, etc.) necessary for carrying out the above-described healthcare guidance providing method can be installed in the cloud server 710 have.

The computer program may be provided to the user's computer devices 720, 730, and 740 such as a consultant and a consultant in the form of a web page by the cloud server 710 executing an interactive platform for consultation. Users can enter a conversation (text, voice, video, etc.) for consultation through the interactive platform for consultation in the form of a web page.

The computer program implementing the healthcare guidance providing method according to the present invention can exhibit functions such as supporting conversation while analyzing and learning consultation contents, diagnosing symptoms, and suggesting solutions.

The interactive platform for consultation provided by the cloud server 710 may be serviced through an application for a mobile terminal.

As described above, the present invention can learn a document analysis model that can extract a user's intentional intention from a conversation between a user and a chat bot and select a response sentence corresponding thereto. The present invention can also be used in a variety of professional counseling fields such as psychiatric consultation between psychiatrists and patients.

The features, structures, effects and the like described in the embodiments are included in one embodiment of the present invention and are not necessarily limited to only one embodiment. Furthermore, the features, structures, effects and the like illustrated in the embodiments can be combined and modified by other persons skilled in the art to which the embodiments belong. Therefore, it should be understood that the present invention is not limited to these combinations and modifications.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of illustration, It can be seen that various modifications and applications are possible. For example, each component specifically shown in the embodiments can be modified and implemented. It is to be understood that all changes and modifications that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

300: User Intention-Based Healthcare Guidance Recommendation System
310, 530: Document Intentional Extraction Neural Network Model
320, 560: Guidance document database (DB) 330: Guidance document qualifications
340: Target intention similarity 350, 540: User Intent
510: User input sentence 520: Input sentence qualification
550: Similarity Analysis Model 570: Guidance Document

Claims (11)

Learning a document intentionally extracted neural network model based on an example of a healthcare guidance document and user intention information; And
A step of recommending a health care instruction by inputting a user intention through a chat bot; and a method for providing health care instructions using a chat bot having a user dialogue intention extracting function. The method of claim 1, wherein the step of learning comprises: dividing the natural language instruction document and the user's intention information into text units of an eighth word; Extracting each of the divided unitary texts as input qualities based on a previously learned word-level neural network model; Learning a document intentionally extracted neural network model based on the extracted input qualities; And storing the learned guide document qualities in a guide document database (DB). 3. The method of claim 2, wherein learning the document intentionally extracted neural network model comprises learning the document intentionally extracted neural network model using the extracted input quality and the target similarity degree of the target intention. A method of providing healthcare instructions using a health chat bot. 3. The method according to claim 2, wherein the step of storing the instruction in the instruction document DB includes storing the instruction document qualities of the extracted natural language instruction document and the text of the natural language instruction document together. How to provide healthcare guidelines using. The method of claim 1, wherein the recommending step comprises:
Receiving a user's message from the chat bot; Extracting a message of an input user as an input quality; Recognizing the user intention using the extracted input quality and extracting the user intention; Searching the retrieved user's intention qualities for a healthcare guidance document based on the similarity analysis; And recommending the searched guide document to the user. The method of claim 1, further comprising: 6. The method according to claim 5, wherein the user's message includes at least one of a dialogue expressed in voice and a sentence input in the form of text. How to provide instructions. 6. The method according to claim 5, wherein the step of extracting with the user's intention qualities extracts the intention characteristics of the user using the learned intentional feature extraction neural network model based on the extracted input features. How to Provide Healthcare Guidance Using Functional Chat Bot. The method according to claim 5, wherein the similarity analysis is performed by comparing the user's intention qualities with previously generated guide document qualities existing in the guide document database storing the guide document qualities learned in the learning step, And a guidance document having a quality of a guidance document most similar to the guidance document. The method for providing healthcare guidance using a chat bot having a user dialogue intention extracting function. A computer-readable recording medium storing a computer program for executing a method for providing healthcare instructions using a chat bot having a user dialogue intention extracting function according to any one of claims 1 to 8, media. 8. A computer-readable recording medium storing a computer program for executing a method of providing a healthcare instruction using a chat bot having a user interaction intention extraction function according to any one of claims 1 to 8 and a data storage unit for storing data;
A processor for reading and executing the computer program from the data storage device;
And a memory for providing a data processing workspace to the processor,
Wherein the processor analyzes the conversation between the counselor and the counselor by executing the computer program to extract a conversation intention and provides a health care instruction suitable for the extracted conversation intention. 11. The computer device of claim 10, wherein the computer device is implemented as a cloud server device.

Images