Information acquisition method based on reference resolution, information acquisition device, and robot

Abstract

The application provides an information acquisition method, an information acquisition device and a robot based on reference disambiguation. The method comprises the following steps: adopting a target classification model to classify a received current question to obtain a subject and a predicate corresponding to the current question; performing merging processing on the predicate corresponding to the current question and a current main classification set to obtain an updated main classification set, and performing merging processing on the subject corresponding to the current question and a current auxiliary classification set to obtain an updated auxiliary classification set, wherein the main classification set is a set composed of predicates of multiple questions, and the auxiliary classification set is a set composed of subjects of multiple questions; combining the predicates in the updated main classification set and the subjects in the updated auxiliary classification set to obtain at least one target question, and displaying an answer corresponding to the target question on a display screen, so as to solve the problem that it is difficult to hit the question of a user and give a targeted answer in the prior art.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to an information acquisition method, an information acquisition device, a computer-readable storage medium, and a robot based on referential disambiguation. Background Technology

[0002] Relation extraction involves identifying which entities in a sentence have relationships with each other. It can extract structured data that machines can understand from complex unstructured text. After relation extraction, unstructured text yields structured graph data, enabling cross-textual associations between entities. A relationship refers to the connection between entities in the text, specifically a series of subject-verb-object triples: (subject S, verb P, object O). Currently, the mainstream relation extraction method uses multi-turn question-and-answer techniques. This involves identifying multiple questions and providing answers. However, this method struggles to accurately and quickly pinpoint the questions users want answered. Summary of the Invention

[0003] The main objective of this application is to provide an information acquisition method, information acquisition device, computer-readable storage medium, and robot based on referential disambiguation, so as to at least solve the problems in the prior art that are difficult to accurately and quickly target users' questions and provide targeted answers.

[0004] To achieve the above objectives, according to one aspect of this application, an information acquisition method based on referential disambiguation is provided, comprising: employing a target classification model to classify a received current question to obtain the subject and predicate corresponding to the current question, wherein the target classification model is a classification model trained on multiple sets of sample data using machine learning, each set of sample data including a training question and the subject and predicate corresponding to the training question; merging the predicate corresponding to the current question with a current main classification set to obtain an updated main classification set; merging the subject corresponding to the current question with a current sub-classification set to obtain an updated sub-classification set, wherein the main classification set is a set composed of predicates of multiple questions, and the sub-classification set is a set composed of subjects of multiple questions; combining the predicates in the updated main classification set and the subjects in the updated sub-classification set to obtain at least one target question, and displaying the answer corresponding to the target question on a display screen.

[0005] Optionally, the priority of each predicate in the main category set is sorted according to a first target time order, which is the order in which each predicate enters the main category set from latest to earliest. The predicate corresponding to the current question is merged with the current main category set to obtain an updated main category set, including: determining the similarity between the predicate corresponding to the current question and a first predicate to obtain a target similarity, wherein the first predicate is the predicate with the highest priority in the current main category set; if the target similarity is lower than a preset threshold, the predicate corresponding to the current question is merged into the current main category set to obtain an updated main category set.

[0006] Optionally, after obtaining the updated main category set, the information acquisition method further includes: updating the priority of each predicate in the updated main category set to obtain a further updated main category set; if the number of predicates in the further updated main category set exceeds a preset predicate threshold, deleting the predicate with the lowest priority in the further updated main category set.

[0007] Optionally, the priority of each subject in the sub-category set is sorted according to the second target time order, where the second target time is the time order in which each subject enters the sub-category set from latest to earliest. Merging the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set includes: incorporating the subject corresponding to the current question into the current sub-category set to obtain the updated sub-category set.

[0008] Optionally, after incorporating the subject corresponding to the current question into the current sub-category set to obtain an updated sub-category set, the information acquisition method further includes: updating the priority of each subject in the updated sub-category set to obtain a further updated sub-category set; if the number of subjects in the further updated sub-category set exceeds a preset subject threshold, deleting the subject with the lowest priority in the further updated sub-category set.

[0009] Optionally, the priorities of the predicates in the main category set are sorted according to a first target time order, which is the order in which the predicates entered the main category set from latest to earliest. The priorities of the subjects in the sub-category set are sorted according to a second target time order, which is the order in which the subjects entered the sub-category set from latest to earliest. Based on the predicates in the updated main category set and the subjects in the updated sub-category set, at least one target problem is obtained by combining the subject with the highest priority in the updated sub-category set with the predicate with the highest priority in the updated main category set to obtain the target problem.

[0010] Optionally, the priority of each predicate in the main category set is sorted according to a first target time order, which is the order in which each predicate enters the main category set from late to early. The priority of each subject in the sub-category set is sorted according to a second target time order, which is the order in which each subject enters the sub-category set from late to early. When multiple answers to multiple target questions can be displayed, at least one target question is obtained by combining the predicates in the updated main category set and the subjects in the updated sub-category set. This includes: determining target subjects from the updated sub-category set, in descending order of priority, based on the number of target questions for which answers can be displayed; and combining each target subject with the predicates in the updated main category set to obtain multiple target questions.

[0011] According to another aspect of this application, an information acquisition method based on referential disambiguation is provided, comprising: a classification unit, configured to classify a received current question using a target classification model to obtain the subject and predicate corresponding to the current question, wherein the target classification model is a classification model trained on multiple sets of sample data using machine learning, and each set of sample data includes a training question and the subject and predicate corresponding to the training question; a merging unit, configured to merge the predicate corresponding to the current question with a current main classification set to obtain an updated main classification set, and merge the subject corresponding to the current question with a current sub-classification set to obtain an updated sub-classification set, wherein the main classification set is a set composed of predicates of multiple questions, and the sub-classification set is a set composed of subjects of multiple questions; and a combination unit, configured to combine the predicates in the updated main classification set and the subjects in the updated sub-classification set to obtain at least one target question, and display the answer corresponding to the target question on a display screen.

[0012] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to perform any of the aforementioned information acquisition methods based on referential disambiguation.

[0013] According to another aspect of this application, a robot is provided, comprising: an information acquisition device based on referential disambiguation, the information acquisition device being used to perform any of the aforementioned information acquisition methods based on referential disambiguation.

[0014] Applying the technical solution of this application, firstly, a pre-trained target classification model is used to classify the received current question, obtaining the subject and predicate of the current question. Then, the predicate of the current question is merged with existing predicates in the current main classification set, and the subject of the current question is merged with existing subjects in the current sub-classification set, resulting in updated main classification sets and updated sub-classification sets, respectively. Finally, based on the predicates in the updated main classification set and the subjects in the updated sub-classification set, at least one target question is obtained, and the corresponding answer to the target question is displayed on the screen, thus recommending the answer to the target question to the user. Compared to existing multi-turn question-and-answer methods that recommend answers to user-defined target questions, this solution categorizes the current question to obtain its subject and predicate. It then merges the predicate of the current question with existing predicates in the current main category set, and merges the subject of the current question with existing subjects in the current sub-category set. This ensures that the resulting target question is the one the user most wants an answer for, and that the target question is more accurate. This allows for more targeted targeting of the user's most desired answer and faster recommendation of the corresponding answer, thus solving the problem of existing technologies struggling to accurately and quickly identify and address user-defined questions. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0016] Figure 1 A hardware structure block diagram of a mobile terminal performing an information acquisition method based on referential disambiguation according to an embodiment of this application is shown.

[0017] Figure 2 A flowchart illustrating an information acquisition method based on referential disambiguation according to an embodiment of this application is shown.

[0018] Figure 3 A schematic diagram of an information acquisition device based on referential disambiguation provided according to an embodiment of this application is shown.

[0019] The above figures include the following reference numerals:

[0020] 102. Processor; 104. Memory; 106. Transmission device; 108. Input / output device. Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] As described in the background section, existing technologies struggle to accurately and quickly pinpoint user questions and provide targeted answers. To address these issues, embodiments of this application provide an information acquisition method, an information acquisition device, a computer-readable storage medium, and a robot based on referential disambiguation.

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0026] The methods and embodiments provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal based on a referential disambiguation information acquisition method according to an embodiment of the present invention. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal. For example, the mobile terminal may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0027] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the information acquisition method based on referential disambiguation in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the method described. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of such networks may include wireless networks provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0028] This embodiment provides an information acquisition method based on referential disambiguation that runs on a mobile terminal, computer terminal, or similar computing device. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0029] Figure 2 This is a flowchart of an information acquisition method based on referential disambiguation according to an embodiment of this application. For example... Figure 2 As shown, the information acquisition method includes the following steps:

[0030] Step S201: Use a target classification model to classify the received current question and obtain the subject and predicate corresponding to the current question. The target classification model is a classification model obtained by training multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question.

[0031] In one specific embodiment, the target classification model can be obtained by training on classification models such as support vector machines, decision trees, and Naive Bayes.

[0032] Specifically, if the current question received is "What is the weather like in Beijing?", then by classifying the current question using the target classification model, we can obtain the subject as "Beijing" and the predicate as "weather".

[0033] In practical robot applications, the current question received by the robot can be issued by the user via voice, or by the user clicking on a display screen. This application does not limit the form in which the current question is issued.

[0034] Step S202: Merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set; merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set; the main category set is a set composed of the predicates of multiple questions; the sub-category set is a set composed of the subjects of multiple questions.

[0035] In step S202, the predicate corresponding to the current problem is merged with the current main classification set, that is, the predicate obtained by classifying the current problem is merged with the predicates already in the current main classification set.

[0036] In step S202, the subject corresponding to the current problem is merged with the current sub-category set, that is, the subject obtained by classifying the current problem is merged with the subject already existing in the current sub-category set.

[0037] Step S203: Based on the predicate in the updated main category set and the subject in the updated sub-category set, combine them to obtain at least one target question, and display the answer corresponding to the target question on the display screen.

[0038] Specifically, after obtaining the target question, the corresponding answer can be retrieved and displayed on the robot's screen or on a screen communicating with the robot. Secondly, after obtaining the answer, it's not limited to displaying it as text on the robot's screen or the screen communicating with the robot; it can also be played back as voice. Of course, it's also possible to simultaneously provide the user with the answer by playing back the answer as voice and displaying the target question on at least one of the robot's screen or the screen communicating with the robot. This not only makes the robot more intelligent but also more user-friendly.

[0039] In the aforementioned information acquisition method based on referential disambiguation, firstly, a pre-trained target classification model is used to classify the received current question, obtaining the subject and predicate of the current question. Then, the predicate of the current question is merged with existing predicates in the current main category set, and the subject of the current question is merged with existing subjects in the current sub-category set, resulting in updated main category sets and updated sub-category sets, respectively. Finally, the predicates in the updated main category set and the subjects in the updated sub-category set are combined to obtain at least one target question, and the corresponding answer is displayed on the screen, thus recommending the answer to the target question to the user. Compared to existing multi-turn question-and-answer methods that recommend answers to user-defined target questions, this solution categorizes the current question to obtain its subject and predicate. It then merges the predicate of the current question with existing predicates in the current main category set, and merges the subject of the current question with existing subjects in the current sub-category set. This ensures that the resulting target question is the one the user most wants an answer for, and that the target question is more accurate. This allows for more targeted targeting of the user's most desired answer and faster recommendation of the corresponding answer, thus solving the problem of existing technologies struggling to accurately and quickly identify and address user-defined questions.

[0040] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0041] In practical applications, there are situations where a user asks about the weather in a certain location one question and then immediately asks about the temperature. Therefore, to obtain the target question—the question the user most wants answered—more accurately and reasonably, in the specific implementation, the priority of each predicate in the main category set is sorted according to a first target time order. This first target time order is the order in which each predicate entered the main category set from latest to earliest. Step S202 can be implemented through steps S2021 and S2022. Wherein,

[0042] Step S2021: Determine the similarity between the predicate corresponding to the current problem and the first predicate to obtain the target similarity. The first predicate is the predicate with the highest priority in the current main classification set.

[0043] Specifically, this application does not limit the specific method for determining the similarity between the predicate corresponding to the current question and the first predicate. Any feasible method in the prior art for determining the similarity between the predicate corresponding to the current question and the first predicate can be used. In one specific embodiment, a method based on semantic distance or a method based on a statistical model can be used to determine the similarity between the predicate corresponding to the current question and the first predicate.

[0044] Step S2022: If the target similarity is lower than a preset threshold, merge the predicate corresponding to the current question into the current main classification set to obtain the updated main classification set.

[0045] Specifically, the size of the preset threshold can be flexibly adjusted according to the actual application scenario, and this application does not limit the size of the preset threshold.

[0046] In one specific embodiment, if the previous question is "What is the weather like in Beijing?" and the current question is "What is the population of Beijing?", then the predicate of the previous question is "weather" and the predicate of the current question is "population". That is, "weather" is the predicate with the highest priority in the main category set. When the similarity between "weather" and "population" is lower than a preset threshold, the main category set is updated to (population, weather). If the current question is "What is the temperature in Beijing?", then the predicate of the previous question is "weather" and the predicate of the current question is "temperature". When the similarity between "weather" and "temperature" is higher than a preset threshold, the main category set does not need to be updated, so the predicate with the highest priority in the main category set remains "weather".

[0047] The information acquisition method of this application further includes step S204: after obtaining the updated main category set, the priority of each predicate in the updated main category set is updated to obtain a further updated main category set; if the number of predicates in the further updated main category set exceeds a preset predicate threshold, the predicate with the lowest priority in the further updated main category set is deleted. In this embodiment, updating the priority of each predicate in the updated main category set ensures that the highest priority in the main category set is what the user most wants to focus on. Furthermore, deleting the predicate with the lowest priority in the further updated main category set when the number of predicates in the further updated main category set exceeds the preset predicate threshold saves memory.

[0048] In one specific embodiment, if the predicates in the updated main category set include (population, weather, area, attractions), then according to the first-order priority, "population" has the highest priority, followed by "weather," then "area" has a lower priority than "weather," and "attractions" has the lowest priority. If the preset predicate threshold is 3, then "attractions," with the lowest priority, can be deleted. That is, since "attractions" entered the main category set the earliest, after multiple user inquiries, the user is most concerned about "population." Of course, if the predicates in the updated main category set include (population, weather, area, attractions, snacks), and the preset predicate threshold remains 3, then both "attractions" and "snacks" can be deleted simultaneously, ensuring that the updated main category set contains the most recently updated and most user-concerned categories.

[0049] In practical applications, users may focus on the same object in multiple locations; for example, a user might simultaneously monitor the weather in Beijing and Shanghai. Therefore, in some embodiments, step S202 can be implemented by step S2023. Specifically, in step S2023, the priority of each subject in the sub-category set is sorted according to a second target time order, where the second target time is the order in which each subject enters the sub-category set from latest to earliest. The subject corresponding to the current question is merged into the current sub-category set to obtain an updated sub-category set. In other words, the subject corresponding to the current question can be directly merged into the current sub-category set to obtain an updated sub-category set.

[0050] In one specific embodiment, the user's previous question might be "What's the weather like in Beijing?". Classifying this previous question using a target classification model yields a subject of "Beijing" and a predicate of "weather." In the current question, the user asks "What about Shanghai?". Classifying this current question using the target classification model yields a subject of "Shanghai," so "Shanghai" can be directly added to the current sub-category set (Beijing), resulting in an updated sub-category set (Shanghai, Beijing).

[0051] The information acquisition method of this application also includes steps S205 and S206. Specifically:

[0052] Step S205: After incorporating the subject corresponding to the current question into the current sub-category set to obtain the updated sub-category set, the priority of each subject in the updated sub-category set is updated to obtain the updated sub-category set again.

[0053] Step S206: If the number of subjects in the updated sub-category set exceeds a preset subject threshold, delete the subject with the lowest priority in the updated sub-category set.

[0054] In this embodiment, the priority of each subject in the updated subcategory set is updated to obtain a further updated subcategory set. This ensures that the highest priority subject in the further updated subcategory set is the one the user most wants to focus on. Simultaneously, to save memory, if the number of subjects in the further updated subcategory set exceeds a preset subject threshold, the further updated subcategory set is deleted.

[0055] In one specific implementation, if the current subject is "Shanghai" and the current subcategory set is "Beijing", then after merging, the updated subcategory set is (Shanghai, Beijing). That is, "Shanghai" is given the highest priority, and "Beijing" has the next lowest priority.

[0056] Specifically, this application does not impose a limit on the size of the preset subject threshold. In practical applications, the preset subject threshold can be flexibly set according to the actual application scenario.

[0057] In another specific embodiment, if the updated sub-category set is (Shanghai, Beijing, Guangzhou, Shenzhen), where "Shanghai" has the highest priority, "Beijing" has a lower priority than "Shanghai," "Guangzhou" has a lower priority than "Beijing," and "Shenzhen" has a lower priority than "Guangzhou," and if the preset subject threshold is 3, then "Shenzhen," with the lowest priority, can be deleted, resulting in a sub-category set of (Shanghai, Beijing, Guangzhou). Alternatively, if the updated sub-category set is (Shanghai, Beijing, Guangzhou, Shenzhen, Suzhou), then "Suzhou," with the lowest priority, and "Shenzhen," with the second lowest priority, can be deleted, resulting in a sub-category set of (Shanghai, Beijing, Guangzhou).

[0058] In a specific implementation, step S203 can also be implemented through step S2031. Specifically, in step S2031, the priorities of the predicates in the main category set are sorted according to a first target time order, which is the order in which the predicates entered the main category set from latest to earliest. The priorities of the subjects in the sub-category set are sorted according to a second target time order, which is the order in which the subjects entered the sub-category set from latest to earliest. The subject with the highest priority in the updated sub-category set is combined with the predicate with the highest priority in the updated main category set to obtain the target question. This ensures that the target question obtained by combination is the one that the user is most concerned about, and further targets the user's most pressing concerns more specifically.

[0059] In one specific embodiment, if the updated main category set is (population, weather, area), and "population" has the highest priority, and the updated sub-category set is (Shanghai, Beijing, Guangzhou), meaning "Shanghai" has the highest priority, then the combined target question is "What is the population of Shanghai?", and based on this target question, "How much population does Shanghai have?" can be recommended to the user.

[0060] In one embodiment, step S203 can also be implemented through steps S2032 and S2033. Specifically:

[0061] Step S2032: The priority of each predicate in the main category set is sorted according to a first target time order, which is the time order in which each predicate enters the main category set from late to early. The priority of each subject in the sub-category set is sorted according to a second target time order, which is the time order in which each subject enters the sub-category set from late to early. When multiple answers to target questions can be displayed, based on the number of target questions for which answers can be displayed, target subjects with the same number of target questions for which answers can be displayed are determined from the updated sub-category set in descending order of priority.

[0062] Step S2033 involves combining each target subject with the predicate in the updated main category set to obtain multiple target questions. This allows for the identification of a larger number of target questions and the recommendation of answers to the questions that users are most concerned about.

[0063] In one specific embodiment, if there are two target questions with displayed answers, and the updated sub-category set is (Shanghai, Beijing, Guangzhou), then two subjects can be determined based on their priority from high to low: "Shanghai" and "Beijing". Simultaneously, if the updated main category set is (population, weather, area), then "Shanghai" and "Beijing" can be merged with the highest-priority "population" to recommend the population of Shanghai and the population of Beijing to the user.

[0064] In another specific embodiment, if there are four target questions with displayed answers, and the updated sub-category set is (Shanghai, Beijing, Guangzhou) and the updated main category set is (population, weather, area), then three subjects can be determined first: "Shanghai," "Beijing," and "Guangzhou" can be combined with the highest priority "population" to obtain the target questions "What is the population of Shanghai?", "What is the population of Beijing?", and "What is the population of Guangzhou?". However, at this point, one question is still missing. Therefore, "Shanghai," the highest priority subject in the updated sub-category set, can be combined with "weather," which has a lower priority than "population" in the updated main category set, to recommend "Shanghai's weather" to the user.

[0065] This application also provides an information acquisition device based on referential disambiguation. It should be noted that this information acquisition device based on referential disambiguation can be used to execute the information acquisition method based on referential disambiguation provided in this application. This information acquisition device is used to implement the embodiments and preferred embodiments described herein; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0066] The following describes the information acquisition device based on referential disambiguation provided in the embodiments of this application.

[0067] Figure 3 This is a schematic diagram of an information acquisition device based on referential disambiguation according to an embodiment of this application. Figure 3 As shown, the information acquisition device includes:

[0068] The classification unit 10 is used to classify the received current question using a target classification model to obtain the subject and predicate corresponding to the current question. The target classification model is a classification model trained on multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question.

[0069] In one specific embodiment, the target classification model can be obtained by training on classification models such as support vector machines, decision trees, and Naive Bayes.

[0070] Specifically, if the current question received is "What is the weather like in Beijing?", then by classifying the current question using the target classification model, we can obtain the subject as "Beijing" and the predicate as "weather".

[0071] In practical robot applications, the current question received by the robot can be issued by the user via voice, or by the user clicking on a display screen. This application does not limit the form in which the current question is issued.

[0072] The merging unit 20 is used to merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set, and to merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set. The main category set is a set composed of the predicates of multiple questions, and the sub-category set is a set composed of the subjects of multiple questions.

[0073] In the merging unit, the predicate corresponding to the current problem is merged with the current main category set, which means merging the predicate obtained by classifying the current problem with the predicates already existing in the current main category set.

[0074] In the merging unit, the subject corresponding to the current problem is merged with the current sub-category set, which means merging the subject obtained by classifying the current problem with the subjects already existing in the current sub-category set.

[0075] The combination unit 30 is used to combine the predicates in the updated main category set and the subjects in the updated sub-category set to obtain at least one target question, and to display the answer corresponding to the target question on the display screen.

[0076] Specifically, after obtaining the target question, the corresponding answer can be retrieved and displayed on the robot's screen or on a screen communicating with the robot. Secondly, after obtaining the answer, it's not limited to displaying it as text on the robot's screen or the screen communicating with the robot; it can also be played back as voice. Of course, it's also possible to simultaneously provide the user with the answer by playing back the answer as voice and displaying the target question on at least one of the robot's screen or the screen communicating with the robot. This not only makes the robot more intelligent but also more user-friendly.

[0077] In the aforementioned information acquisition device based on referential disambiguation, the classification unit uses a pre-trained target classification model to classify the received current question, obtaining the subject and predicate of the current question. The merging unit merges the predicate of the current question with existing predicates in the current main classification set, and merges the subject of the current question with existing subjects in the current sub-classification set, respectively obtaining updated main classification sets and updated sub-classification sets. The combination unit combines the predicates in the updated main classification set and the subjects in the updated sub-classification set to obtain at least one target question, and displays the corresponding answer to the target question on the display screen, thus recommending the answer corresponding to the target question to the user. Compared to existing multi-turn question-and-answer methods that recommend answers to user-defined target questions, this solution categorizes the current question to obtain its subject and predicate. It then merges the predicate of the current question with existing predicates in the current main category set, and merges the subject of the current question with existing subjects in the current sub-category set. This ensures that the resulting target question is the one the user most wants an answer for, and that the target question is more accurate. This allows for more targeted targeting of the user's most desired answer and faster recommendation of the corresponding answer, thus solving the problem of existing technologies struggling to accurately and quickly identify and address user-defined questions.

[0078] In practical applications, there are instances where a user asks about the weather in a certain location in one question, and then immediately asks about the temperature. Therefore, to obtain the target question—the question the user most wants answered—more accurately and reasonably, the implementation process involves sorting the predicates in the main category set according to a first target time order. This first target time order is the order in which the predicates entered the main category set from latest to earliest. The merging unit includes a first determining module and a first merging module. The first determining module determines the similarity between the predicate corresponding to the current question and the first predicate, obtaining a target similarity. The first predicate is the predicate with the highest priority in the current main category set. The first merging module merges the predicate corresponding to the current question into the current main category set when the target similarity is lower than a preset threshold, resulting in an updated main category set.

[0079] Specifically, this application does not limit the specific method for determining the similarity between the predicate corresponding to the current question and the first predicate. Any feasible method in the prior art for determining the similarity between the predicate corresponding to the current question and the first predicate can be used. In one specific embodiment, a method based on semantic distance or a method based on a statistical model can be used to determine the similarity between the predicate corresponding to the current question and the first predicate.

[0080] Specifically, the size of the preset threshold can be flexibly adjusted according to the actual application scenario, and this application does not limit the size of the preset threshold.

[0081] In one specific embodiment, if the previous question is "What is the weather like in Beijing?" and the current question is "What is the population of Beijing?", then the predicate of the previous question is "weather" and the predicate of the current question is "population". That is, "weather" is the predicate with the highest priority in the main category set. When the similarity between "weather" and "population" is lower than a preset threshold, the main category set is updated to (population, weather). If the current question is "What is the temperature in Beijing?", then the predicate of the previous question is "weather" and the predicate of the current question is "temperature". When the similarity between "weather" and "temperature" is higher than a preset threshold, the main category set does not need to be updated, so the predicate with the highest priority in the main category set remains "weather".

[0082] The information acquisition device of this application further includes a first updating unit and a first deleting unit. The first updating unit, after obtaining the updated main category set, updates the priority of each predicate in the updated main category set to obtain a further updated main category set. The first deleting unit is used to delete the predicate with the lowest priority in the further updated main category set if the number of predicates in the further updated main category set exceeds a preset predicate threshold. In this embodiment, updating the priority of each predicate in the updated main category set ensures that the highest priority predicate in the main category set is the one the user most wants to focus on. Furthermore, deleting the predicate with the lowest priority in the further updated main category set when the number of predicates exceeds the preset predicate threshold saves memory.

[0083] In one specific embodiment, if the predicates in the updated main category set include (population, weather, area, attractions), then according to the first-order priority, "population" has the highest priority, followed by "weather," then "area" has a lower priority than "weather," and "attractions" has the lowest priority. If the preset predicate threshold is 3, then "attractions," with the lowest priority, can be deleted. That is, since "attractions" entered the main category set the earliest, after multiple user inquiries, the user is most concerned about "population." Of course, if the predicates in the updated main category set include (population, weather, area, attractions, snacks), and the preset predicate threshold remains 3, then both "attractions" and "snacks" can be deleted simultaneously, ensuring that the updated main category set contains the most recently updated and most user-concerned categories.

[0084] In practical applications, users may focus on the same object in multiple locations; for example, a user might simultaneously monitor the weather in Beijing and Shanghai. Therefore, in some embodiments, the merging unit further includes a second merging module. The priority of each subject in the sub-category set is sorted according to a second target time order, where the second target time is the order in which each subject entered the sub-category set from latest to earliest. The second merging module is used to merge the subject corresponding to the current question into the current sub-category set, resulting in an updated sub-category set. In other words, the subject corresponding to the current question can be directly merged into the current sub-category set to obtain an updated sub-category set.

[0085] In one specific embodiment, the user's previous question might be "What's the weather like in Beijing?". Classifying this previous question using a target classification model yields a subject of "Beijing" and a predicate of "weather." In the current question, the user asks "What about Shanghai?". Classifying this current question using the target classification model yields a subject of "Shanghai," so "Shanghai" can be directly added to the current sub-category set (Beijing), resulting in an updated sub-category set (Shanghai, Beijing).

[0086] The information acquisition device of this application further includes a second updating unit and a second deleting unit. The second updating unit is used to update the priority of each subject in the updated sub-category set after incorporating the subject corresponding to the current question into the current sub-category set to obtain an updated sub-category set. The second deleting unit is used to delete the subject with the lowest priority in the updated sub-category set if the number of subjects in the updated sub-category set exceeds a preset subject threshold.

[0087] In this embodiment, the priority of each subject in the updated subcategory set is updated to obtain a further updated subcategory set. This ensures that the highest priority subject in the further updated subcategory set is the one the user most wants to focus on. Simultaneously, to save memory, if the number of subjects in the further updated subcategory set exceeds a preset subject threshold, the further updated subcategory set is deleted.

[0088] In one specific implementation, if the current subject is "Shanghai" and the current subcategory set is "Beijing", then after merging, the updated subcategory set is (Shanghai, Beijing). That is, "Shanghai" is given the highest priority, and "Beijing" has the next lowest priority.

[0089] Specifically, this application does not impose a limit on the size of the preset subject threshold. In practical applications, the preset subject threshold can be flexibly set according to the actual application scenario.

[0090] In another specific embodiment, if the updated sub-category set is (Shanghai, Beijing, Guangzhou, Shenzhen), where "Shanghai" has the highest priority, "Beijing" has a lower priority than "Shanghai," "Guangzhou" has a lower priority than "Beijing," and "Shenzhen" has a lower priority than "Guangzhou," and if the preset subject threshold is 3, then "Shenzhen," with the lowest priority, can be deleted, resulting in a sub-category set of (Shanghai, Beijing, Guangzhou). Alternatively, if the updated sub-category set is (Shanghai, Beijing, Guangzhou, Shenzhen, Suzhou), then "Suzhou," with the lowest priority, and "Shenzhen," with the second lowest priority, can be deleted, resulting in a sub-category set of (Shanghai, Beijing, Guangzhou).

[0091] In its implementation, the combination unit further includes a first combination module. Specifically, the priorities of the predicates in the main category set are sorted according to a first target time order, which is the order in which the predicates entered the main category set from latest to earliest. The priorities of the subjects in the sub-category set are sorted according to a second target time order, which is the order in which the subjects entered the sub-category set from latest to earliest. The first combination module is used to combine the highest priority subject in the updated sub-category set with the highest priority predicate in the updated main category set to obtain the target question. This ensures that the combined target question is the one that the user is most concerned about, further targeting the user's most pressing concerns more specifically.

[0092] In one specific embodiment, if the updated main category set is (population, weather, area), and "population" has the highest priority, and the updated sub-category set is (Shanghai, Beijing, Guangzhou), meaning "Shanghai" has the highest priority, then the combined target question is "What is the population of Shanghai?", and based on this target question, "How much population does Shanghai have?" can be recommended to the user.

[0093] In one embodiment, the combination unit further includes a second determining module and a second combining module. The priorities of the predicates in the main category set are sorted according to a first target time order, which is the order in which the predicates entered the main category set from latest to earliest. The priorities of the subjects in the sub-category set are sorted according to a second target time order, which is the order in which the subjects entered the sub-category set from latest to earliest. The second determining module is used to, when multiple target questions can have their answers displayed, determine target subjects from the updated sub-category set in descending order of priority, based on the number of target questions for which answers can be displayed. The second combining module is used to combine each target subject with the predicates in the updated main category set to obtain multiple target questions. This allows for the identification of a larger number of target questions and the recommendation of answers to more questions of greatest interest to the user.

[0094] In one specific embodiment, if there are two target questions with displayed answers, and the updated sub-category set is (Shanghai, Beijing, Guangzhou), then two subjects can be determined based on their priority from high to low: "Shanghai" and "Beijing". Simultaneously, if the updated main category set is (population, weather, area), then "Shanghai" and "Beijing" can be merged with the highest-priority "population" to recommend the population of Shanghai and the population of Beijing to the user.

[0095] In another specific embodiment, if there are four target questions with displayed answers, and the updated sub-category set is (Shanghai, Beijing, Guangzhou) and the updated main category set is (population, weather, area), then three subjects can be determined first: "Shanghai," "Beijing," and "Guangzhou" can be combined with the highest priority "population" to obtain the target questions "What is the population of Shanghai?", "What is the population of Beijing?", and "What is the population of Guangzhou?". However, at this point, one question is still missing. Therefore, "Shanghai," the highest priority subject in the updated sub-category set, can be combined with "weather," which has a lower priority than "population" in the updated main category set, to recommend "Shanghai's weather" to the user.

[0096] The information acquisition device based on referential disambiguation includes a processor and a memory. The classification unit, merging unit, and combination unit are all stored as program units in the memory, and the processor executes the program units stored in the memory to achieve the corresponding functions. All modules are located in the same processor; alternatively, the modules may be located in different processors in any combination.

[0097] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured, and by adjusting kernel parameters, questions that are difficult to accurately and quickly answer in existing technologies can be addressed.

[0098] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0099] This invention provides a computer-readable storage medium including a stored program, wherein the program, when running, controls the device where the computer-readable storage medium is located to execute the information acquisition method based on referential disambiguation.

[0100] Specifically, information acquisition methods based on referential disambiguation include:

[0101] Step S201: Use a target classification model to classify the received current question and obtain the subject and predicate corresponding to the current question. The target classification model is a classification model obtained by training multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question.

[0102] Step S202: Merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set; merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set; the main category set is a set composed of the predicates of multiple questions; the sub-category set is a set composed of the subjects of multiple questions.

[0103] Step S203: Based on the predicate in the updated main category set and the subject in the updated sub-category set, combine them to obtain at least one target question, and display the answer corresponding to the target question on the display screen.

[0104] This invention provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor is configured to execute the information acquisition method based on referential disambiguation through the computer program.

[0105] Specifically, information acquisition methods based on referential disambiguation include:

[0106] Step S201: Use a target classification model to classify the received current question and obtain the subject and predicate corresponding to the current question. The target classification model is a classification model obtained by training multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question.

[0107] Step S202: Merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set; merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set; the main category set is a set composed of the predicates of multiple questions; the sub-category set is a set composed of the subjects of multiple questions.

[0108] Step S203: Based on the predicate in the updated main category set and the subject in the updated sub-category set, combine them to obtain at least one target question, and display the answer corresponding to the target question on the display screen.

[0109] In a typical embodiment of this application, a robot is also provided. The robot includes an information acquisition device based on referential disambiguation, the information acquisition device being used to execute any of the described information acquisition methods based on referential disambiguation.

[0110] The robot includes an information acquisition device based on referential disambiguation, which is used to execute any of the aforementioned information acquisition methods based on referential disambiguation. Compared with the existing multi-turn question-and-answer method that recommends answers to target questions for users, this solution classifies the current question to obtain the subject and predicate of the current question. It then merges the predicate of the current question with existing predicates in the current main category set, and merges the subject of the current question with existing subjects in the current sub-category set. This ensures that the combined target question is the question the user most wants an answer for, and that the target question is more accurate, thus more effectively targeting the question the user most wants an answer for, and quickly recommending answers to the target question for the user. This solves the problem in existing technologies that struggle to accurately and quickly target user questions and provide targeted answers.

[0111] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs at least the following steps:

[0112] Step S201: Use a target classification model to classify the received current question and obtain the subject and predicate corresponding to the current question. The target classification model is a classification model obtained by training multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question.

[0113] Step S202: Merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set; merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set; the main category set is a set composed of the predicates of multiple questions; the sub-category set is a set composed of the subjects of multiple questions.

[0114] Step S203: Based on the predicate in the updated main category set and the subject in the updated sub-category set, combine them to obtain at least one target question, and display the answer corresponding to the target question on the display screen.

[0115] The devices mentioned in this article can be servers, PCs, tablets, mobile phones, etc.

[0116] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having at least the following method steps:

[0117] Step S201: Use a target classification model to classify the received current question and obtain the subject and predicate corresponding to the current question. The target classification model is a classification model obtained by training multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question.

[0118] Step S202: Merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set; merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set; the main category set is a set composed of the predicates of multiple questions; the sub-category set is a set composed of the subjects of multiple questions.

[0119] Step S203: Based on the predicate in the updated main category set and the subject in the updated sub-category set, combine them to obtain at least one target question, and display the answer corresponding to the target question on the display screen.

[0120] It will be apparent to those skilled in the art that the modules or steps of the present invention can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using device-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular hardware and software combination.

[0121] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0122] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0123] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0124] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0125] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0126] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0127] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0128] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0129] As can be seen from the above description, the embodiments described in this application achieve the following technical effects:

[0130] 1) In the information acquisition method based on reference disambiguation of this application, firstly, a pre-trained target classification model is used to classify the received current question, obtaining the subject and predicate of the current question. Then, the predicate of the current question is merged with the predicates already existing in the current main classification set, and the subject of the current question is merged with the subjects already existing in the current sub-classification set, respectively, to obtain updated main classification sets and updated sub-classification sets. Finally, based on the predicates in the updated main classification set and the subjects in the updated sub-classification set, at least one target question is obtained, and the answer corresponding to the target question is displayed on the screen, thus realizing the recommendation of the answer corresponding to the target question to the user. Compared to existing multi-turn question-and-answer methods that recommend answers to user-defined target questions, this solution categorizes the current question to obtain its subject and predicate. It then merges the predicate of the current question with existing predicates in the current main category set, and merges the subject of the current question with existing subjects in the current sub-category set. This ensures that the resulting target question is the one the user most wants an answer for, and that the target question is more accurate. This allows for more targeted targeting of the user's most desired answer and faster recommendation of the corresponding answer, thus solving the problem of existing technologies struggling to accurately and quickly identify and address user-defined questions.

[0131] 2) In the information acquisition device based on referential disambiguation of this application, the classification unit is used to classify the received current question using a pre-trained target classification model to obtain the subject and predicate of the current question. The merging unit is used to merge the predicate of the current question with the predicates already existing in the current main classification set, and to merge the subject of the current question with the subjects already existing in the current sub-classification set, to obtain updated main classification sets and updated sub-classification sets, respectively. The combination unit is used to combine the predicates in the updated main classification set and the subjects in the updated sub-classification set to obtain at least one target question, and to display the answer corresponding to the target question on the display screen, thus realizing the recommendation of the answer corresponding to the target question to the user. Compared to existing multi-turn question-and-answer methods that recommend answers to user-defined target questions, this solution categorizes the current question to obtain its subject and predicate. It then merges the predicate of the current question with existing predicates in the current main category set, and merges the subject of the current question with existing subjects in the current sub-category set. This ensures that the resulting target question is the one the user most wants an answer for, and that the target question is more accurate. This allows for more targeted targeting of the user's most desired answer and faster recommendation of the corresponding answer, thus solving the problem of existing technologies struggling to accurately and quickly identify and address user-defined questions.

[0132] 3) The robot of this application includes an information acquisition device based on referential disambiguation, which is used to execute any of the aforementioned information acquisition methods based on referential disambiguation. Compared with the existing multi-turn question-and-answer method that recommends answers to target questions for users, this solution classifies the current question to obtain the subject and predicate of the current question, merges the predicate of the current question with the predicates already existing in the current main category set, and merges the subject of the current question with the subjects already existing in the current sub-category set. This ensures that the target question obtained by combination is the question that the user most wants to answer, and that the obtained target question is more accurate, thus more specifically hitting the question that the user most wants to answer, and recommending the answer to the target question to the user more quickly. This solves the problem that it is difficult for the existing technology to accurately and quickly hit the user's question and provide a targeted answer.

[0133] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An information acquisition method based on referential disambiguation, characterized in that, include: A target classification model is used to classify the received current question to obtain the subject and predicate corresponding to the current question. The target classification model is a classification model trained by machine learning on multiple sets of sample data. Each set of sample data includes a training question and the subject and predicate corresponding to the training question. The predicate corresponding to the current question is merged with the current main category set to obtain the updated main category set. The subject corresponding to the current question is merged with the current sub-category set to obtain the updated sub-category set. The main category set is a set consisting of the predicates of multiple questions, and the sub-category set is a set consisting of the subjects of multiple questions. The predicates in the updated main category set and the subjects in the updated sub-category set are combined to obtain at least one target question, and the answer corresponding to the target question is displayed on the screen. The priority of each predicate in the main category set is ordered according to a first target time order, which is the order in which each predicate entered the main category set from latest to earliest. The predicate corresponding to the current problem is merged with the current main category set to obtain the updated main category set, including: Determine the similarity between the predicate corresponding to the current problem and the first predicate to obtain the target similarity, where the first predicate is the predicate with the highest priority in the current main classification set; If the target similarity is lower than a preset threshold, the predicate corresponding to the current question is merged into the current main category set to obtain the updated main category set.

2. The information acquisition method according to claim 1, characterized in that, After obtaining the updated main category set, the information acquisition method further includes: The priority of each predicate in the updated main category set is updated to obtain the updated main category set again; If the number of predicates in the updated main category set exceeds a preset predicate threshold, delete the predicate with the lowest priority in the updated main category set.

3. The information acquisition method according to claim 1, characterized in that, The priority of each subject in the sub-category set is sorted according to the second target time order, which is the order in which each subject entered the sub-category set from latest to earliest. The subject corresponding to the current question is merged with the current sub-category set to obtain the updated sub-category set, which includes: The subject corresponding to the current question is incorporated into the current sub-category set to obtain the updated sub-category set.

4. The information acquisition method according to claim 3, characterized in that, After incorporating the subject corresponding to the current question into the current sub-category set to obtain the updated sub-category set, the information acquisition method further includes: The priority of each subject in the updated sub-category set is updated to obtain the updated sub-category set again; If the number of subjects in the updated sub-category set exceeds a preset subject threshold, delete the subject with the lowest priority in the updated sub-category set.

5. The information acquisition method according to claim 1, characterized in that, The priority of each predicate in the main category set is ordered according to a first target time order, which is the order in which each predicate enters the main category set from latest to earliest. The priority of each subject in the subcategory set is ordered according to a second target time order, which is the order in which each subject enters the subcategory set from latest to earliest. By combining the predicates in the updated main category set and the subjects in the updated subcategory set, at least one target problem is obtained, including: The target problem is obtained by combining the subject with the highest priority in the updated sub-category set with the predicate with the highest priority in the updated main category set.

6. The information acquisition method according to claim 1, characterized in that, The priority of each predicate in the main category set is ordered according to a first target time order, which is the order in which each predicate enters the main category set from latest to earliest. The priority of each subject in the subcategory set is ordered according to a second target time order, which is the order in which each subject enters the subcategory set from latest to earliest. When multiple answers to the target questions can be displayed, at least one target question is obtained by combining the predicates in the updated main category set and the subjects in the updated subcategory set, including: Based on the number of target questions with displayable answers, target subjects with the same number of target questions with displayable answers are determined from the updated sub-category set in descending order of priority; Each of the target subjects is combined with the predicates in the updated main category set to obtain multiple target questions.

7. An information acquisition device based on referential disambiguation, characterized in that, include: The classification unit is used to classify the received current question using a target classification model to obtain the subject and predicate corresponding to the current question. The target classification model is a classification model trained on multiple sets of sample data using machine learning. Each set of sample data includes a training question and the subject and predicate corresponding to the training question. The merging unit is used to merge the predicate corresponding to the current question with the current main category set to obtain the updated main category set, and to merge the subject corresponding to the current question with the current sub-category set to obtain the updated sub-category set. The main category set is a set composed of the predicates of multiple questions, and the sub-category set is a set composed of the subjects of multiple questions. The combination unit is used to combine the predicates in the updated main category set and the subjects in the updated sub-category set to obtain at least one target question, and to display the answer corresponding to the target question on the display screen. The priority of each predicate in the main classification set is sorted according to a first target time order, which is the order in which each predicate entered the main classification set from latest to earliest. The merging unit includes: The first determining module is used to determine the similarity between the predicate corresponding to the current problem and the first predicate to obtain the target similarity, wherein the first predicate is the predicate with the highest priority in the current main classification set; The first merging module is used to merge the predicates corresponding to the current question into the current main classification set when the target similarity is lower than a preset threshold, so as to obtain the updated main classification set.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the information acquisition method based on referential disambiguation as described in any one of claims 1 to 6.

Images