A system that guides operators to ask questions to rescue callers

Abstract

Embodiments of the present application provide a system for guiding an operator to question a rescue sender. The system uses the positive example probability of any second reference group included in any first reference group among a plurality of first reference groups as a cut-off point to perform a classification test on any first reference group to obtain the classification sensitivity and classification specificity corresponding to any second reference group, and based on the classification sensitivity and classification specificity corresponding to a plurality of second reference groups within any first reference group, obtain a receiver operating characteristic curve corresponding to any first reference group. Based on the receiver operating characteristic curves corresponding to each first reference group within a plurality of first reference groups, determine a target first reference group from the plurality of first reference groups, and determine target question semantics from the distance between the coordinate point corresponding to any second reference group among the plurality of second reference groups and the perfect classification coordinate point.

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims priority to Chinese Patent Application No. 202210805597.1, entitled "SYSTEM FOR HELPING OPERATOR TO QUESTION HELP-SEEKER," filed with the China Patent Office on July 8, 2022, the entire contents of which are incorporated by reference.

[0002] The present application relates to the field of computers, and in particular to a system that guides an operator to ask questions of a rescue caller. The present application also relates to a system that guides an operator to determine whether a rescuee is a positive case. The present application further relates to a system that guides an investigator to ask questions of a researchee. The present application further relates to a system that guides an investigator to determine whether a researchee is a positive case. [Background technology]

[0003] Out-of-hospital cardiac arrest (OHCA) is the most serious clinical condition occurring outside of a hospital, and once it occurs, patients can die quickly. Although the golden time for rescue after cardiac arrest is only 3-5 minutes, the average response time of ambulances in China is usually more than 15 minutes, and on-site cardiopulmonary resuscitation (CPR) is the only hope for survival of OHCA patients.

[0004] In recent years, both domestically and internationally, emergency medical command center (EMC) dispatchers (e.g., emergency telephone number: 120) have attempted to encourage and guide emergency callers (callers) or other on-site personnel to provide CPR (i.e., telephone CPR) to OHCA patients when receiving distress calls. Dispatchers lacking specialized first aid knowledge often have low rates of telephone CPR implementation and success, and are prone to overlooking the diagnosis of OHCA patients. Therefore, technical issues that urgently need to be resolved include how to guide emergency callers in their questions and how to provide dispatchers with reference data to determine the on-site condition of the rescued person (e.g., whether the rescued person is suspected of having an OHCA) and the progress of the rescue provided by the emergency caller or other on-site personnel (e.g., whether the rescue caller initiated CPR, whether an automated external defibrillator (AED) was used, and whether return of spontaneous circulation (ROSC) was achieved after CPR). Summary of the Invention [Problem to be solved by the invention]

[0005] An embodiment of the present application provides a system that guides an operator to ask questions to a rescue caller, and is configured to help a dispatcher determine the on-site status of the rescued person (e.g., whether the rescued person is suspected of having an OHCA patient) and the progress of rescue by the rescue caller or other personnel at the scene for the rescued person (e.g., whether the rescue caller started CPR, whether an AED was used, whether return of spontaneous circulation (ROSC) was achieved after CPR, etc.). An embodiment of the present application also provides a system that determines the probability of a positive case for a rescued person, and is configured to help a dispatcher determine the on-site status of the rescued person (e.g., whether the rescued person is suspected of having an OHCA patient) and the progress of rescue by the rescue caller or other personnel at the scene for the rescued person (e.g., whether the rescue caller started CPR, whether an AED was used, whether return of spontaneous circulation (ROSC) was achieved after CPR, etc.). [Means for solving the problem]

[0006] An embodiment of the present application provides a system for guiding an operator to ask questions to a rescue caller, the system including: a first reference group acquisition module, a classification test module, a receiver operating characteristic curve acquisition module, a candidate question semantics acquisition module, and a target question semantics acquisition module, the first reference group acquisition module is configured to select a plurality of unprocessed first reference groups from an initial reference group, the first reference groups being acquired by classifying a plurality of historical rescue recipients in the initial reference group based on an entity-attribute cluster, the entity-attribute cluster being composed of a plurality of entity-attribute pairs corresponding to the same core question, the entity-attribute pair including an entity and an attribute, the entity representing the semantics of a question asked by an operator in a historical rescue call, and the attribute representing the semantics of an answer from a rescue caller to the question, the classification test module using a positive example probability of any second reference group included in any first reference group among the plurality of first reference groups as a cutoff point to select any of the first reference groups. the target question semantics acquisition module is configured to: perform a classification test on the group to obtain a classification sensitivity and a classification specificity corresponding to the arbitrary second reference group, wherein the cutoff point is for determining a positive example label of the first reference group, and the second reference group is obtained by classifying the first reference group based on an entity attribute; the receiver operating characteristic curve acquisition module is configured to obtain a receiver operating characteristic curve corresponding to the arbitrary first reference group based on the classification sensitivity and classification specificity corresponding to a plurality of second reference groups within the arbitrary first reference group; the candidate question semantics acquisition module is configured to determine a target first reference group from the plurality of first reference groups based on the receiver operating characteristic curve corresponding to each first reference group within the plurality of first reference groups, and to determine an entity of each entity-attribute pair in an entity-attribute cluster corresponding to the target first reference group as a candidate question semantic; and the target question semantics acquisition module is configured to:A target question semantic is determined from the candidate question semantic based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups on a receiver operating characteristic curve corresponding to the target first reference group to a perfect classification coordinate point, where the perfect classification coordinate point is a coordinate point on the receiver operating characteristic curve with an abscissa of 0 and an ordinate of 1, and the target question semantic is configured to guide the operator to ask a next question.

[0007] In some embodiments, the positive example probability of any second reference group included in any first reference group among the plurality of first reference groups is used as a cutoff point, and a classification test is performed on the any first reference group to obtain classification sensitivity and classification specificity corresponding to the any second reference group. members of a second reference group whose positive example probability in the first reference group is equal to or greater than the cutoff point as preset positive examples, and members of the second reference group whose positive example probability in the first reference group is less than the cutoff point as preset counterexamples; obtaining the classification sensitivity according to the number of members of the preset positive examples that are actually classified as positive examples and the number of members of the first reference group that are actually classified as positive examples; and obtaining the classification specificity according to the number of members of the preset counterexamples that are actually classified as counterexamples and the number of members of the first reference group that are actually classified as counterexamples.

[0008] In some embodiments, obtaining a receiver operating characteristic curve corresponding to any of the first reference groups based on the classification sensitivity and classification specificity corresponding to a plurality of second reference groups within the arbitrary first reference group includes: obtaining a coordinate point corresponding to any of the second reference groups among the plurality of second reference groups, using the classification sensitivity of the arbitrary second reference group as the ordinate and the absolute value of the difference between the classification specificity of the arbitrary second reference group and 1 as the abscissa; and connecting the coordinate points corresponding to each second reference group within the first reference group to obtain a receiver operating characteristic curve corresponding to the arbitrary first reference group.

[0009] In some embodiments, determining a target first reference group from the plurality of first reference groups based on a receiver operating characteristic curve corresponding to each first reference group in the plurality of first reference groups includes calculating an area under the curve of a receiver operating characteristic curve corresponding to each first reference group in the plurality of first reference groups, and using the first reference group corresponding to the largest area under the curve as the target first reference group.

[0010] In several embodiments, determining a target question semantic from the candidate question semantic based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups to a perfect classification coordinate point on a receiver operating characteristic curve corresponding to the target first reference group includes calculating a plurality of distances from a coordinate point corresponding to each second reference group among the plurality of second reference groups to a perfect classification coordinate point, and determining the second reference group corresponding to the shortest distance among the plurality of distances as the target second reference group, and determining an entity of an entity-attribute pair corresponding to the target second reference group as the target question semantic.

[0011] In some embodiments, the initial reference group is obtained by extracting personal characteristic information of the rescuee from the information said by the rescue caller, and determining the initial reference group based on the personal characteristic information.

[0012] One embodiment of the present application provides a system for guiding an operator to determine whether a rescuee is a positive case, the system including an entity attribute cluster and entity attribute pair determination module, a first reference group and a second reference group determination module, a preset positive case member number acquisition module, and a positive case probability determination module.The entity-attribute cluster and entity-attribute pair determination module is configured to determine an entity-attribute cluster currently corresponding to the rescuee and an entity-attribute pair currently corresponding to the rescuee based on a question of the operator in any dialogue between the operator and the rescue caller and an answer of the rescue caller to the question, wherein the entity-attribute cluster is composed of a plurality of entity-attribute pairs corresponding to the same core question, the entity-attribute pair includes an entity and an attribute, the entity represents the semantics of the question of the operator in a historical rescue call, and the attribute represents the semantics of the answer of the rescue caller to the question; the first reference group and second reference group determination module is configured to obtain a first reference group corresponding to the entity-attribute cluster and a second reference group corresponding to the entity-attribute pair from an initial reference group, wherein the first reference group is an entity the second reference group is obtained by classifying the first reference group based on an entity attribute cluster; the preset positive example member number acquisition module is configured to acquire the number of members of the first reference group who are preset positive examples using a positive example probability of the second reference group as a cutoff point, wherein the cutoff point is for determining a positive example label of the first reference group; the positive example probability determination module is configured to acquire the positive example probability of the rescuee based on a ratio of the number of members of the first reference group who are preset positive examples and are actually classified as positive examples to the number of members in the first reference group that are preset positive examples; and the positive example probability is configured to guide an operator to determine whether the rescuee is a positive example.

[0013] In several embodiments, the system further includes a positive example probability trend scatter plot acquisition module and a linear regression equation slope acquisition module, wherein the positive example probability trend scatter plot acquisition module is configured to acquire a positive example probability trend scatter plot based on multiple positive example probabilities of the rescuee obtained in multiple interactions between the operator and the rescue caller, and the linear regression equation slope acquisition module is configured to fit the positive example probability trend scatter plot using a linear regression equation and solve the linear regression equation to acquire the slope of the linear regression equation, wherein the positive example probability trend scatter plot and the slope of the linear regression equation are intended to represent the trend in the probability that the rescuee is a positive example.

[0014] In some embodiments, obtaining a positive example probability trend scatter diagram based on multiple positive example probabilities of the rescuee obtained in multiple interactions between the operator and the rescue caller includes obtaining coordinates on the positive example probability trend scatter diagram, with the number of interactions corresponding to an arbitrary positive example probability among the multiple positive example probabilities as the abscissa and the arbitrary positive example probability as the ordinate. In some embodiments, determining an entity-attribute cluster currently corresponding to the rescuee and an entity-attribute pair currently corresponding to the rescuee includes obtaining a question from the operator and extracting question semantics information from the question, obtaining an answer from the rescue caller to the question and extracting answer semantics information from the answer, determining a core question explained in the question and obtaining the entity-attribute cluster based on the core question, and determining the entity-attribute pair based on the question semantics information and the answer semantics information.

[0015] In some embodiments, obtaining the number of members of the first reference group that are preset positive examples using the positive example probability of the second reference group as a cutoff point includes determining the number of members of the second reference group that have a positive example probability in the first reference group that is equal to or greater than the cutoff point as the number of members of the preset positive examples.

[0016] In several embodiments, the system further includes a rate ratio acquisition module and a rate difference acquisition module, wherein the rate ratio acquisition module is configured to acquire a rate ratio based on the ratio between the current positive case probability of the rescuee and the positive case probability of the rescuee acquired in the previous interaction, and the rate difference acquisition module is configured to acquire a rate difference based on the difference between the current positive case probability of the rescuee and the positive case probability of the rescuee acquired in the previous interaction, wherein the rate difference and the rate ratio are configured to guide the operator to judge the value of the current positive case probability of the rescuee in determining whether the current rescuee is a positive case.

[0017] One embodiment of the present application provides a system for guiding an investigator to ask questions to an investigator, the system including a first reference group acquisition module, a classification test module, a receiver operating characteristic curve acquisition module, a candidate question semantics acquisition module, and a target question semantics acquisition module, the first reference group acquisition module is configured to select a plurality of unprocessed first reference groups from an initial reference group, wherein the first reference group is acquired by classifying a plurality of historical investigators in the initial reference group based on an entity-attribute cluster, the entity-attribute cluster being composed of a plurality of entity-attribute pairs corresponding to the same core question, the entity-attribute pair including an entity and an attribute, the entity representing the semantics of a question asked by an investigator in a historical investigation call, and the attribute representing the semantics of an answer given by the investigator to the question, the classification test module being configured to calculate a positive example probability of any second reference group included in any first reference group among the plurality of first reference groups at a cutoff point. the cutoff point is for determining a positive example label of the first reference group, and the second reference group is obtained by classifying the first reference group based on an entity attribute; the receiver operating characteristic curve acquisition module is configured to acquire a receiver operating characteristic curve corresponding to the first reference group based on the classification sensitivity and classification specificity corresponding to a plurality of second reference groups within the first reference group; the candidate question semantics acquisition module is configured to determine a target first reference group from the plurality of first reference groups based on the receiver operating characteristic curve corresponding to each first reference group within the plurality of first reference groups, and to determine an entity of each entity-attribute pair in an entity-attribute cluster corresponding to the target first reference group as a candidate question semantic; and the target question semantics acquisition module is configured to:and determining a target question semantic from the candidate question semantic based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups on a receiver operating characteristic curve corresponding to the target first reference group to a perfect classification coordinate point, wherein the perfect classification coordinate point is a coordinate point on the receiver operating characteristic curve with an abscissa of 0 and an ordinate of 1, and the target question semantic is configured to guide the researcher to ask a next question.

[0018] One embodiment of the present application provides a system for guiding an investigator to determine whether an interviewee is a positive case, the system including an entity-attribute cluster and entity-attribute pair determination module, a first reference group and a second reference group determination module, a preset positive case member number acquisition module, and a positive case probability determination module, the entity-attribute cluster and entity-attribute pair determination module being configured to determine an entity-attribute cluster currently corresponding to the interviewee and an entity-attribute pair currently corresponding to the interviewee based on the interviewer's questions and the interviewee's answers to the questions in any interaction between the investigator and the interviewee, wherein the entity-attribute cluster is composed of multiple entity-attribute pairs corresponding to the same core question, the entity-attribute pair including an entity and an attribute, the entity representing the semantics of the interviewer's questions in a history survey call, and the attribute representing the semantics of the interviewee's answers to the questions, the first reference group and the second reference group determination module being configured to determine the entity-attribute cluster and entity-attribute pair currently corresponding to the interviewee based on the questions of the interviewer in any interaction between the investigator and the interviewee, the positive example probability determination module is configured to obtain a positive example probability of the first reference group based on a ratio of the number of members of the first reference group that are actually classified as positive examples to the number of members of the first reference group that are preset positive examples, the positive example probability being configured to guide an investigator to determine whether the investigator is a positive example. [Effects of the Invention]

[0019] In an embodiment provided by the present application, the entity-attribute clusters are composed of entity-attribute pairs with the same subject in historical emergency calls, where the entity-attribute pairs are composed of the operator's question semantics and the emergency caller's answer semantics, structuring a large number of chaotic historical conversations in the emergency dispatch database, thereby classifying the historical rescuer subjects according to the historical conversations to obtain a first reference group and a second reference group. The embodiment provided by the present application performs a classification test on the first reference group by using the positive case probability of any second reference group as a cutoff point, thereby enabling continuous separation of the second reference group obtained by classification by entity attributes and calculating the first reference group obtained by classification based on the entity-attribute clusters. Furthermore, the classification sensitivity and classification specificity obtained by using the positive case probability of the second reference group as a cutoff point can more accurately reflect the effectiveness of the conversation represented by the entity-attribute pair corresponding to the second reference group in determining the situation at the scene of the rescuee and the rescue caller's rescue progress for the rescuee.

[0020] In the embodiment provided by the present application, a receiver operating characteristic curve corresponding to a first reference group is obtained based on the classification sensitivity and classification specificity corresponding to an arbitrary second reference group, and the effectiveness of multiple entity attribute clusters in determining the situation of the rescuee at the scene and the progress of the rescue by the rescue sender to the rescuee is quantified, and in this way, a core question of the entity attribute cluster representation useful for determining the situation of the rescuee at the scene and the progress of the rescue by the rescue sender to the rescuee can be obtained.

[0021] In an embodiment provided by the present application, the effectiveness of the entity-attribute pair in determining the situation at the scene of the rescuee and the progress of the rescue by the rescue sender to the rescuee is quantified by calculating the distance from the coordinate point corresponding to any second reference group among multiple second reference groups on the receiver operating characteristic curve corresponding to the obtained target first reference group to the perfect classification coordinate point, and in this way, the question semantics of the entity-attribute pair expression that is useful for determining the situation at the scene of the rescuee and the progress of the rescue by the rescue sender to the rescuee can be obtained.

[0022] In an embodiment provided by the present application, the entity attribute cluster currently corresponding to the rescuee and the entity attribute pair currently corresponding to the rescuee are determined through the operator's questions and the rescue caller's answers to those questions, and a first reference group corresponding to the entity attribute cluster and a second reference group corresponding to the entity attribute pair are obtained from the initial reference group, thereby obtaining the rescuee's historical reference group through dialogue. In an embodiment provided by the present application, the positive case probability of the second reference group is used as a cutoff point to obtain the number of members in the first reference group that are preset positive cases, and the positive case probability of the rescuee is obtained based on the ratio of the number of members in the first reference group that are preset positive cases actually classified as positive cases to the number of members in the first reference group that are preset positive cases, thereby making full use of the patient record resources of the rescue dispatch database and more accurately obtaining the probability of whether the rescuee is a positive case through dialogue between the operator and the rescue caller. [Brief explanation of the drawings]

[0023] The present application will be further illustrated by non-limiting exemplary embodiments, which are illustrated in detail in the drawings, in which like numerals represent like structures.

[0024] [Figure 1]FIG. 1 is a schematic diagram of an application scenario of a system that guides an operator to ask questions to a rescue caller and a system that guides the operator to determine whether a rescuee is a positive case, according to some embodiments of the present application. [Figure 2] 1 is an exemplary schematic diagram of a system for guiding an operator to ask a rescue caller questions, according to some embodiments of the present application. [Figure 3] 1 is an exemplary schematic diagram of a system for guiding an operator to determine whether a rescuee is a positive case, according to some embodiments of the present application. FIG. [Figure 4] 2 is an exemplary schematic diagram of an entity-attribute cluster, an entity-attribute pair, a first reference group, and a second reference group according to some embodiments of the present application; [Figure 5] FIG. 2 is an exemplary schematic diagram of a receiver operating characteristic curve shown in some examples of the present application. [Figure 6A-D] FIG. 1 is an exemplary schematic diagram of a positive example probability trend scatter plot shown in some examples of the present application. [Figure 7A-D] FIG. 10 is an exemplary schematic diagram of another positive example probability trend scatter plot shown in some examples of the present application. DETAILED DESCRIPTION OF THE INVENTION

[0025] In order to more clearly explain the technical solutions of the embodiments of the present application, the drawings necessary for the description of the embodiments are briefly introduced below. Obviously, the drawings in the following description are merely examples or embodiments of the present application, and those skilled in the art can also apply the present application to other similar scenarios based on these drawings without any creative efforts. Unless otherwise clear from the language environment or unless otherwise stated, the same reference numerals in the drawings represent the same structures or operations.

[0026] It will be understood that the terms "system," "device," "unit," and / or "module" as used herein are a means of distinguishing between different components, elements, parts, portions, or assemblies at different levels, although other terms may be substituted if they achieve the same purpose.

[0027] As set forth in this application and the claims, unless the context clearly indicates otherwise, words such as "a," "one," "an," "one," and / or "the" do not specifically refer to the singular but may include the plural. Generally speaking, the terms "comprise" and "containing" only include the specifically identified steps and elements, and do not constitute an exclusive list, as a method or device may also include other steps or elements.

[0028] This application uses flowcharts to describe operations performed by systems according to embodiments of the application. It should be understood that the preceding and following operations are not necessarily performed in exact order. Instead, these steps can be processed in reverse order or simultaneously. At the same time, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0029] FIG. 1 is a schematic diagram of an application scenario of a system for guiding an operator to ask questions to a rescue caller and a system for guiding the operator to determine whether a rescuee is a positive case, according to some embodiments of the present application.

[0030] The system for guiding an operator to ask questions to a help caller and the system for guiding the operator to determine whether a help recipient is a positive example provided by the embodiments of the present application can be applied to various task scenarios, such as a telephone help scenario in an emergency medical dispatch center, a telephone recommendation scenario in an online sales platform, etc. For example, an online platform can use the system provided by the present application to obtain target question semantics configured to guide an operator to ask the next question, and the operator can determine the true purchase intention of the call help user through the question by referring to the target question semantics.

[0031] As an example, the application scenario of the system in this application for guiding operators to ask questions to emergency callers will be described using the telephone emergency services of an emergency medical dispatch center as an example.

[0032] As shown in FIG. 1, the application scenario 100 may include a server 110, a terminal 120, and a network .

[0033] In some embodiments, server 110 and terminal 120 may exchange data or information over network 130. For example, server 110 may obtain information and / or data in terminal 120 over network 130 or transmit information and / or data to terminal 120 over network 130.

[0034] The terminal 120 is an electronic device configured to allow an operator (e.g., a dispatcher at an emergency medical dispatch center) to respond to a rescue call and provide telephone emergency assistance guidance to the operator via the terminal 120. In various embodiments, the terminal 120 can acquire the operator's question and the rescue caller's answer and send the question and answer to the server 110 for processing. In various embodiments, the terminal 120 can present the target question semantics, the rescue recipient's positive case probability, rate ratio, rate difference, and positive case probability trend scatter plot received from the server 110 to the operator in various ways (e.g., voice prompts, text prompts, etc.). In various embodiments, if the terminal 120 has high processing power, the terminal 120 can process the operator's question and the rescue caller's answer to acquire the target question semantics, the rescue recipient's positive case probability, rate ratio, rate difference, and positive case probability trend scatter plot, etc., without being limited to the representations herein. The terminal 120 can be one or any combination of devices with input and / or output capabilities, such as a mobile device, a tablet computer, etc.

[0035] Server 110 may be a single server or a collection of servers, which may be centralized or distributed (e.g., server 110 is a distributed system), dedicated, or concurrently served by other devices or systems. In various embodiments, server 110 may be local or remote. In various embodiments, server 110 may be implemented on a cloud platform or may be virtually hosted. By way of example only, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-tier cloud, etc., or any combination thereof.

[0036] In some embodiments, the server 110 can maintain a medical dispatch database of the emergency center and classify patients in the dispatch database based on entity-attribute clusters and entity-attribute pairs. In some embodiments, the server 110 can obtain a first reference group and a second reference group that need to participate in the calculation from the medical dispatch database based on questions between the operator and the medical dispatcher, and transmit data information of the first reference group or the second reference group to the terminal 120.

[0037] In various embodiments, network 130 may be any one or more of a wired network or a wireless network. For example, network 130 may include a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), etc., or any combination thereof.

[0038] For ease of understanding, the technical solution of the present application is described below in conjunction with drawings and examples.

[0039] 2 is an exemplary schematic diagram of a system for guiding an operator to ask questions to a rescue caller, according to some embodiments of the present application. As shown in FIG. 2, the system for guiding an operator to ask questions to a rescue caller includes a first reference group acquisition module 210, a classification test module 220, a receiver operating characteristic curve acquisition module 230, a candidate question semantics acquisition module 240, and a target question semantics acquisition module 250.

[0040] The first reference group acquisition module 210 is configured to select a plurality of unprocessed first reference groups from the initial reference group. The first reference group is acquired by classifying a plurality of historical rescue recipients (e.g., patients suspected of OHCA) from the initial reference group based on an entity-attribute cluster. The entity-attribute cluster is composed of a plurality of entity-attribute pairs corresponding to the same core question. The core question is a subject that the operator is trying to understand through the question. For example, the core question may be a subject such as help, consciousness, breathing, heart, or talking. The entity-attribute pair includes an entity and an attribute, where the entity represents the semantics of the operator's question in the historical rescue call, and the attribute represents the semantics of the rescue caller's answer to the question. For example, in the case where the operator asks "Hello, how can I help you?" and the rescue caller replies "It's not a big deal, but my father is not feeling well," the semantics of "hello" are extracted from the question and the semantics of "not feeling well" are extracted from the answer. Therefore, the entity-attribute pair corresponding to the question and answer is "hello-not feeling well." Since the core question of the question is "help," the entity-attribute pair corresponding to the question and answer belongs to the "help" entity-attribute cluster. As shown in Figure 4, answers based on the same core question constitute an entity-attribute cluster, and different entity-attribute clusters are independent of each other. The coordinates (black dots) on the entity-attribute cluster represent entities extracted from the core question, each of which corresponds to at least one different attribute. The entity and attribute together constitute an entity-attribute pair.

[0041] In specific implementation, every conversation between the operator and the emergency caller is converted into text using a speech-to-text algorithm, and then text recognition is performed based on the converted text to obtain questions and answers represented by text. In some embodiments, the questions and answers in text format can be stored in different databases, and when the emergency caller calls for help, the questions and answers are retrieved from the question database and the answer database, respectively, and the questions and answers are matched to obtain entity-attribute clusters and entity-attribute pairs corresponding to the conversation.

[0042] The first aid dialogues stored in the rescue dispatch database are large and messy, and are difficult to standardize and structure, making them difficult to use. In the embodiment provided by the present application, the first aid dialogues in the rescue dispatch database are structured through the concepts of entity-attribute clusters and entity-attribute pairs, so that the existing dialogue data resources can be used to guide the dispatcher to ask questions to the rescue caller, determine the situation of the rescued person at the scene, and the rescue progress of the rescue caller to the rescued person, thereby reducing the expertise requirements of the emergency dispatcher and allowing ordinary dispatchers to provide effective first aid services to patients.

[0043] The initial reference group is a group consisting of historical rescue recipients selected from a rescue dispatch database. In specific implementation, personal characteristic information of the rescue recipient is extracted from the information provided by the rescue caller, and an initial reference group can be determined from a database (e.g., a rescue dispatch database) based on the personal characteristic information. The personal characteristic information includes, but is not limited to, information indicating the rescue recipient's personal characteristics, such as gender, age, body type, time of rescue, and geographic location. For example, in a dialogue scenario in which an operator asks, "Hello, xxx Command Center, how can I help you?" and the rescue caller replies, "It's no big deal, but my father is not feeling well," two characteristic information items, "adult" and "male," can be extracted from the semantics of the rescue caller's response, and 500,000 historical rescue recipient cases that meet these characteristic information can be selected from the database as the initial reference group. As the dialogue progresses, a reference group more similar to the rescue recipient's situation can be obtained from the database based on the entity-attribute clusters and entity-attribute pairs corresponding to each dialogue.

[0044] The initial reference group can be further divided into multiple first reference groups based on the entity attribute cluster, for example, the initial reference group can include a first reference group corresponding to a "help" entity attribute cluster, a first reference group corresponding to an "awareness" entity attribute cluster, a first reference group corresponding to a "breathing" entity attribute cluster, etc.

[0045] In specific implementation, a plurality of unprocessed first reference groups can be selected from the initial reference group, and then a target question semantic can be obtained based on the plurality of first reference groups to guide the operator to ask the next question. As just one example, the entity attribute cluster corresponding to the current dialogue is the "help" entity attribute cluster, and the first reference groups corresponding to the other four entity attribute clusters included in the initial reference group have not yet been processed, which for ease of explanation are respectively referred to as the consciousness cluster, the breathing cluster, the heart cluster, and the speaking cluster. Then, the first reference group corresponding to the consciousness cluster, the first reference group corresponding to the breathing cluster, the first reference group corresponding to the heart cluster, and the first reference group corresponding to the speaking cluster can be selected as a plurality of first reference groups for subsequent processing.

[0046] The classification test module 220 is configured to perform a classification test on any of the first reference groups using a cutoff point that is a positive example probability of any of the second reference groups included in any of the first reference groups among the plurality of first reference groups, and to obtain classification sensitivity and classification specificity corresponding to any of the second reference groups.

[0047] As shown in FIG. 4 , the second reference group is obtained by classifying the first reference group based on entity attributes. For example, the consciousness cluster in the above example includes four entity-attribute pairs: “coma-yes,” “responsive-shock,” “awake-unresponsive,” and “responsive-still.” (The actual number of entity-attribute pairs included may be much more than four, but this is simplified for ease of explanation.) Based on the four entity-attribute pairs, the first reference group corresponding to the consciousness cluster can be divided into four second reference groups. For example, the posture guidance cluster (entity-attribute cluster) includes four entity-attribute pairs: “lying-flat,” “lying-on-back-okay,” “lying-yes,” and “lying-on-back-okay.” (The actual number of entity-attribute pairs included may be much more than four, but this is simplified for ease of explanation.) Based on the four entity-attribute pairs, the first reference group corresponding to the posture guidance cluster can be divided into four second reference groups.

[0048] The probability of a positive case in the second reference group is the ratio of members who were actually classified as positive cases to the total number of people in the second reference group, The calculation formula is expressed as follows: P c = Number of patients actually classified as positive / Total number of patients Equation (1)

[0049] By way of example only, if a second reference group has a total of 100,000 people, of which 5,000 members are identified as positive cases, then the probability of a positive case in the second reference group is 0.005.

[0050] Positive examples, also called "positive examples," are proper examples or instances of a concept. All positive examples of each concept contain common essential features, and each concept has positive examples and counterexamples (negative examples). For example, elephants, lions, tigers, cats, dogs, and whales are positive examples of the concept of mammals, while fish and turtles are counterexamples of the concept of mammals.

[0051] In a scenario involving a telephone rescue task at an emergency medical dispatch center, when a dispatcher tries to determine whether a patient is in a certain state, a patient in that state is a positive example, and a patient not in that state is a counterexample. For example, when a dispatcher determines through dialogue whether a rescuee is a patient suspected of OHCA (since only an emergency physician can confirm whether a rescuee is an OHCA patient, the dispatcher can only determine that the rescuee is a patient suspected of OHCA), a positive example would be a patient diagnosed by the emergency physician as an OHCA patient, and a counterexample would be a patient diagnosed by the emergency physician as a non-OHCA patient. For example, when a dispatcher determines through dialogue the progress of a rescuer's rescue of a rescuee, a positive example would be a patient for whom the rescuer has initiated CPR, and a counterexample would be a patient for whom the rescuer has not yet initiated CPR. For example, when a dispatcher determines through dialogue whether a rescuee has achieved ROSC, a positive example would be a patient who has achieved ROSC, and a counterexample would be a patient who has not achieved ROSC.

[0052] In other application scenarios, the positive examples and counterexamples may refer to target groups different from the positive examples and counterexamples in the emergency medical dispatch center telephone rescue task scenario, and are not limited by the expressions in this specification. For example, in an online sales platform telephone recommendation scenario, a user who is willing to purchase a specific product would be a positive example, and a user who is not willing to purchase a specific product would be a counterexample.

[0053] A preset positive example is an object assumed to be a positive example; that is, the true classification of a preset positive example may be either a positive example or a counterexample. A preset counterexample is an object assumed to be a counterexample; that is, the true classification of a preset counterexample may be either a counterexample or a positive example. In a scenario of a telephone rescue task at an emergency medical dispatch center, a preset positive example is a patient that the system determines and assumes to be in a specific state by performing a classification test or other calculation, and a preset counterexample is a patient that the system determines and assumes not to be in a specific state by performing a classification test or other calculation. For example, if a dispatcher determines through dialogue whether a rescuee is a patient suspected of OHCA, a preset positive example is a patient whose doctor has diagnosed OHCA, as determined and assumed by the system, and a preset counterexample is a patient whose doctor has diagnosed non-OHCA, as determined and assumed by the system. For example, if a dispatcher determines through dialogue the progress of a rescuer's rescue of a rescuee by a rescuer, a preset positive example is a patient whose doctor has assumed that the rescuer has initiated CPR, and a preset counterexample is a patient whose doctor has assumed that the rescuer has not initiated CPR. For example, when a dispatcher determines through dialogue whether a rescuee has achieved ROSC, the preset positive example is a patient who has achieved ROSC as assumed by the system, and the preset counterexample is a patient who has not achieved ROSC as assumed by the system.

[0054] The cutoff point is for determining a positive example label of the first reference group. In some embodiments, the positive example probability of the second reference group can be set as the cutoff point, and members of the second reference group whose positive example probability in the first reference group is equal to or greater than the cutoff point can be set as preset positive examples, and members of the second reference group whose positive example probability in the first reference group is less than the cutoff point can be set as preset counterexamples.

[0055] For example, the "Help" entity-attribute cluster contains four entity-attribute pairs: "hello-passed out", "hello-dying", "hello-please come now", and "hello-ambulance". (The actual number of entity-attribute pairs may be much more than four, but this is simplified for ease of explanation.) After sorting, the positive probability of the second reference group corresponding to each entity-attribute pair is P=0.165 for "hello-passed out", P=0.035 for "hello-dying", P=0.019 for "hello-please come now", and P=0.011 for "hello-ambulance". If P=0.035 (hello-dying) is used as the cutoff point, the member of the second reference group corresponding to the "hello-passed out" entity-attribute pair (whose positive probability of 0.165 is greater than the cutoff point of 0.035) is "hello". The members of the second reference group corresponding to the entity-attribute pair "Chiha - I'm about to die" (whose positive probability of 0.035 is equal to the cutoff point of 0.035) can be used as preset positive examples, and the members of the second reference group corresponding to the entity-attribute pair "Hello - Please come soon" (whose positive probability of 0.019 is less than the cutoff point of 0.035) and the members of the second reference group corresponding to the entity-attribute pair "Hello - Ambulance" (whose positive probability of 0.011 is less than the cutoff point of 0.035) can be used as preset counterexamples.

[0056] For example, the palm positioning cluster (entity attribute cluster) contains four entity attribute pairs: "Put your hands together with your carpals centered between the patient's nipples - OK," "Put your carpals centered between the patient's nipples - OK," "Put your hands together in front of the patient's chest - OK," and "Put your hands in front of the patient's chest - OK" (the actual number of entity attribute pairs may be much more than four, but this has been simplified for ease of explanation). After sorting, the positive probability of the second reference group corresponding to each entity-attribute pair was P=0.857 for "Put your hands together and place the carpus in the center between the patient's nipples - OK", P=0.811 for "Put your hands together and place the carpus in the center between the patient's nipples - OK", P=0.789 for "Put your hands together and place them in front of the patient's chest - OK", and P=0.667 for "Put your hands together and place them in front of the patient's chest - OK". When P=0.811 (put your hands together and place the carpus in the center between the patient's nipples - OK) is used as the cutoff point, the positive probability of the entity-attribute pair "put your hands together and place the carpus in the center between the patient's nipples - OK" is greater than the cutoff point. The members of the second reference group corresponding to the entity-attribute pair "Place your wrist in the center between the patient's nipples - OK" (whose positive example probability is equal to the cutoff point) can be preset positive examples, and the members of the second reference group corresponding to the entity-attribute pair "Place your palm in the center between the patient's nipples - OK" (whose positive example probability is equal to the cutoff point) can be preset counterexamples. The members of the second reference group corresponding to the entity-attribute pair "Place your palm in the center between the patient's nipples - OK" (whose positive example probability is less than the cutoff point) can be preset counterexamples.

[0057] Classification tests may include screening tests and diagnostic tests, which use quick and easy test examinations or other methods to detect unidentified suspected cases from apparently healthy individuals. In a telephone rescue scenario at an emergency medical dispatch center, dispatchers determine the true medical condition of the rescued person through dialogue with the rescue caller. Typically, to evaluate the reliability of a classification test, it is necessary to calculate classification sensitivity and classification specificity. The classification test table is as follows:

[0058] [Table 1]

[0059] In specific implementation, the classification sensitivity can be obtained according to the number of members of the preset positive cases (true positives (a) in Table 1) that are actually classified as positive cases and the number of members of the first reference group that are actually classified as positive cases (the sum of true positives (a) and false negatives (c) in Table 1). The calculation formula is as follows: TIFF0007749870000002.tif14140

[0060] In actual implementation, the classification sensitivity can be obtained according to the number of members of the preset counterexamples (true negatives (d) in Table 1) that are actually classified as counterexamples and the number of members of the first reference group (the sum of false positives (b) and true negatives (d) in Table 1) that are actually classified as counterexamples. The calculation formula is as follows: TIFF0007749870000003.tif14140

[0061] As just one example, the breathing cluster contains four entity-attribute pairs (the actual number is much greater than four, but is simplified for ease of explanation). After sorting based on the positive example probability of the second reference group, the positive example probabilities of each entity-attribute pair and the corresponding second reference group are "breathing - none" (P = 0.400), "breathing - discomfort" (P = 0.210), "gasping - choking" (P = 0.145), and "gasping - very weak" (P = 0.074), respectively. Based on the above assumptions, a classification test was performed on the first reference group corresponding to the "breathing cluster" using the positive example probability of the second reference group corresponding to "breathing - none" as the cutoff point, resulting in Table 2.

[0062] [Table 2]

[0063] In Table 2, X represents the cutoff point, and X=1 means that members that meet the cutoff point classification condition are used as preset positive examples. In this example, if the positive example probability (0.400) of the second reference group corresponding to "breathing-none" is used as the cutoff point, members of the second reference group corresponding to "breathing-none" are used as preset positive examples, and members of the second reference group (whose positive example probability is less than 0.400) whose remaining entities correspond to the attribute pair are used as preset counterexamples.

[0064] In Table 2, the number of members of the preset positive examples that are actually classified as positive examples (a) is 480, and the number of members of the first reference group that are actually classified as positive examples (a+c) is 480+1120=1600. In this case, based on Equation (2), TIFF0007749870000005.tif20168 is calculated and the number of preset counterexample members (d) whose actual classification is a counterexample is 12680. The number of members of the first reference group that are actually classified as counterexamples (b+d) is 720+12680=13400. Based on Equation (3), TIFF0007749870000006.tif18168 can be calculated.

[0065] According to the above method, when the probability of positive cases of the second reference groups corresponding to "breathing - discomfort," "gasping - choking," and "gasping - somewhat weak" are used as cutoff points, the classification sensitivity and classification specificity corresponding to each second reference group can be obtained.

[0066] The receiver operating characteristic curve acquisition module 230 is configured to acquire a receiver operating characteristic curve corresponding to any first reference group based on the classification sensitivity and classification specificity corresponding to a plurality of second reference groups within the any first reference group.

[0067] A receiver operating characteristic curve (ROC) is a line connecting points where the difference between the classification specificity (Sp) and 1 obtained by subjects using different criteria under specific stimulus conditions is the abscissa and the classification sensitivity (Se) is the ordinate. A ROC curve is characterized by the fact that it must pass through (0,0) and (1,1), and points on the curve appear only on the line connecting these two points, and it increases monotonically. Therefore, an ROC curve is meaningful only when the AUC is greater than 0.5.

[0068] In a specific implementation, the classification sensitivity of any of the multiple second reference groups is taken as the ordinate, and the difference between 1 and the classification specificity of any of the multiple second reference groups is taken as the abscissa, to obtain a coordinate point corresponding to any of the multiple second reference groups. Then, the coordinate points corresponding to each of the multiple second reference groups in the first reference group are connected to obtain a receiver operating characteristic curve corresponding to any of the multiple first reference groups. As just one example, the classification sensitivity and classification specificity corresponding to the multiple second reference groups in the first reference group corresponding to the "breathing cluster" obtained in the above example can be used to obtain a receiver operating characteristic curve such as that shown in FIG. 5.

[0069] The candidate question semantics acquisition module 240 is configured to determine a target first reference group from the plurality of first reference groups based on a receiver operating characteristic curve corresponding to each of the first reference groups in the plurality of first reference groups, and to determine an entity of each entity-attribute pair in the entity-attribute cluster corresponding to the target first reference group as a candidate question semantic.

[0070] The larger the area under the receiver operating characteristic curve, the better the classification effect of the classification test corresponding to that curve. Therefore, in some embodiments, the area under the curve (AUC) of the receiver operating characteristic curve corresponding to each first reference group within a plurality of first reference groups can be calculated (e.g., calculated using the trapezoidal rule), and then the first reference group corresponding to the largest area under the curve can be used as the target first reference group. As just one example, for three first reference groups corresponding to entity attribute clusters named "Respiration Cluster," "Heart Cluster," and "Speech Cluster," respectively, the above method can be used to obtain the corresponding receiver operating characteristic curves. The calculated AUCs of the receiver operating characteristic curves are 0.761, 0.733, and 0.662, respectively. The largest AUC corresponds to the first reference group corresponding to the "Respiration Cluster." Therefore, the first reference group corresponding to the "Respiration Cluster" is used as the target first reference group.

[0071] After the target first reference group is obtained, the entities of each entity-attribute pair in the entity-attribute cluster corresponding to the target first reference group may be used as candidate question semantics. By way of example only, the "breathing cluster" corresponding to the target first reference group obtained in the above example includes the entity-attribute pairs "breathing-none," "breathing-uncomfortable," "gasping-choking," and "gasping-slightly weak," so "breathing" and "gasping" can be used as candidate question semantics. A target question semantic can then be selected from the candidate question semantics.

[0072] The target question semantics acquisition module is configured to determine the target question semantics from the candidate question semantics based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups to a perfect classification coordinate point on a receiver operating characteristic curve corresponding to the target first reference group.

[0073] The perfect classification coordinate point is the coordinate point on the receiver operating characteristic curve where the abscissa is 0 and the ordinate is 1. The coordinate point (0,1) on the receiver operating characteristic curve represents a classification test result with a classification specificity (Sp) of 1 and a classification sensitivity of 1, which is a perfect result with the highest classification accuracy. On the receiver operating characteristic curve, the distance from the coordinate point corresponding to an arbitrary second reference group to the perfect classification coordinate point can reflect the accuracy of the classification result obtained by performing a classification test on the first reference group using the positive example probability of the arbitrary second reference group as the cutoff point. The distance from the coordinate point corresponding to an arbitrary second reference group to the perfect classification coordinate point on the receiver operating characteristic curve can be obtained using the following formula: TIFF0007749870000007.tif14138

[0074] In specific implementation, a plurality of distances from the coordinate points corresponding to each of the plurality of second reference groups to the perfect classification coordinate points can be calculated, and the second reference group corresponding to the shortest distance among the plurality of distances can be set as the target second reference group, and the entity of the entity-attribute pair corresponding to the target second reference group can be set as the target question semantics. As just one example, the "breathing cluster" corresponding to the target first reference group in the above example includes the entity-attribute pairs "breathing - none," "breathing - discomfort," "gasping - choking," and "gasping - slightly weak." In the receiver operating characteristic curve, the distances from the four coordinate points corresponding to the four entity-attribute pairs to (0,1) are 0.702, 0.548, 0.419, and 0.537, respectively. Therefore, "gasping" of "gasping - choking" is used as the target question semantics.

[0075] The target question semantics are configured to guide the operator to ask the following questions: In some embodiments, the operator is a dispatcher at an emergency medical dispatch center, and the dispatcher's questions are to determine whether the victim requires on-scene cardiopulmonary resuscitation.

[0076] In some embodiments, the operator is a dispatcher at an emergency medical dispatch center, and the dispatcher's questions are intended to guide the rescue caller in performing CPR on the recipient at the scene.

[0077] 3 is an exemplary schematic diagram of a system for guiding an operator to determine whether a rescuee is a positive case according to some embodiments of the present application. As shown in FIG. 3, the system for guiding an operator to determine whether a rescuee is a positive case includes an entity-attribute cluster and entity-attribute pair determination module 310, a first reference group and a second reference group determination module 320, a preset positive case member count acquisition module 330, and a positive case probability determination module 340.

[0078] The entity attribute cluster and entity attribute pair determination module 310 is configured to determine an entity attribute cluster currently corresponding to the rescuee and an entity attribute pair currently corresponding to the rescuee based on an operator's question and the rescue caller's answer to that question in any interaction between the operator and the rescue caller.

[0079] An entity-attribute cluster consists of multiple entity-attribute pairs corresponding to the same core question, where an entity-attribute pair includes an entity and an attribute, where the entity represents the semantics of the operator's question in a historical emergency call, and the attribute represents the semantics of the emergency caller's answer to the question.

[0080] In specific implementation, the operator's question can be obtained, question semantic information can be extracted from the question, the rescue caller's answer to the question can be obtained, answer semantic information can be extracted from the answer, the core question explained in the question can be determined, an entity attribute cluster can be obtained based on the core question, and an entity attribute pair can be determined based on the question semantic information and the answer semantic information. Details of the entity attribute cluster and the entity attribute pair are not further described here, and refer to FIG. 2.

[0081] The first reference group and second reference group determination module 320 is configured to obtain a first reference group corresponding to an entity attribute cluster and a second reference group corresponding to an entity attribute pair from the initial reference group, where the first reference group is obtained by classifying a plurality of historical rescuees in the initial reference group based on the entity attribute cluster, and the second reference group is obtained by classifying the first reference group based on the entity attribute.

[0082] For a detailed description of the initial reference group, first reference group, and second reference group, please refer to FIG. 2 and will not be described again here.

[0083] The preset positive example member number obtaining module 330 is configured to obtain the number of members of the first reference group that are preset positive examples, using the positive example probability of the second reference group as a cutoff point.

[0084] The cutoff point is for determining the positive example label of the first reference group, and a detailed description of the cutoff point will not be further described here, with reference to FIG.

[0085] In some embodiments, the number of members in the second reference group whose positive example probability in the first reference group is equal to or greater than the cutoff point can be used as the preset number of positive example members. For a detailed description of this embodiment, please refer to the example of step S220, and it will not be described again here.

[0086] The positive case probability determination module 340 is configured to obtain the positive case probability of the rescuee based on the ratio of the number of members of the first reference group who are preset positive cases and are actually classified as positive cases to the number of members in the first reference group who are preset positive cases.

[0087] As just one example, the "help" entity-attribute cluster contains four entity-attribute pairs (the actual number of entity-attribute pairs may be much more than four, but is simplified for ease of explanation), and after sorting based on the positive example probability of the second reference group, these four entity-attribute pairs and the corresponding positive example probability of each entity-attribute pair in the second reference group are "hello-I fainted" (P=0.165), "hello-I'm dying" (P=0.035), "hello-please come now" (P=0.019), and "hello-ambulance" (P=0.011), respectively. According to the current rescue caller's response, "Hello, my father is dying," if we use the positive case probability (0.035) of the second reference group corresponding to the entity-attribute pair "Hello, my father is dying" as the cutoff point, we can use the members of the second reference group corresponding to the entity-attribute pairs "Hello, he fainted" (the corresponding positive case probability of the second reference group is 0.165, which is greater than the cutoff point) and "Hello, he is dying" (the corresponding positive case probability of the second reference group is 0.035, which is equal to the cutoff point) as preset positive cases. There are a total of 100,000 cases. Of these 100,000 members, 4,000 cases were confirmed by doctors as OHCA, and the positive case probability of the rescuee is 0.04.

[0088] The positive case probability of the rescuee is configured to guide the operator to determine whether the rescuee is a positive case. In some embodiments, the operator is a dispatcher of an emergency medical dispatch center, and a positive case is a rescuee who requires on-site cardiopulmonary resuscitation.

[0089] In some embodiments, the operator is a dispatcher at an emergency medical dispatch center, and the positive case is a patient who is encouraged by the dispatcher and on whom CPR is initiated by the rescue caller or other bystander.

[0090] In some embodiments, the operator is a dispatcher at an emergency medical dispatch center and the positive case is a ROSC recipient following cardiopulmonary resuscitation in the field.

[0091] In some embodiments, the system for guiding an operator to ask a rescue dispatcher also includes a positive case probability trend scatter plot obtainment module and a linear regression equation slope obtainment module.

[0092] The positive example probability trend scatter diagram acquisition module is configured to acquire a positive example probability trend scatter diagram based on a plurality of positive example probabilities of the rescuee obtained in a plurality of interactions between the operator and the rescue sender.

[0093] The linear regression equation slope obtaining module is configured to fit the positive example probability trend scatter plot using a linear regression equation and solve the linear regression equation to obtain the slope of the linear regression equation.

[0094] The positive case probability trend scatter diagram and the slope of the linear regression equation represent the trend of the probability that the rescuee is a positive case. In specific implementation, the coordinates on the positive case probability trend scatter diagram can be obtained by taking the number of interactions corresponding to any one of the multiple positive case probabilities as the abscissa and the any one of the positive case probabilities as the ordinate.

[0095] Figure 6 A to Figure 6D shows a positive probability trend scatter diagram drawn based on multiple positive probability of rescuees obtained from multiple interactions, where Figures 6A to 6D Figure 6 shows the positive probability trend scatter plot obtained from multiple interactions between different rescue callers and operators. A to Figure 6D The four slopes of are 0.045, 0.0351, 0.1047, and 0.1291, respectively. Based on this, The patient in Figure 6D is judged to be the most likely outcome.

[0096] Figure 7 A to Figure 7D 1 shows another positive probability trend scatter plot drawn based on multiple positive probability of rescuees obtained from multiple interactions, where Figures 7A to 7D Figure 7 shows the positive probability trend scatter plot obtained from multiple interactions between different rescue callers and operators. A is,The predicted probability of the result rapidly increases to 0.9 after only 3 iterations, ,indicating that the rescued patient is likely to belong to the actual ,classification result. 7B is ,shows that the predicted probability of the outcome first increases, then decreases, and then increases again, but the overall trend still tends towards the actual classification result. 7C is ,shows that the first few iterations failed to "capture" the positive label, but the resulting,prediction probability rapidly increased in the final iterations.,Fig. 7D is , which indicates that it is unlikely to belong to the actual classification result. A to Figure 7D The four slopes of are 0.297, 0.138, 0.121, and 0.068, respectively. The patient in Figure 7A The outcome is still judged to be the most likely.

[0097] In the positive probability trend scatter plot, because each rescuee is independent and the questions and answers for each entity attribute cluster are also independent, the variance of the random variable P (positive probability) at different x (abscissa) is assumed to be equal. According to equations (2) and (3), combining expertise, as x increases, y (y = P, 0 <= P <= 1) generally increases. Therefore, the slope (b) of the linear regression equation (y = a + bx) can be used as the basis for determining the results. The larger b, the higher the probability of a future rescuee being a positive case. In specific implementation, the least squares method can be used to solve the linear regression equation to determine the slope of the equation. The calculation formula is as follows: TIFF0007749870000008.tif19139

[0098] In some embodiments, other regression models can be used to fit the scatter plot, including but not limited to the formulas herein. For example, the scatter plot can be fitted using functions such as an inverse function, a power function, a logarithmic function, a composite function, a growth function, an exponential function, etc.

[0099] In several embodiments, the positive case probability trend scatter plot and the slope of the linear regression equation can be displayed to the operator (e.g., via the screen of terminal 120) to guide the operator in determining whether the rescuee is a positive case.

[0100] In some embodiments, the system for guiding an operator to ask a rescue dispatcher further includes a rate ratio acquisition module and a rate difference acquisition module.

[0101] The rate ratio acquisition module is configured to acquire a rate ratio based on the ratio between the current positive case probability of the rescuee and the positive case probability of the rescuee acquired in the previous interaction. The calculation formula for the rate ratio is: TIFF0007749870000009.tif13139

[0102] By way of example only, if P1 is 0.04 and P2 is 0.1, then RR can be calculated to be 2.5 based on equation (6).

[0103] The rate ratio (RR) can represent the value (contribution) of the cutoff point selected in this round of interaction to predicting the actual classification; if RR>1, the cutoff point selected in this round of interaction has discriminatory value for the actual classification; if RR≦1, the cutoff point selected in this round of interaction does not have discriminatory value for the actual classification.

[0104] The rate difference acquisition module is configured to acquire a rate difference based on the difference between the positive case probability of the current rescuee and the positive case probability of the rescuee acquired in the previous interaction. The calculation formula for the rate difference is as follows: RD=P c -P c-1 Formula (7)

[0105] By way of example only, if P1 is 0.04 and P2 is 0.1, then RR can be calculated to be 0.06 based on equation (7).

[0106] The rate difference (RD) can represent the value of the cutoff point selected in this round of interaction for predicting the actual classification. If RD>0, the cutoff point selected in this round of interaction has discriminatory value for the actual classification, and if RR≦0, the cutoff point selected in this round of interaction has no discriminatory value for the actual classification.

[0107] In several embodiments, the rate difference and rate ratio can be displayed to the operator to guide the operator in determining the value of the positive case probability of the current rescuee in determining whether the current rescuee is a positive case.

[0108] If RR>1, the label has a discriminatory value for the actual classification, if RR≦1, the label has no discriminatory value for the actual classification, if RD>0, the label has a discriminatory value for the actual classification, if RD≦0, the label has no discriminatory value for the actual classification.

[0109] Another embodiment of the present application provides a system for guiding a researcher to ask questions to a research subject, the system including a first reference group acquisition module, a classification test module, a receiver operating characteristic curve acquisition module, a candidate question semantics acquisition module, and a target question semantics acquisition module.

[0110] The first reference group obtaining module is configured to select a plurality of unprocessed first reference groups from the initial reference group.

[0111] The first reference group is obtained by classifying multiple historical survey subjects in the initial reference group based on entity-attribute clusters, and the entity-attribute clusters are composed of multiple entity-attribute pairs corresponding to the same core question, and the entity-attribute pairs include an entity and an attribute, where the entity represents the semantics of the survey subject's question in the historical survey call, and the attribute represents the semantics of the survey subject's answer to the question.

[0112] For a detailed description of the initial reference group, first reference group, and second reference group, please refer to FIG. 2 and will not be described again here.

[0113] The classification test module is configured to perform a classification test on any of the first reference groups using a positive example probability of any of the second reference groups included in any of the first reference groups among the plurality of first reference groups as a cutoff point, and obtain classification sensitivity and classification specificity corresponding to the any of the second reference groups.

[0114] The cutoff point is for determining the positive example labels of the first reference group, and the second reference group is obtained by classifying the first reference group based on entity attributes. A detailed description of the second reference group and the cutoff point will be given in FIG. 2 and will not be further described here.

[0115] The receiver operating characteristic curve acquisition module is configured to acquire a receiver operating characteristic curve corresponding to the arbitrary first reference group based on classification sensitivities and classification specificities corresponding to a plurality of second reference groups within the arbitrary first reference group.

[0116] The candidate question semantics acquisition module is configured to determine a target first reference group from the plurality of first reference groups based on a receiver operating characteristic curve corresponding to each first reference group in the plurality of first reference groups, and to determine an entity of each entity-attribute pair in an entity-attribute cluster corresponding to the target first reference group as a candidate question semantic.

[0117] The target question semantics acquisition module is configured to determine a target question semantics from the candidate question semantics based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups on a receiver operating characteristic curve corresponding to the target first reference group to a perfect classification coordinate point, where the perfect classification coordinate point is a coordinate point on the receiver operating characteristic curve with an abscissa of 0 and an ordinate of 1, and the target question semantics is configured to guide the researcher to ask a next question.

[0118] In the above embodiment of the system for guiding the researcher to ask questions to the researchee, the specific processing of each module and the resulting technical effects can be referred to the relevant description in the corresponding embodiment of Figure 2, and the description will be omitted.

[0119] Another embodiment of the present application provides a system for guiding an investigator to determine whether an individual being investigated is a positive case, the system including an entity-attribute cluster and entity-attribute pair determination module, a first reference group and a second reference group determination module, a preset positive case member count acquisition module, and a positive case probability determination module.

[0120] The entity-attribute cluster and entity-attribute pair determination module is configured to determine an entity-attribute cluster currently corresponding to the surveyee and an entity-attribute pair currently corresponding to the surveyee based on the surveyee's questions and the surveyee's answers to those questions in any interaction between the surveyee and the surveyee.

[0121] The entity-attribute cluster is composed of multiple entity-attribute pairs corresponding to the same core question, and the entity-attribute pair includes an entity and an attribute, where the entity represents the semantics of the researcher's question in the historical survey call, and the attribute represents the semantics of the researchee's answer to the question.

[0122] For a detailed description of the entity-attribute clusters and entity-attribute pairs, please refer to the relevant content of FIG. 2 and will not be further described here.

[0123] The first and second reference group determination module is configured to obtain a first reference group corresponding to the entity-attribute cluster and a second reference group corresponding to the entity-attribute pair from an initial reference group.

[0124] The first reference group is obtained by classifying a plurality of historical subjects in the initial reference group based on entity attribute clusters, and the second reference group is obtained by classifying the first reference group based on entity attributes.

[0125] For a detailed description of the first and second reference groups, please refer to the relevant content of FIG. 2 and will not be further described here.

[0126] The preset positive example member number obtaining module is configured to obtain the number of members of the first reference group that are preset positive examples, using the positive example probability of the second reference group as a cutoff point.

[0127] The cutoff point is used to determine the positive example labels of the first reference group, and a detailed description of the cutoff point is omitted here, and reference is made to the related content of FIG.

[0128] The positive case probability determination module is configured to obtain the positive case probability of the subject based on the ratio of the number of members of the first reference group who are preset positive cases and are actually classified as positive cases to the number of members in the first reference group who are preset positive cases.

[0129] The positive example probability is configured to guide the researcher in determining whether the researchee is a positive example. For details about the positive example probability and positive examples, please refer to the relevant content of Figure 2 and will not be further described here.

[0130] For the specific processing of each module in the above embodiment of the system that guides researchers to determine whether the research subject is a positive case and the resulting technical effects, please refer to the relevant description in the corresponding embodiment of Figure 3, and further details will be omitted here.

[0131] The above-mentioned system that guides researchers in asking questions to interviewees and the system that guides researchers in determining whether interviewees are positive examples can be applied to a variety of application scenarios, such as disease and health status surveys (an abbreviation for epidemiological surveys) and product after-sales service surveys. For example, the following will explain the task of investigating infectious diseases as an example.

[0132] Flow surveillance is key to controlling infectious diseases, and the information collected from it can play an important role in effectively containing infectious diseases. The purpose of flow surveillance is to determine the movement trajectory of the subject over a certain period of time, the people they encountered, and what happened, thereby providing a basis for identifying the route of infection, identifying the source of infection, identifying close contacts, implementing isolation measures, and clarifying the scope of disinfection.

[0133] In some embodiments, the terminal used by the investigator can acquire the investigator's questions and the interviewee's answers, and send the questions and answers to a server for processing (or be processed by the terminal used by the investigator). In some embodiments, the terminal used by the investigator can display the target question semantics, the interviewee's positive case probability, rate ratio, rate difference, and positive case probability trend scatter plot received from the server through various methods (e.g., voice prompts, text prompts, etc.), which can help the investigator determine the interviewee's flow investigation status (e.g., whether the interviewee is a close contact, whether the interviewee needs to be quarantined, etc.).

[0134] In the scenario of an infectious disease flow investigation task, if an investigator is trying to determine whether the subjects are a certain type of group of people for whom specific measures should be taken to prevent the spread of the infectious disease, subjects who belong to this type of group are positive examples, and subjects who do not belong to this type of group are counterexamples. For example, if an investigator (flow investigator) determines through dialogue whether the subjects are close contacts, positive examples could be subjects who belong to close contacts, and counterexamples could be subjects who belong to non-close contacts.

[0135] In a scenario involving a disease flow investigation task, preset positive examples are people that the system assumes belong to a specific type of person who needs to take specific measures to prevent the spread of the disease in order to perform classification tests or other calculations, and preset counterexamples are people that the system (the terminal or server used by the investigator) assumes do not belong to a specific type who needs to take specific measures to prevent the spread of the disease in order to perform classification tests or other calculations. For example, if an investigator (flow investigator) determines through dialogue whether a person being investigated is a close contact, preset positive examples may be people who belong to the close contact category that the system assumes, and preset counterexamples may be people who belong to the non-close contact category that the system assumes.

[0136] Although the basic concepts have been described above, it will be apparent to those skilled in the art that the above detailed disclosure is merely illustrative and does not limit the present invention. Although not expressly stated herein, those skilled in the art may make various changes, improvements, and modifications to the present application. Such changes, improvements, and modifications are proposed in the present application, and such changes, improvements, and modifications will still fall within the spirit and scope of the exemplary embodiments of the present application.

[0137] At the same time, this application uses certain words to describe embodiments of the application. For example, "one embodiment," "one embodiment," and / or "multiple embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of the application. Therefore, it should be emphasized and noted that multiple references to "one embodiment," "one embodiment," or "one alternative embodiment" in different places in this application do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments of the application may be combined as appropriate.

[0138] Additionally, unless expressly recited in the claims, the order of processing elements and sequences, the use of numbers and letters, or the use of other designations are not intended to limit the order of the processes and methods of the present application. While the above disclosure discusses, by way of various examples, several embodiments of the present invention that are presently believed to be useful, it is to be understood that such details are for illustrative purposes only, and that the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations within the spirit and scope of the embodiments of the present application. For example, the system components described above can be implemented through hardware devices, but can also be implemented through software-only solutions, such as installing the described system on an existing server or mobile device.

[0139] Similarly, it should be noted that in the foregoing description of embodiments of the present application, multiple features may be combined in a single embodiment, drawing, or description to simplify the disclosure and thereby facilitate understanding of one or more embodiments of the present invention. However, this method of disclosure does not imply that the subject matter of the present application requires more features than are recited in the claims. In fact, an embodiment may have fewer features than all of the features of a single disclosed embodiment.

[0140] In some examples, numerical values ​​are used to describe the quantities of components and properties. It should be understood that such numerical values, constituting the description of the examples, are in some instances modified by the modifiers "about," "approximately," or "generally." Unless otherwise specified, "about," "approximately," or "generally" imply that the stated numerical value is permitted to vary by ±20%. Accordingly, in some examples, the numerical parameters used in the specification and claims are approximations that may vary depending on the desired characteristics of the particular example. In some examples, the numerical parameters should take into account the number of significant digits specified and employ conventional digit conservation techniques. While the numerical ranges and parameters constituting the broadness of ranges in some examples of the present application are approximations, in certain examples, such numerical values ​​are set as precisely as possible.

[0141] Each patent, patent application, patent application publication, and other materials, such as articles, books, manuals, publications, documents, etc., cited in this application are incorporated herein by reference in their entirety. Any filing history documents that are inconsistent or inconsistent with the content of this application are excluded, as are any documents (now or later attached to this application) that limit the broadest scope of the claims of this application. In the event of an inconsistency or conflict between the explanations, definitions, and / or term usage in the materials attached to this application and the content set forth in this application, the explanations, definitions, and / or term usage in this application shall govern.

[0142] Finally, it should be understood that the embodiments described herein are used only to illustrate the principles of the embodiments of the present application. Other variations are possible within the scope of the present application. Thus, by way of example, and not limitation, alternative configurations of the embodiments of the present application are considered consistent with the teachings of the present application. Thus, the embodiments of the present application are not limited to the embodiments explicitly introduced and described herein.

Claims

1. 1. A system for guiding an operator to ask a rescue dispatcher a question, the system including: a first reference group acquisition module, a classification test module, a receiver operating characteristic curve acquisition module, a candidate question semantics acquisition module, and a target question semantics acquisition module; The first reference group acquisition module is configured to select a plurality of unprocessed first reference groups from an initial reference group, and the first reference groups are acquired by classifying a plurality of historical rescue recipients in the initial reference group based on an entity-attribute cluster, and the entity-attribute cluster is composed of a plurality of entity-attribute pairs corresponding to the same core question, and the entity-attribute pair includes an entity and an attribute, the entity representing the semantics of the operator's question in the historical rescue call, and the attribute representing the semantics of the rescue caller's answer to the question; the classification test module is configured to perform a classification test on any one of the plurality of first reference groups using a positive example probability of any one of the first reference groups as a cutoff point to obtain classification sensitivity and classification specificity corresponding to the any one of the second reference groups, wherein the cutoff point is for determining a positive example label of the first reference group, and the second reference group is obtained by classifying the first reference group based on entity attributes; the receiver operating characteristic curve acquisition module is configured to acquire a receiver operating characteristic curve corresponding to the arbitrary first reference group based on classification sensitivities and classification specificities corresponding to a plurality of second reference groups within the arbitrary first reference group; the candidate question semantics acquisition module is configured to determine a target first reference group from the plurality of first reference groups based on a receiver operating characteristic curve corresponding to a first reference group in the plurality of first reference groups, and determine an entity of each entity-attribute pair in an entity-attribute cluster corresponding to the target first reference group as a candidate question semantic; The target question semantics acquisition module is configured to determine a target question semantics from the candidate question semantics based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups on a receiver operating characteristic curve corresponding to the target first reference group to a perfect classification coordinate point, wherein the perfect classification coordinate point is a coordinate point on the receiver operating characteristic curve with an abscissa of 0 and an ordinate of 1, and the target question semantics is configured to guide the operator to ask a next question.

2. performing a classification test on any of the first reference groups using a positive example probability of any of the second reference groups included in any of the first reference groups among the plurality of first reference groups as a cutoff point, and obtaining classification sensitivity and classification specificity corresponding to the any of the second reference groups; members of a second reference group whose positive example probability in the first reference group is equal to or greater than the cutoff point as preset positive examples, and members of the second reference group whose positive example probability in the first reference group is less than the cutoff point as preset counterexamples; obtaining the classification sensitivity according to the number of members of the preset positive examples that are actually classified as positive examples and the number of members of the first reference group that are actually classified as positive examples; The system of claim 1, further comprising: obtaining the classification specificity according to the number of members of the preset counterexamples that are actually classified as counterexamples and the number of members of the first reference group that are actually classified as counterexamples.

3. Obtaining a receiver operating characteristic curve corresponding to the arbitrary first reference group based on classification sensitivities and classification specificities corresponding to a plurality of second reference groups within the arbitrary first reference group, Obtaining a coordinate point corresponding to an arbitrary second reference group among the plurality of second reference groups, using the classification sensitivity of the arbitrary second reference group as the ordinate and the absolute value of the difference between the classification specificity of the arbitrary second reference group and 1 as the abscissa; and connecting coordinate points corresponding to each second reference group within the first reference group to obtain a receiver operating characteristic curve corresponding to any of the first reference groups.

4. determining a target first reference group from the plurality of first reference groups based on a receiver operating characteristic curve corresponding to each first reference group in the plurality of first reference groups; calculating an area under a receiver operating characteristic curve corresponding to each first reference group in the plurality of first reference groups; and using the first reference group corresponding to the largest area under the curve among the areas under the curve as a target first reference group.

5. determining a target question semantics from the candidate question semantics based on a distance from a coordinate point corresponding to any second reference group among the plurality of second reference groups to a perfect classification coordinate point on a receiver operating characteristic curve corresponding to the target first reference group, calculating a plurality of distances from a coordinate point corresponding to each second reference group among the plurality of second reference groups to a perfect classification coordinate point; The system of claim 1, further comprising: determining a second reference group corresponding to the shortest distance among the plurality of distances as a target second reference group, and determining an entity of an entity-attribute pair corresponding to the target second reference group as the target question semantics.

6. The system described in claim 1, characterized in that the initial reference group is obtained by extracting personal characteristic information of the rescuee from the information provided by the rescue caller and determining the initial reference group based on the personal characteristic information.

7. A system for guiding an operator to determine whether a rescuee is a positive example, the system including: an entity-attribute cluster and entity-attribute pair determination module; a first reference group and a second reference group determination module; a preset positive example member number acquisition module; and a positive example probability determination module; The entity-attribute cluster and entity-attribute pair determination module is configured to determine an entity-attribute cluster currently corresponding to the rescuee and an entity-attribute pair currently corresponding to the rescuee based on a question of the operator in any dialogue between the operator and the rescue caller and an answer of the rescue caller to the question, wherein the entity-attribute cluster is composed of multiple entity-attribute pairs corresponding to the same core question, and the entity-attribute pair includes an entity and an attribute, the entity representing the semantics of the operator's question in a historical rescue call, and the attribute representing the semantics of the rescue caller's answer to the question; The first reference group and second reference group determination module is configured to obtain a first reference group corresponding to the entity attribute cluster and a second reference group corresponding to the entity attribute pair from an initial reference group, where the first reference group is obtained by classifying a plurality of historical rescuees in the initial reference group based on the entity attribute cluster, and the second reference group is obtained by classifying the first reference group based on the entity attribute; the preset positive example member number obtaining module is configured to obtain a number of members of the first reference group that are preset positive examples using a positive example probability of the second reference group as a cutoff point, where the cutoff point is for determining a positive example label of the first reference group; The positive example probability determination module is configured to obtain a positive example probability of the rescuee based on a ratio between the number of members of the first reference group who are preset positive examples and who are actually classified as positive examples, and the number of members in the first reference group who are preset positive examples, and the positive example probability is configured to guide an operator to determine whether the rescuee is a positive example.

8. The system further includes a positive example probability trend scatter plot acquisition module and a linear regression equation slope acquisition module; the positive example probability trend scatter diagram acquisition module is configured to acquire a positive example probability trend scatter diagram based on a plurality of positive example probabilities of the rescuee obtained in a plurality of interactions between the operator and the rescue caller; the linear regression equation slope acquisition module is configured to fit the positive example probability trend scatter plot using a linear regression equation and solve the linear regression equation to obtain a slope of the linear regression equation; The system according to claim 7, wherein the positive case probability trend scatter diagram and the slope of the linear regression equation are intended to represent the trend in the probability that the rescuee is a positive case.

9. Obtaining a positive example probability trend scatter diagram based on a plurality of positive example probabilities of the rescuee obtained in a plurality of interactions between the operator and the rescue caller includes: The system of claim 8, further comprising acquiring coordinates on the positive example probability trend scatter plot, with the number of interactions corresponding to an arbitrary positive example probability among the plurality of positive example probabilities as the abscissa and the arbitrary positive example probability as the ordinate.

10. Determining an entity-attribute cluster currently corresponding to the rescuee and an entity-attribute pair currently corresponding to the rescuee includes: obtaining a question of the operator and extracting question semantic information from the question; obtaining the rescue caller's answer to the question and extracting answer semantic information from the answer; determining a core question described in the question, and obtaining the entity attribute cluster based on the core question; and determining the entity-attribute pairs based on the question semantics information and the answer semantics information.

11. Obtaining the number of members of the first reference group that are preset positive examples using the positive example probability of the second reference group as a cutoff point, The system of claim 7, further comprising: setting the number of members of the second reference group whose probability of positive cases in the first reference group is equal to or greater than the cutoff point as the preset number of positive case members.

12. The system further includes a rate ratio acquisition module and a rate difference acquisition module; The rate ratio acquisition module is configured to acquire a rate ratio based on a ratio between a current positive case probability of the rescuee and a positive case probability of the rescuee acquired in a previous interaction; The rate difference acquisition module is configured to acquire a rate difference according to a difference between a current positive case probability of the rescuee and a positive case probability of the rescuee acquired in a previous interaction; The system described in claim 7, characterized in that the rate difference and the rate ratio are configured to guide the operator to judge the value of determining whether the current rescuee is a positive case based on the positive case probability of the current rescuee.

Images