WO2021135603A1 - Intention recognition method, server and storage medium - Google Patents

Abstract

An intention recognition method, a server and a storage medium, wherein the intention recognition method comprises: acquiring original sentence information of a user (S21); inputting the original sentence information into a preset shared named entity analysis engine to obtain an analysis result outputted by the shared named entity analysis engine (S22); if the analysis result indicates that only shared named entities are comprised in the original sentence information, then detecting whether a target round of dialogue corresponding to the original sentence information is a first round of dialogue (S23); and if the target round of dialogue is the first round of dialogue, then outputting intention categories corresponding to shared named entity categories to which the shared named entities belong, and determining a target intention category selected by the user from among the intention categories (S24). The intention recognition method is able to reduce the error rate of intention recognition and improve the accuracy of intention recognition.

Description

Intention recognition method, server and storage medium

This application claims the priority of the Chinese patent application filed with the State Intellectual Property Office on December 31, 2019, the application number is 201911417222.2, and the application name is "Intent Recognition Method, Server and Storage Medium", the entire content of which is incorporated herein by reference. Applying.

Technical field

This application relates to the field of artificial intelligence technology, and in particular to an intention recognition method, server and storage medium.

Background technique

With the rapid development of artificial intelligence technology, the application of man-machine dialogue technology in daily life is becoming more and more extensive. The most important thing in man-machine dialogue technology is the recognition of user intent, that is, the recognition of the intent expressed by the voice data input by the user. . Existing intent recognition methods usually first convert the voice data input by the user into corresponding original sentence information, and then input the original sentence information into the trained intent recognition model to obtain the user's intent category. However, when the original sentence information in the first round of human-machine dialogue contains only shared named entities, since shared named entities are usually applied to at least two types of user intent, the intent category determined directly through the intent recognition model is not It must be the real intention that the user wants to express. It can be seen that when the existing intention recognition method only includes the original sentence information of the shared named entity in the process of recognizing the first round of human-machine dialogue, there is a problem that the error rate of the intention recognition is high and the accuracy of the intention recognition is low.

Summary of the invention

The embodiments of the present application provide an intention recognition method, server, and storage medium, which can reduce the error rate of intention recognition and improve the accuracy of intention recognition.

In the first aspect, an embodiment of the present application provides an intention recognition method, including:

Get the user's original sentence information;

Inputting the original sentence information into a preset shared named entity analysis engine to obtain an analysis result output by the shared named entity analysis engine;

If the analysis result indicates that the original sentence information only contains a shared named entity, detecting whether the target dialogue round corresponding to the original sentence information is the first round of dialogue;

If the target dialogue round is the first round of dialogue, output the intent category corresponding to the shared named entity category to which the shared named entity belongs, and determine the target intent category selected by the user in the intent category.

In a possible implementation of the first aspect, the inputting the original sentence information into a preset shared named entity analysis engine to obtain an analysis result output by the shared named entity analysis engine includes:

Identifying named entities included in the original sentence information;

Identify the shared named entity among the named entities according to a preset list of shared named entity categories, and determine the shared named entity category to which the shared named entity belongs;

Determining the start position and the end position of the shared named entity in the original sentence information;

According to the starting position and the ending position of each of the shared named entities, it is analyzed whether the original sentence information only includes the shared named entity, and the analysis result is obtained.

In a possible implementation of the first aspect, according to the start position and the end position of each of the shared named entities, analyze whether the original sentence information includes only the shared named entity, and obtain the The analysis results include:

Determining the shared named entity whose starting position is the first position of the original sentence information as a candidate shared named entity;

If the end position of one of the candidate shared named entities is the end position of the original sentence information, it is determined that only the shared named entity is included in the original sentence information.

In a possible implementation of the first aspect, after the shared named entity whose starting position is the first position of the original sentence information is determined as a candidate shared named entity, the method further includes:

If the ending positions of all the candidate shared named entities are not the end positions of the original sentence information, the shared naming with the starting position being the position after the ending position of any one of the candidate shared named entities is executed in a loop The step of determining whether the entity is a new candidate shared named entity, and detecting whether the end position of the new candidate shared named entity is the end position of the original sentence information;

After traversing all the shared named entities, if the end positions of all the new candidate shared named entities are not the end positions of the original sentence information, it is determined that the original sentence information does not only include the shared named entity.

In a possible implementation of the first aspect, according to the start position and the end position of each of the shared named entities, analyze whether the original sentence information includes only the shared named entity, and obtain the The analysis results include:

Defining a flag bit array with the same length as the length of the original sentence information, and setting the value of each flag bit in the flag bit array to a first preset value;

The shared named entity whose starting position is the first position of the original sentence information is determined as the first target shared named entity, and the value of the flag corresponding to the end position of the first target shared named entity is updated to the first target shared named entity. Two preset values;

Determine the shared named entity whose starting position is not the first position of the original sentence information as the second target shared named entity, and detect the mark corresponding to the previous position of the starting position of each second target shared named entity Whether the value of the bit is the second preset value;

If the value of the flag bit corresponding to the previous position of the start position of the second target shared named entity is the second preset value, set the value of the flag bit corresponding to the end position of the second target shared named entity The value is updated to the second preset value;

After traversing all the shared named entities, if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the second preset value, it is determined that the original sentence information only contains the shared named entity ；

After traversing all the shared named entities, if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the first preset value, it is determined that the original sentence information does not only include the shared name entity.

In a possible implementation of the first aspect, according to the start position and the end position of each of the shared named entities, analyze whether the original sentence information includes only the shared named entity, and obtain the The analysis results include:

If there is no shared named entity whose starting position is the first position of the original sentence information in the shared named entity, it is determined that the original sentence information does not only include the shared named entity.

In a possible implementation of the first aspect, after detecting whether the target dialogue round corresponding to the original sentence information is the first round of dialogue, the method further includes:

If the target dialogue round is not the first round of dialogue, acquiring historical original sentence information of the user in the historical dialogue round before the target dialogue round;

According to the historical original sentence information, the target intention category corresponding to the original sentence information is determined.

In the second aspect, an embodiment of the present application provides a server, including:

The first obtaining unit is used to obtain the original sentence information of the user;

The second obtaining unit is configured to input the original sentence information into a preset shared named entity analysis engine to obtain the analysis result output by the shared named entity analysis engine;

The first detecting unit is configured to detect whether the target dialogue round corresponding to the original sentence information is the first round of dialogue if the analysis result indicates that the original sentence information contains only shared named entities;

The first determining unit is configured to, if the target dialogue round is the first round of dialogue, output an intent category corresponding to the shared named entity category to which the shared named entity belongs, and determine that the user selects among the intent categories The target intent category.

In a third aspect, an embodiment of the present application provides a server, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The processor executes the computer program when the computer program is executed. The intention recognition method as described in the first aspect above.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the intention recognition as described in the first aspect is realized. method.

In a fifth aspect, the embodiments of the present application provide a computer program product, which when the computer program product runs on a server, causes the server to execute the intention recognition method described in any one of the above-mentioned first aspects.

Compared with the prior art, the embodiments of this application have the following beneficial effects:

In an intention recognition method provided by an embodiment of the present application, after obtaining the original sentence information of the user, the original sentence information is not directly input into the traditional intent recognition model to determine the intention category expressed by the user, but the The original sentence information is input into the preset shared named entity analysis engine, and the shared named entity analysis engine is used to analyze whether the original sentence information contains only the shared named entity, and the original sentence information contains only the shared named entity, and the original sentence information When the corresponding target dialogue round is the first round of dialogue, by outputting the intent category corresponding to the shared named entity category to which the shared named entity belongs, the user can select the expressed target intent category from the intent categories. The category is obtained through further confirmation by the user, so it can reduce the error rate of intent recognition and improve the accuracy of intent recognition.

Description of the drawings

FIG. 1 is a schematic structural diagram of a human-machine dialogue system to which an intention recognition method provided by an embodiment of the present application is applicable;

FIG. 2 is a schematic flowchart of an intention recognition method provided by an embodiment of the present application;

FIG. 3 is a specific schematic flowchart of S22 in an intention recognition method provided by an embodiment of the present application;

FIG. 4 is a specific schematic flowchart of S224 in an intention recognition method provided by an embodiment of the present application;

FIG. 5 is a specific schematic flowchart of S224 in an intention recognition method provided by another embodiment of the present application;

FIG. 6 is a schematic flowchart of an intention recognition method provided by another embodiment of the present application;

FIG. 7 is a structural block diagram of a server provided by an embodiment of the present application;

FIG. 8 is a schematic structural diagram of a server provided by another embodiment of the present application.

Detailed ways

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are proposed for a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from obstructing the description of this application.

It should be understood that when used in the specification and appended claims of this application, the term "comprising" indicates the existence of the described features, wholes, steps, operations, elements and/or components, but does not exclude one or more other The existence or addition of features, wholes, steps, operations, elements, components, and/or collections thereof.

It should also be understood that the term "and/or" used in the specification and appended claims of this application refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

As used in the description of this application and the appended claims, the term "if" can be construed as "when" or "once" or "in response to determination" or "in response to detecting ". Similarly, the phrase "if determined" or "if detected [described condition or event]" can be interpreted as meaning "once determined" or "in response to determination" or "once detected [described condition or event]" depending on the context ]" or "in response to detection of [condition or event described]".

In addition, in the description of the specification of this application and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

Reference to "one embodiment" or "some embodiments" described in the specification of this application means that one or more embodiments of this application include a specific feature, structure, or characteristic described in combination with the embodiment. Therefore, the sentences "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. appearing in different places in this specification are not necessarily All refer to the same embodiment, but mean "one or more but not all embodiments" unless it is specifically emphasized otherwise. The terms "including", "including", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

Please refer to FIG. 1. FIG. 1 is a schematic architecture diagram of a human-machine dialogue system to which an intention recognition method provided by an embodiment of the present application is applicable. As shown in FIG. 1, the man-machine dialogue system 100 provided in this embodiment includes a man-machine dialogue terminal 110 and a man-machine dialogue server 120. Among them, the human-machine dialogue terminal 110 includes, but is not limited to, mobile phones, tablet computers, smart TVs, wearable devices, in-vehicle devices, augmented reality (AR)/virtual reality (VR) devices, notebook computers, and ultra mobile devices. For mobile terminals such as ultra-mobile personal computers (UMPC), netbooks, and personal digital assistants (PDAs), the embodiment of the application does not impose any restriction on the specific type of the human-machine dialogue terminal 110.

In this embodiment, when a user conducts a man-machine dialogue with the man-machine dialogue system 100, the man-machine dialogue terminal 110 in the man-machine dialogue system 100 can establish a wireless communication connection or a wired communication connection with the man-machine dialogue server 120, thereby realizing Wireless communication or wired communication between the man-machine dialogue terminal 110 and the man-machine dialogue server 120. Specifically, during the man-machine dialogue, the man-machine dialogue terminal 110 may collect voice data from the user through its voice collection module. The man-machine dialogue terminal 110 may convert the voice data from the user into corresponding original sentence information, and then send the original sentence information to the man-machine dialogue server 120 through wireless communication or wired communication; or, the man-machine dialogue terminal 110 may directly The voice data from the user is sent to the man-machine dialogue server 120 through wireless communication or wired communication, and the man-machine dialogue server 120 converts the voice data from the user into corresponding original sentence information. After obtaining the user's original sentence information, the human-machine dialogue server 120 recognizes the target intention category of the original sentence information, and feeds back the target intention category to the human-machine dialogue terminal 110 through wireless communication or wired communication.

Please refer to FIG. 2. FIG. 2 is a schematic flowchart of an intention recognition method provided by an embodiment of the present application. In this embodiment, the execution subject of the process is a server, and the server may specifically be a man-machine dialogue in a man-machine dialogue system. server. As shown in FIG. 2, an intention recognition method provided by this embodiment includes S21 to S24, which are described in detail as follows:

S21: Obtain the original sentence information of the user.

In this embodiment, the user's original sentence information refers to text information obtained by translating the voice data from the user in the process of man-machine dialogue. In specific applications, in order to accurately identify the type of intent that the user wants to express in the process of man-machine dialogue, it is usually necessary to conduct at least one round of man-machine dialogue, and each round of man-machine dialogue needs to convert the voice data from the user into the corresponding Original sentence information, and intent recognition is performed on the original sentence information in the human-machine dialogue process.

In specific applications, when conducting a human-machine dialogue, the man-machine dialogue terminal in the man-machine dialogue system can collect the user's voice data through its voice collection module. In a possible implementation of this embodiment, the human-machine dialogue terminal can perform voice-to-text processing on the collected user’s voice data to obtain the original sentence information corresponding to the user’s voice data, and combine it with the user’s voice The original sentence information corresponding to the data is sent to the man-machine dialogue server in the man-machine dialogue system, and the man-machine dialogue server receives the user's original sentence information sent by the man-machine dialogue terminal. In another possible implementation of this embodiment, the human-machine dialogue terminal can directly send the collected voice data of the user to the human-machine dialogue server, and the human-machine dialogue server performs audio-to-text processing on the user’s voice data. Obtain the original sentence information corresponding to the user's voice data.

As an example and not a limitation, the human-machine dialogue terminal or the man-machine dialogue server may convert the voice data from the user into corresponding original sentence information based on Automatic Speech Recognition (ASR) technology.

S22: Input the original sentence information into a preset shared named entity analysis engine, and obtain an analysis result output by the shared named entity analysis engine.

In this embodiment, the preset shared named entity analysis engine is pre-configured with an analysis algorithm for analyzing whether the sentence information contains only shared named entities, that is, the preset shared named entity analysis engine can analyze whether the sentence information contains only shared named entities. Contains shared named entities.

It should be noted that a named entity refers to an object identified by a name, and it can be an object represented by any noun. Generally, named entities can be divided into different categories such as person names, place names, organization names, and song names. Each named entity category usually includes multiple named entities of the same category, for example, "place names" under the named entity category It can include multiple named entities belonging to place names, such as Beijing, Shanghai, and Guangzhou.

Shared named entities refer to named entities that can be included in at least two types of intents and shared by at least two types of intents. Exemplarily, because taxi intention and navigation intention usually need to know the origin and/or destination, and the origin and destination belong to the named entity of "place name", that is, the named entity of "place name" is usually included in In the taxi intent and navigation intent, therefore, the named entity of the “place name” category is a shared named entity.

In this embodiment, after obtaining the user's original sentence information, the man-machine dialogue server inputs the user's original sentence information into the preset shared named entity analysis engine to analyze the original sentence information through the shared named entity analysis engine Whether to include only shared named entities in the file, and then get the analysis results output by the shared named entity analysis engine.

In a specific embodiment of the present application, the shared named entity analysis engine can analyze whether the original sentence information contains only shared named entities through S221 to S224 as shown in FIG. 3, as detailed below:

S221: Identify the named entity contained in the original sentence information.

In this embodiment, when the man-machine dialogue server analyzes whether the original sentence contains only the shared named entity through the shared named entity analysis engine, it needs to first identify the named entity contained in the original sentence information.

In a possible implementation of this embodiment, the human-machine dialogue server can perform a named entity recognition (NER) operation on the original sentence information based on a preset named entity recognition tool. Among them, the preset named entity recognition tool can identify all the named entities contained in the original sentence information, and can obtain the information of each named entity. It is understandable that the named entity contained in the original sentence information may be one or at least two, which is specifically determined according to the actual situation, and there is no limitation here.

Among them, the information of the named entity may include, but is not limited to, the category of the named entity to which the named entity belongs and the start position and end position of the named entity in the original sentence information. Among them, the start position refers to the position of the first character in the named entity in the original sentence information, the end position refers to the position of the last character in the named entity in the original sentence information, and the characters in the named entity are in the original sentence information. The position of the character can be identified by the order of the character in the original sentence information. Exemplarily, assuming that the original sentence information is "Take a taxi to Beijing Botanical Garden", the order of the characters from left to right in the original sentence information in the original sentence information can be 0, 1, 2, 3, 4, 5. 6 and 7, therefore, the position of each character from left to right in the original sentence information can be identified by 0, 1, 2, 3, 4, 5, 6, and 7, respectively. If the preset named entity recognition tool is used to perform named entity recognition on the original sentence information of "Take a taxi to Beijing Botanical Garden", it can be recognized that the original sentence information contains "Beijing", "Botanical Garden" and "Beijing Botanical Garden". Three named entities, and the named entity category to which the named entity "Beijing" belongs is "place name", the starting position of the original sentence information is 3, and the ending position is 4; the named entity "Botanical Garden" belongs to The named entity category of "Beijing Botanical Garden" is "place name", which has a starting position of 5 and an ending position of 7 in the original sentence information; the named entity category of the named entity of "Beijing Botanical Garden" is "place name", which is in the original sentence information The start position is 3 and the end position is 7.

S222: Identify the shared named entity in the named entity according to the preset list of shared named entity categories, and determine the shared named entity category to which the shared named entity belongs.

After identifying the named entities contained in the original sentence information, it is necessary to further identify whether there are shared named entities among these named entities. In this embodiment, the human-machine dialogue server can identify the shared named entity among the named entities according to a preset list of shared named entity categories. Among them, the preset shared named entity category list is used to store pre-configured shared named entity categories and intent categories corresponding to each shared named entity category.

In a possible implementation of this embodiment, the preset shared named entity category list may be obtained according to a preset named entity category configuration file. Specifically, before identifying the shared named entity contained in the original sentence information, the human-machine dialogue system can be configured with corresponding intent categories according to the functions that can be realized by the human-machine dialogue system, where different functions correspond to different intent categories. Exemplarily, assuming that the human-machine dialogue system can realize functions such as navigation or taxiing, the user may express the intention of taxiing or navigating to the human-machine dialogue system when communicating with the human-machine dialogue system. Therefore, it can be a human-machine dialogue system. Configure taxi intent or navigation intent, etc. Since the intent of each category must usually include at least one category of named entities, for example, taxi intent and navigation intent must generally include named entities in the category of "place name", so the necessary information can be included in each category of intent. , Configure the corresponding named entity category for each intent category, the man-machine dialogue server can store the named entity category configured for each intent category in the preset named entity category configuration file, that is, the named entity category configuration file is used to store the pre-defined The named entity category configured for each intent category, for example, please refer to Table 1. Table 1 shows part of the content stored in the named entity category configuration file, where named entity category 2 is configured in both the intent A and the intent. In B, therefore, named entity category 2 is a shared named entity category.

Table 1

After the man-machine dialogue server obtains the pre-configured named entity category configuration file, it can perform the shared named entity detection on the named entity category configuration file, that is, check whether there is at least one named entity category configured in at least two of the named entity category configuration files. In the intent category, if it is detected that at least one named entity category is configured in at least two intent categories, it is determined that the at least one named entity category is a shared named entity category. For example, named entity category 2 in Table 1 is configured at the same time In Intent A and Intent B, therefore, named entity category 2 in Table 1 is a shared named entity category. The human-machine dialogue server can associate each detected shared named entity category with its corresponding at least two intent categories and store them in a preset shared named entity category list, that is, the shared named entity category list is used to store each shared named entity category The intent category corresponding to it. Exemplarily, please refer to Table 2. Table 2 shows part of the content stored in the shared named entity category list, where the intention categories corresponding to the shared named entity category 2 include intention A and intention B. In a specific application, the man-machine dialogue server can store a preset list of shared named entity categories in its memory.

Table 2

In this embodiment, when the man-machine dialogue server recognizes the shared named entity in the named entity contained in the original sentence information, it can obtain a preset list of shared named entity categories from its memory, and then according to the preset shared named entity category The shared named entity category contained in the list identifies the shared named entity among the named entities contained in the original sentence information, and determines the shared named entity category to which each shared named entity belongs. Specifically, if the first named entity contained in the original sentence information belongs to the first shared named entity category in the list of shared named entity categories, the first named entity is identified as a shared named entity, and the shared named entity to which the shared named entity belongs is determined The named entity category is the first shared named entity category. It should be noted that there may be one or at least two shared named entities included in the original sentence information. Exemplarily, suppose the original sentence information is "Take a taxi to Beijing Botanical Garden", the named entities contained in the original sentence information include "Beijing", "Botanical Garden" and "Beijing Botanical Garden", suppose that the list of shared named entity categories contains "place name" This shared named entity category, because "Beijing", "Botanical Garden" and "Beijing Botanical Garden" are named entities of the "place name" category, therefore, the human-computer dialogue server will be "Beijing", "Botanical Garden" and "Beijing Botanical Garden". "These three named entities are all identified as shared named entities, and the shared named entity category to which the three shared named entities "Beijing", "Botanical Garden" and "Beijing Botanical Garden" belong is "place name".

S223: Determine the start position and the end position of the shared named entity in the original sentence information.

Since in S221, after the named entity recognition operation is performed on the original sentence information, the start position and end position of each named entity contained in the original sentence information in the original sentence information have been obtained, therefore, this embodiment determines in S222 After the shared named entities in the named entities are extracted, the start position and end position of each shared named entity in the original sentence information can be directly obtained.

S224: According to the start position and the end position of each of the shared named entities, analyze whether the original sentence information only includes the shared named entity, and obtain the analysis result.

In this embodiment, after the man-machine dialogue server determines the start position and end position of each shared named entity contained in the original sentence information in the original sentence information, it can be based on the start position of each shared named entity in the original sentence information. Start position and end position to detect whether only shared named entities are included in the original sentence information.

In an embodiment of the present application, S224 may be specifically implemented through S2241 to S2244 as shown in FIG. 4, which are described in detail as follows:

S2241: Determine the shared named entity whose starting position is the first position of the original sentence information as a candidate shared named entity.

In this embodiment, when the man-machine dialogue server detects whether the original sentence information contains only the shared named entity based on the start position and end position of each shared named entity in the original sentence information, it can first detect the content contained in the original sentence information. Whether there is a shared named entity whose starting position is the first position of the original sentence information in the shared named entity. The first position of the original sentence information refers to the position of the first character in the original sentence information, and the end position of the original sentence information refers to the position of the last character in the original sentence information. Exemplarily, the first position of the original sentence information "Taking a taxi to Beijing Botanical Garden" is the position where the first character "打" is located, that is, the identification of the first position of the original sentence information "Taking a taxi to Beijing Botanical Garden" is 0; the original sentence information The last position of "Taking a taxi to Beijing Botanical Garden" is the position where the last character "door" is located, that is, the identification of the last position of the original sentence information "Taking a taxi to Beijing Botanical Garden" is 7.

When the man-machine dialogue server detects that there is a shared named entity whose starting position is the first position of the original sentence information in the shared named entity contained in the original sentence information, it determines all the shared named entities whose starting position is the first position of the original sentence information It is a candidate shared named entity, and detects whether the end position of each candidate shared named entity is the end position of the original sentence information. Exemplarily, assuming that the original sentence information is "Beijing Botanical Garden", since the shared named entities "Beijing" and "Beijing Botanical Garden" contained in the original sentence information are the first positions of the original sentence information, the The shared named entities "Beijing" and "Beijing Botanical Garden" are both identified as candidate shared named entities. Further, the man-machine dialogue server separately detects whether the end positions of "Beijing" and "Beijing Botanical Garden" in the original sentence information are the last positions of the original sentence information. In this example, the end position of "Beijing" in the original sentence information is not the end position of the original sentence information, and the end position of "Beijing Botanic Garden" in the original sentence information is the end position of the original sentence information.

In this implementation, if the man-machine dialogue server detects that the end position of a candidate shared named entity is the end position of the original sentence information, S2242 is executed; if the man-machine dialogue server detects that the end position of all candidate shared named entities is not At the end of the original sentence information, S2243~2244 are executed. It should be noted that in this embodiment, S2242 and S2243 to S2244 are parallel steps, that is, when the man-machine dialogue server executes S2242, S2243 to S2244 are not executed; that is, when the man-machine dialogue server executes S2243 to S2244, S2242 is not executed.

S2242: If the end position of one of the candidate shared named entities is the end position of the original sentence information, it is determined that the original sentence information only includes the shared named entity.

In this implementation, when the man-machine dialogue server detects that the end position of a candidate shared named entity in the original sentence information is the end position of the original sentence information, because the candidate shared named entity is at the start position in the original sentence information It is the first position of the original sentence information, so it means that all the characters in the original sentence information constitute the candidate shared named entity, which means that the original sentence information only contains the shared named entity and does not contain other information. At this time, the man-machine dialogue The server determines that only shared named entities are included in the original sentence information. Exemplarily, in combination with the example in S2241, since the end position of the candidate shared named entity "Beijing Botanical Garden" in the original sentence information "Beijing Botanical Garden" is the end position of the original sentence information, it is determined that only the original sentence information "Beijing Botanical Garden" Contains shared named entities.

S2243: If the end positions of all the candidate shared named entities are not the end positions of the original sentence information, execute the cyclically executing the position that sets the start position to the end position of any one of the candidate shared named entities. The step of determining that the shared named entity is a new candidate shared named entity, and detecting whether the end position of the new candidate shared named entity is the end position of the original sentence information.

In this embodiment, when the human-machine dialogue server detects that the end positions of all candidate shared named entities are not the end positions of the original sentence information, it indicates that none of the candidate shared named entities starts from the first position of the original sentence information to the original sentence. The end position of the information ends. At this time, for each candidate shared named entity, the man-machine dialogue server detects whether there is a shared named entity located after the candidate shared named entity and adjacent to the candidate shared named entity in the original sentence information. That is, it is detected whether there is a shared named entity whose starting position is a position after the ending position of any candidate shared named entity in the original sentence information. If the man-machine dialogue server detects that there is at least one shared named entity whose starting position is the end position of any candidate shared named entity in the original sentence information, it will determine the at least one shared named entity as a new candidate shared entity Named entities. The human-machine dialogue server detects whether the end position of each new candidate shared named entity is the end position of the original sentence information.

Specifically, if the human-machine dialogue server detects that the end position of at least one candidate shared named entity among the new candidate shared named entities is the end position of the original sentence information, it means that the original sentence information is only composed of the new candidate shared named entity and The candidate shared named entity that is adjacent to the new candidate shared named entity and is located before the new candidate shared named entity is composed of the candidate shared named entity, which means that the original sentence information only contains the shared named entity. Exemplarily, if the original sentence information is "Botanic Garden Zoo", since the start position of the shared named entity "Zoo" is a position after the end position of the candidate shared named entity "Botanical Garden", the shared named entity "Zoo" will be shared. It is determined as a new candidate shared named entity. Furthermore, since the end position of the new candidate shared named entity "zoo" is the end position of the original sentence information, it is determined that the original sentence information "Botanical Garden Zoo" only contains the shared named entity.

If the human-machine dialogue server detects that the ending position of all new candidate shared named entities is not the end position of the original sentence information, it will continue to loop to detect whether there is a starting position in the original sentence information that is the end position of any candidate shared named entity If there is a shared named entity in the latter position, the shared named entity whose starting position is the ending position of any candidate shared named entity is determined as a new candidate shared named entity, and each new candidate is detected Whether the end position of the shared named entity is the last position of the original sentence information, until all the shared named entities in the original sentence information are traversed, if after traversing all the shared named entities in the original sentence information, there is no candidate for sharing The end position of the named entity is the end position of the original sentence information, and the man-machine dialogue server executes S2244.

S2244: After traversing all the shared named entities, if the end positions of all the new candidate shared named entities are not the end positions of the original sentence information, it is determined that the original sentence information does not only include shared names entity.

In this implementation, after the man-machine dialogue server has traversed all the shared named entities in the original sentence information, if it detects that the end position of none of the candidate shared named entities is the end position of the original sentence information, it means that the original sentence information except In addition to the shared named entity, it also contains other information. At this time, the man-machine dialogue server determines that the original sentence information does not only include the shared named entity. Exemplarily, assuming that the original sentence information is "How to get to Beijing Botanical Garden", the shared named entity "Beijing" can be determined as a candidate shared named entity according to S2241, and the shared named entity "Botanical Garden" can be determined as a new candidate shared named entity according to S2243 , And the end position of the new shared named entity "Botanic Garden" is not the end position of the original sentence information "How to get to Beijing Botanical Garden", because at this time all the shared named entities in the original sentence information "How to get to Beijing Botanical Garden" have been traversed, And the end position of none of the candidate shared named entities is the end position of the original sentence information "How to get to Beijing Botanical Garden", therefore, it is determined that the end position of original sentence information "How to get to Beijing Botanical Garden" does not only include the shared named entity.

In another embodiment of the present application, if the human-machine dialogue server detects that there is no shared named entity whose starting position is the end position of any candidate shared named entity in the original sentence information, it indicates that the original sentence information There is no shared named entity adjacent to each candidate shared named entity, which means that there are other information between at least two shared named entities in the original sentence information. At this time, the man-machine dialogue server determines that the original sentence information does not only contain shared Named entities. Exemplarily, suppose the original sentence information is "how to get from the botanical garden to the zoo", since the last position of the end position of the candidate shared named entity "botanic garden" is the position of "to", and the start of the shared named entity "zoo" The location is the location of "Long". Therefore, in the original sentence information "How to get from the Botanical Garden to the Zoo", there is no shared named entity whose starting position is the end position of the candidate shared named entity "Botanical Garden". At this time, It is determined that the original sentence information "How to get from the botanical garden to the zoo" does not only include shared named entities.

In another embodiment of the present application, S224 can also be specifically implemented through S2245 to S2240 as shown in FIG. 5, which is described in detail as follows:

S2245: Define a flag bit array with the same length as the length of the original sentence information, and set the value of each flag bit in the flag bit array to a first preset value.

In this embodiment, when the human-machine dialogue server detects whether the original sentence information contains only shared named entities, it can first define a flag bit array with the same length as the length of the original sentence information, where each flag in the flag bit array The bits respectively correspond to the positions of the characters in the original sentence information. Exemplarily, if the original sentence information is "Beijing Botanical Garden", a flag bit array with a length of 5 can be defined. The first flag bit in the flag bit array is the same as the first character "North" in the original sentence information "Beijing Botanical Garden". ”Corresponds to the position, and the second flag bit in the flag bit array corresponds to the position of the second character “京” in the original sentence information “Beijing Botanical Garden”.

In this embodiment, after the man-machine dialogue server defines the flag bit array, it can first set the value of each flag bit in the flag bit array to the first preset value. The first preset value may be any value in the Boolean logic value. For example, the first preset value may be 0 in the Boolean logic value or 1 in the Boolean logic value. It should be noted that this embodiment will also involve a second preset value. The second preset value can also be any value of Boolean logic values, but the second preset value is different from the first preset value. When the first preset value is 0, the second preset value is 1, and when the first preset value is 1, the second preset value is 0.

In this embodiment, the human-machine dialogue server also detects whether the original sentence information contains a shared named entity whose starting position is the first position of the original sentence information. If the man-machine dialogue server detects that the original sentence information contains at least one shared named entity whose starting position is the first position of the original sentence information, S2246 to S2240 are executed.

S2246: Determine the shared named entity whose start position is the first position of the original sentence information as the first target shared named entity, and update the value of the flag bit corresponding to the end position of the first target shared named entity Is the second preset value.

In this embodiment, if the man-machine dialogue server detects that the original sentence information contains at least one shared named entity whose starting position is the first position of the original sentence information, it will share all starting positions as the first position of the original sentence information. The named entity is determined to be the first target shared named entity, and the values of the flag bits corresponding to the end positions of all the first target shared named entities are updated to the second preset value. Exemplarily, assuming that the original sentence information is "How to get to Beijing Botanical Garden", since the starting positions of the shared named entities "Beijing" and "Beijing Botanical Garden" are the first positions of the original sentence information, the named entities "Beijing" and "Beijing Botanic Garden" is determined as the first target shared named entity, and the value of the flag corresponding to the end position of "Beijing" (ie the position of "京") is updated to the second preset value, and the value of "Beijing Botanic Garden" The value of the flag bit corresponding to the end position (that is, the position of the "door") is updated to the second preset value.

S2247: Determine the shared named entity whose starting position is not the first position of the original sentence information as the second target shared named entity, and detect that each of the second target shared named entities corresponds to the previous position of the starting position Whether the value of the flag bit is the second preset value.

S2248: If the value of the flag bit corresponding to the previous position of the start position of the second target shared named entity is the second preset value, set the flag corresponding to the end position of the second target shared named entity The value of the bit is updated to the second preset value.

In this embodiment, the human-machine dialogue server also determines the shared named entity whose starting position is not the first position of the original sentence information as the second target shared named entity. Exemplarily, in combination with the example in S2246, since the starting position of the shared named entity "Botanical Garden" in the original sentence information "How to get to Beijing Botanical Garden" is not the first position of the original sentence information, the shared named entity "Botanical Garden" is determined Share a named entity for the second target.

After determining the second target shared named entity, the man-machine dialogue server detects whether the value of the flag bit corresponding to the previous position of the starting position of each second target shared named entity is the second preset value. If the human-machine dialogue server detects that the value of the flag corresponding to the first position of the second target shared named entity is the second preset value, it indicates that the second target shared named entity is before the start position of the shared named entity. A position is the end position of a first target shared named entity, which means that the second target shared named entity is adjacent to a first target shared named entity in the original sentence information. The value of the flag bit corresponding to the end position of the target shared named entity is updated to the second preset value.

In this embodiment, after the man-machine dialogue server has traversed all the second target shared named entities, it detects whether the updated value of the flag bit corresponding to the end position of the original sentence information is the second preset value. If the man-machine dialogue server detects that the flag bit corresponding to the end position of the original sentence information is updated to the second preset value, execute S2249; if the man-machine dialogue server detects the flag bit corresponding to the end position of the original sentence information After the updated value is the first preset value, S2240 is executed.

In another embodiment of the present application, if the human-machine dialogue server detects that the value of the flag corresponding to the previous position of the start position of a second target shared named entity is the first preset value, it indicates that the second target The shared named entity is not adjacent to any first target shared named entity in the original sentence information. At this time, the man-machine dialogue server does not update the value of the flag bit corresponding to the end position of the second target shared named entity.

S2249: After traversing all the shared named entities, if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the second preset value, then it is determined that the original sentence information only contains shared Named entities.

In this embodiment, after the human-machine dialogue server has traversed all shared named entities, if it detects that the updated value of the flag bit corresponding to the end position of the original sentence information is the second preset value, it indicates that the original sentence information From the first position to the end of the original sentence information, is composed of at least one shared named entity that is adjacent to the end, that is, the original sentence information does not contain other information except the shared named entity. At this time, the man-machine dialogue server It is determined that only shared named entities are included in the original sentence information.

S2240: After traversing all the shared named entities, if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the first preset value, it is determined that the original sentence information does not only contain Shared named entities.

In this embodiment, after the human-machine dialogue server has traversed all the shared named entities, if it detects that the updated value of the flag bit corresponding to the end position of the original sentence information is the first preset value, it indicates that the original sentence information From the first position of the original sentence to the end of the original sentence information, it is not composed of at least one shared named entity that is adjacent to the end. That is, in addition to the shared named entity, the original sentence information also contains other information. At this time, the man-machine dialogue The server determines that the original sentence information does not include only shared named entities.

In another possible implementation manner of this embodiment, S224 may further include the following steps:

If there is no shared named entity whose starting position is the first position of the original sentence information in the shared named entity, it is determined that the original sentence information does not only include the shared named entity.

In this embodiment, when the human-machine dialogue server detects that there is no shared named entity whose starting position is the first position of the original sentence information in the shared named entity contained in the original sentence information, it indicates that the first character in the original sentence information is not Included in the shared named entity means that the original sentence information also contains other information besides the shared named entity. At this time, the man-machine dialogue server determines that the original sentence information does not only include the shared named entity.

S23: If the analysis result indicates that the original sentence information only includes a shared named entity, detect whether the target dialogue round corresponding to the original sentence information is the first round of dialogue.

Generally, when the original sentence information in the first round of human-machine dialogue only contains shared named entities, because there is no more reference information, it is difficult to accurately identify the user’s true intention category. At this time, more rounds of people are needed. Machine dialogue to obtain more reference information and further identify the real intention of the user. When the original sentence information in the non-first round of human-machine dialogue only contains shared named entities, it is usually possible to refer to the information obtained in other rounds of human-machine dialogue to identify the user’s true intention category, so there is no need to do more Rounds of man-machine dialogue. Based on this, in specific applications, after each human-machine dialogue terminal converts the voice data from the user during each round of human-machine dialogue into corresponding original sentence information, it also records the corresponding original sentence information in each round of human-machine dialogue. Dialogue rounds, among them, the dialogue round includes the first round of dialogue and non-first round of dialogue, that is, all other rounds of dialogue except the first round of dialogue are non-first round of dialogues.

In this implementation, when the analysis result output by the shared named entity analysis engine indicates that the original sentence information only contains the shared named entity, the man-machine dialogue server further detects whether the target dialogue round corresponding to the original sentence information is the first round of dialogue. If the human-machine dialogue server detects that the target dialogue round corresponding to the original sentence information is the first round of dialogue, S24 is performed.

S24: If the target dialogue round is the first round of dialogue, output the intent category corresponding to the shared named entity category to which the shared named entity belongs, and determine the target intent category selected by the user in the intent category.

In this embodiment, when the man-machine dialogue server detects that the original sentence information contains only shared named entities, and the target dialogue round corresponding to the original sentence information is the first round of dialogue, it can be based on each shared named entity contained in the original sentence information The category of shared named entity to which it belongs, the intent category corresponding to the category of shared named entity to which each shared named entity belongs is obtained from the list of shared named entity categories. After the human-machine dialogue server obtains the intent categories corresponding to the shared named entity category to which each shared named entity contained in the original sentence information belongs, it outputs these intent categories so that the user can select the target they want to express from these intent categories Intent category.

Specifically, the man-machine dialogue server may send the intent category corresponding to the shared named entity category to which each shared named entity contained in the original sentence information belongs to the man-machine dialogue terminal, and the man-machine dialogue terminal may generate and output corresponding intent categories based on these intent categories. To ask the user for the target intent category selected in these intent categories. The man-machine dialogue terminal sends the target intent category selected by the user in these intent categories to the man-machine dialogue server, and the man-machine dialogue server obtains the user The target intent category selected among these intent categories.

In an embodiment of the present application, after the human-machine dialogue server determines the target intent category, it can further obtain the slot information corresponding to the original sentence information, and then according to the target intent category, the original sentence information, and the slot information corresponding to the original sentence information , To determine clear user instructions. Among them, the slot information refers to the necessary information type to which the shared named entity contained in the original sentence information belongs under the target intention category. Exemplarily, suppose that the target intention category is "hailing a taxi", and the intention category "hailing a taxi" usually needs to include two types of necessary information: "departure" and "destination". Assume that the shared named entity contained in the original sentence information Is "Beijing Botanical Garden", and the necessary information type of "Beijing Botanical Garden" under the target intention category of "Taxi" is the destination, then the slot information corresponding to the original sentence information is the destination. Based on this, the man-machine Based on the target intent category "Taxi", the original sentence information "Beijing Botanical Garden", and the slot information "Destination" corresponding to the original sentence information, the dialogue server determines a clear user instruction that may be "Take a taxi to Beijing Botanical Garden".

In a possible implementation of this embodiment, the human-machine dialogue terminal in the human-machine dialogue system can obtain the necessary information type of the shared named entity contained in the original sentence information under the target intention category by asking the user. Then the slot information corresponding to the original sentence information is obtained, and the man-machine dialogue terminal can send the slot information corresponding to the original sentence information to the man-machine dialogue server.

In another possible implementation of this embodiment, when the shared named entity category to which the shared named entity contained in the original sentence information belongs is "place name", the man-machine dialogue terminal in the man-machine dialogue system can collect the original sentence When the voice data corresponds to the information, the geographic location information of its current location is obtained, and the geographic location information of its current location is sent to the man-machine dialogue server. The man-machine dialogue server can be based on the current location of the man-machine dialogue terminal The geographic location information and the geographic location information corresponding to the shared named entity contained in the original sentence information are used to determine the slot information corresponding to the original sentence information. Specifically, when the geographic location information of the current location of the human-machine dialogue terminal matches the geographic location information corresponding to the shared named entity contained in the original sentence information, the slot information corresponding to the original sentence information is determined as the starting place; When the geographic location information of the current location of the machine dialogue terminal does not match the geographic location information corresponding to the shared named entity contained in the original sentence information, the slot information corresponding to the original sentence information is determined as the destination. It should be noted that the geographic location information of the current location of the human-machine dialogue terminal matches the geographic location information corresponding to the shared named entity contained in the original sentence information. Specifically, the current geographic location of the human-machine dialogue terminal matches the original sentence. The location deviation between the geographic locations corresponding to the shared named entities contained in the information is within a preset range; the geographic location information of the current location of the human-machine dialogue terminal is different from the geographic location information corresponding to the shared named entities contained in the original sentence information. The matching specifically refers to that the position deviation between the current geographic location of the human-machine dialogue terminal and the geographic location corresponding to the shared named entity included in the original sentence information is not within a preset range.

It can be seen from the above that the intention recognition method provided by the embodiments of the present application does not directly input the original sentence information of the user into the traditional intention recognition model to determine the intention category expressed by the user after obtaining the original sentence information of the user. , But input the original sentence information into the preset shared named entity analysis engine, through the shared named entity analysis engine to analyze whether the original sentence information contains only shared named entities, and the original sentence information contains only shared named entities , And when the target dialogue round corresponding to the original sentence information is the first round of dialogue, by outputting the intent category corresponding to the shared named entity category to which the shared named entity belongs, the user can select the expressed target intention from the intent category Category, since the target intention category is obtained through further confirmation by the user, it can reduce the error rate of intention recognition and improve the accuracy of intention recognition.

Please refer to FIG. 6, which is a schematic flowchart of an intention recognition method according to another embodiment of the present application. As shown in FIG. 6, with respect to the embodiments corresponding to FIGS. 3 to 5, an intention recognition method provided in this embodiment may further include S25 to S26 after S23, which is described in detail as follows:

S25: If the target dialogue round is not the first round of dialogue, obtain historical original sentence information of the user in the historical dialogue round before the target dialogue round.

S26: Determine the target intention category corresponding to the original sentence information according to the historical original sentence information.

In this embodiment, when the target dialogue round corresponding to the original sentence information is not the first round of dialogue, because the user’s historical original sentence information in the historical dialogue round before the target dialogue round may contain information that can express the user’s intentions The necessary information for, for example, includes the target intention category expressed by the original sentence information. Therefore, the man-machine dialogue server detects that the original sentence information contains only the shared named entity, and the target dialogue round corresponding to the original sentence information is not the first round of dialogue At the time, the historical original sentence information of the user in the historical dialogue round before the target dialogue round is obtained, and the target intention category expressed by the original sentence information is determined based on the historical original sentence information.

In an embodiment of the present application, after the human-machine dialogue server determines the target intent category, it can further obtain the slot information corresponding to the original sentence information, and then according to the target intent category, the original sentence information, and the slot information corresponding to the original sentence information , To determine clear user instructions. It should be noted that, in this embodiment, the man-machine dialogue server determines the specific way of definite user instructions according to the target intent category, original sentence information, and slot information corresponding to the original sentence information. You can refer to the relevant description in S24 here. No longer.

For example, suppose that the original sentence information in the first round of human-machine dialogue is "I want to take a taxi", and the original sentence information in the second round of human-computer dialogue is "Beijing Botanical Garden". The original sentence information in only contains the shared named entity, and the target intention category expressed by the original sentence information "Beijing Botanical Garden" can be determined as the taxi intention based on the original sentence information "I want to take a taxi" in the first round of dialogue. Further, assuming that the slot information of the original sentence information "Beijing Botanical Garden" obtained by asking the user is "destination", it can be further determined that the clear user instruction is "Take a taxi to Beijing Botanical Garden".

As can be seen from the above, in the intention recognition method provided by this embodiment, when the original sentence information only contains the shared named entity, but the target dialogue round corresponding to the original sentence information is not the first round of dialogue, because the target dialogue round is before The historical original sentence information in the historical dialogue round may contain necessary information that can express the user’s intentions, such as the target intention category expressed by the original sentence information. Therefore, directly pass the historical dialogue round before the target round The historical original sentence information determines the target intention category expressed by the original sentence information, without the need to determine the user's target intention category through the user intention recognition model, thereby improving the efficiency of user intention recognition.

In another embodiment of the present application, if the human-machine dialogue server detects that the original sentence information does not only include the shared named entity, it can perform the following steps:

If the original sentence information does not only include a shared named entity, the original sentence information is input into a preset intention recognition model to obtain the target intention category expressed by the original sentence information.

In this embodiment, when the original sentence information does not only contain the shared named entity, that is, when the original sentence information contains other information in addition to the shared named entity, the man-machine dialogue server can directly input the original sentence information into the preset In the intention recognition model, the target intention category expressed by the original sentence information is obtained. It should be noted that the user intent recognition model in this embodiment may be an intent recognition model based on neural networks, or an intent recognition model based on statistics, or may also be an intent recognition model of other types, which may be based on actual conditions. Demand settings. When the user intention recognition model receives the feature vector corresponding to the original sentence information at the input terminal, it can output the target intention category expressed by the original sentence information.

In an embodiment of the present application, after determining the target intent category expressed by the original sentence information, the human-machine dialogue server can obtain the slot information corresponding to the target intent category based on the necessary information that the target intent category needs to include, and based on The target intent category and the slot information corresponding to the target intent category are given clear user instructions.

It can be seen from the above that the intent recognition method provided by this embodiment uses the user intent recognition model to determine the target intent class expressed by the original sentence information when the original sentence information does not only include a shared named entity, thereby improving The accuracy of user intent recognition.

It can be understood that the size of the sequence number of each step in the above embodiment does not mean the order of execution. The execution sequence of each process should be determined by its function and internal logic, and should not constitute any implementation process of the embodiments of this application. limited.

Corresponding to the intention recognition method described in the foregoing embodiment, FIG. 7 shows a structural block diagram of a server provided by an embodiment of the present application. The server may specifically be a human-machine dialogue server in a human-machine dialogue system. The server includes Each unit is used to execute each step in the foregoing embodiment. For details, please refer to the relevant description in the foregoing embodiment. For ease of description, only the parts related to the embodiment of the present application are shown. Please refer to FIG. 7, the server 70 includes a first acquiring unit 71, a second acquiring unit 72, a first detecting unit 73, and a first determining unit 74. among them:

The first obtaining unit 71 is used to obtain the user's original sentence information.

The second obtaining unit 72 is configured to input the original sentence information into a preset shared named entity analysis engine to obtain an analysis result output by the shared named entity analysis engine.

The first detection unit 73 is configured to detect whether the target dialogue round corresponding to the original sentence information is the first round of dialogue if the analysis result indicates that the original sentence information only includes a shared named entity.

The first determining unit 74 is configured to, if the target dialogue round is the first round of dialogue, output an intent category corresponding to the shared named entity category to which the shared named entity belongs, and determine that the user selects among the intent categories The target intent category.

In a possible implementation of this embodiment, the second acquisition unit 72 specifically includes a named entity recognition unit, a shared named entity recognition unit, a location determination unit, and an analysis unit. among them:

The named entity recognition unit is used to recognize the named entity contained in the original sentence information.

The shared named entity identification unit is used to identify the shared named entity in the named entity according to a preset list of shared named entity categories, and determine the shared named entity category to which the shared named entity belongs

The position determining unit is used to determine the start position and the end position of the shared named entity in the original sentence information.

The analysis unit is used to analyze whether the original sentence information contains only the shared named entity according to the start position and the end position of each of the shared named entities, and obtain the analysis result.

In a possible implementation manner of this embodiment, the analysis unit specifically includes: a second determination unit and a first determination unit. among them:

The second determining unit is configured to determine the shared named entity whose starting position is the first position of the original sentence information as a candidate shared named entity.

The first determining unit is configured to determine that the original sentence information only includes the shared named entity if the end position of one of the candidate shared named entities is the end position of the original sentence information.

In a possible implementation manner of this embodiment, the analysis unit specifically further includes: a third determination unit and a second determination unit. among them:

The third determining unit is configured to, if the end positions of all the candidate shared named entities are not the end positions of the original sentence information, perform the loop execution to set the start position to the end position of any one of the candidate shared named entities. The step of determining that the shared named entity at the location is a new candidate shared named entity, and detecting whether the end position of the new candidate shared named entity is the end position of the original sentence information.

The second determining unit is used to determine that the end position of all the new candidate shared named entities is not the end position of the original sentence information after traversing all the shared named entities. Contains only shared named entities.

In another possible implementation manner of this embodiment, the analysis unit specifically includes: a first definition unit, a first update unit, a first detection unit, a second update unit, a third determination unit, and a fourth determination unit. among them:

The first definition unit is used to define a flag bit array with the same length as the length of the original sentence information, and set the value of each flag bit in the flag bit array to a first preset value.

The first update unit is configured to determine the shared named entity whose starting position is the first position of the original sentence information as the first target shared named entity, and mark the end position of the first target shared named entity corresponding to the mark The value of the bit is updated to the second preset value.

The first detection unit is configured to determine the shared named entity whose starting position is not the first position of the original sentence information as the second target shared named entity, and detect the start position of each of the second target shared named entities Whether the value of the flag bit corresponding to the previous position is the second preset value.

The second update unit is configured to, if the value of the flag bit corresponding to the previous position of the start position of the second target shared named entity is the second preset value, set the end of the second target shared named entity The value of the flag bit corresponding to the position is updated to the second preset value.

The third determining unit is configured to determine the original sentence information if the value of the flag bit corresponding to the end position of the original sentence information is the second preset value after traversing all the shared named entities Contains only shared named entities.

The fourth determining unit is configured to determine the original sentence information if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the first preset value after traversing all the shared named entities Does not only contain shared named entities.

In a possible implementation manner of this embodiment, the analysis unit further includes a fifth determination unit.

The fifth determining unit is used for determining that the original sentence information does not only include the shared named entity if there is no shared named entity whose starting position is the first position of the original sentence information in the shared named entity.

In another embodiment of the present application, the server 70 further includes: a third acquiring unit and a fourth determining unit. among them:

The third obtaining unit is configured to obtain historical original sentence information of the user in the historical dialogue round before the target dialogue if the target dialogue round is not the first round of dialogue.

The fourth determining unit is configured to determine the target intention category corresponding to the original sentence information according to the historical original sentence information.

It can be seen from the above that, after obtaining the original sentence information of the user, the server provided by the embodiment of the present application does not directly input the original sentence information into the traditional intention recognition model to determine the intention category expressed by the user. The original sentence information is input into the preset shared named entity analysis engine, and the shared named entity analysis engine is used to analyze whether the original sentence information contains only shared named entities, and the original sentence information contains only shared named entities, and When the target dialogue round corresponding to the original sentence information is the first round of dialogue, by outputting the intent category corresponding to the shared named entity category to which the shared named entity belongs, so that the user can select the expressed target intent category from the intent categories, Since the target intention category is obtained through further confirmation by the user, the error rate of intention recognition can be reduced, and the accuracy of intention recognition can be improved.

Please refer to FIG. 8. FIG. 8 is a schematic structural diagram of a server provided by another embodiment of the present application. As shown in FIG. 8, the server 800 of this embodiment includes: at least one processor 80 (only one is shown in FIG. 8), a processor, a memory 81, and a memory 81 stored in the memory 81 and available in the at least one processor. A computer program 82 running on the processor 80, when the processor 80 executes the computer program 82, implements the steps in any of the above-mentioned intention recognition method embodiments.

The server 800 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The server may include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art can understand that FIG. 8 is only an example of the server 800, and does not constitute a limitation on the server 800. It may include more or less components than shown, or a combination of certain components, or different components, such as It can also include input and output devices, network access devices, and so on.

The so-called processor 80 may be a central processing unit (Central Processing Unit, CPU), and the processor 80 may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), and application specific integrated circuits (Application Specific Integrated Circuits). , ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.

In some embodiments, the memory 81 may be an internal storage unit of the server 800, such as a hard disk or a memory of the server 800. In other embodiments, the memory 81 may also be an external storage device of the server 800, for example, a plug-in hard disk equipped on the server 800, a smart memory card (Smart Media Card, SMC), and a secure digital (Secure Digital). Digital, SD) card, flash card, etc. Further, the storage 81 may also include both an internal storage unit of the server 800 and an external storage device. The memory 81 is used to store an operating system, an application program, a boot loader (Boot Loader), data, and other programs, such as the program code of the computer program. The memory 81 can also be used to temporarily store data that has been output or will be output.

It should be noted that the information interaction and execution process between the above-mentioned devices/units are based on the same concept as the method embodiment of this application, and its specific functions and technical effects can be found in the method embodiment section. I won't repeat it here.

Those skilled in the art can clearly understand that for the convenience and conciseness of description, only the division of the above functional units and modules is used as an example. In practical applications, the above functions can be allocated to different functional units and modules as required. Module completion, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist alone physically, or two or more units can be integrated into one unit. The above-mentioned integrated units can be hardware-based Formal realization can also be realized in the form of a software functional unit. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the present application. For the specific working process of the units and modules in the foregoing system, reference may be made to the corresponding process in the foregoing method embodiment, which will not be repeated here.

The embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned intention recognition method can be realized.

The embodiment of the present application provides a computer program product. When the computer program product runs on a mobile terminal, the steps in the above-mentioned intention recognition method can be realized when the mobile terminal is executed.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the implementation of all or part of the processes in the above-mentioned embodiment methods in this application can be accomplished by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. The computer program can be stored in a computer-readable storage medium. When executed by the processor, the steps of the foregoing method embodiments can be implemented. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms. The computer-readable medium may include at least: any entity or device capable of carrying computer program code to the server, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signals, telecommunications signals, and software distribution media. Such as U disk, mobile hard disk, floppy disk or CD-ROM, etc. In some jurisdictions, according to legislation and patent practices, computer-readable media cannot be electrical carrier signals and telecommunication signals.

In the above-mentioned embodiments, the description of each embodiment has its own focus. For parts that are not described in detail or recorded in an embodiment, reference may be made to related descriptions of other embodiments.

A person of ordinary skill in the art may realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.

In the embodiments provided in this application, it should be understood that the disclosed apparatus/network equipment and method may be implemented in other ways. For example, the device/network device embodiments described above are only illustrative. For example, the division of the modules or units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units. Or components can be combined or integrated into another system, or some features can be omitted or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be separately on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

Finally, it should be noted that the above are only specific implementations of this application, but the scope of protection of this application is not limited to this. Any changes or substitutions within the technical scope disclosed in this application shall be covered by this application. Within the scope of protection applied for. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims (10)

1. An intention recognition method, which is characterized in that it includes:

   Get the user's original sentence information;

   Inputting the original sentence information into a preset shared named entity analysis engine to obtain an analysis result output by the shared named entity analysis engine;

   If the analysis result indicates that the original sentence information only contains a shared named entity, detecting whether the target dialogue round corresponding to the original sentence information is the first round of dialogue;

   If the target dialogue round is the first round of dialogue, output the intent category corresponding to the shared named entity category to which the shared named entity belongs, and determine the target intent category selected by the user in the intent category.
2. 3. The intention recognition method of claim 1, wherein the inputting the original sentence information into a preset shared named entity analysis engine to obtain the analysis result output by the shared named entity analysis engine comprises:

   Identifying named entities included in the original sentence information;

   Identify the shared named entity among the named entities according to a preset list of shared named entity categories, and determine the shared named entity category to which the shared named entity belongs;

   Determining the start position and the end position of the shared named entity in the original sentence information;

   According to the starting position and the ending position of each of the shared named entities, it is analyzed whether the original sentence information only includes the shared named entity, and the analysis result is obtained.
3. 3. The intention recognition method according to claim 2, wherein said analyzing whether said original sentence information contains only shared named entities according to said starting position and said ending position of each of said shared named entities, Obtaining the analysis result includes:

   Determining the shared named entity whose starting position is the first position of the original sentence information as a candidate shared named entity;

   If the end position of one of the candidate shared named entities is the end position of the original sentence information, it is determined that only the shared named entity is included in the original sentence information.
4. 5. The intention recognition method of claim 3, wherein after determining the shared named entity whose starting position is the first position of the original sentence information as a candidate shared named entity, the method further comprises:

   If the ending positions of all the candidate shared named entities are not the end positions of the original sentence information, the shared naming with the starting position being the position after the ending position of any one of the candidate shared named entities is executed in a loop The step of determining whether the entity is a new candidate shared named entity, and detecting whether the end position of the new candidate shared named entity is the end position of the original sentence information;

   After traversing all the shared named entities, if the end positions of all the new candidate shared named entities are not the end positions of the original sentence information, it is determined that the original sentence information does not only include the shared named entity.
5. 3. The intention recognition method according to claim 2, wherein said analyzing whether said original sentence information contains only shared named entities according to said starting position and said ending position of each of said shared named entities, Obtaining the analysis result includes:

   Defining a flag bit array with the same length as the length of the original sentence information, and setting the value of each flag bit in the flag bit array to a first preset value;

   The shared named entity whose starting position is the first position of the original sentence information is determined as the first target shared named entity, and the value of the flag corresponding to the end position of the first target shared named entity is updated to the first target shared named entity. Two preset values;

   Determine the shared named entity whose starting position is not the first position of the original sentence information as the second target shared named entity, and detect the mark corresponding to the previous position of the starting position of each second target shared named entity Whether the value of the bit is the second preset value;

   If the value of the flag bit corresponding to the previous position of the start position of the second target shared named entity is the second preset value, set the value of the flag bit corresponding to the end position of the second target shared named entity The value is updated to the second preset value;

   After traversing all the shared named entities, if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the second preset value, it is determined that the original sentence information only contains the shared named entity ；

   After traversing all the shared named entities, if it is detected that the value of the flag bit corresponding to the end position of the original sentence information is the first preset value, it is determined that the original sentence information does not only include the shared name entity.
6. 3. The intention recognition method according to claim 2, wherein said analyzing whether said original sentence information contains only shared named entities according to said starting position and said ending position of each of said shared named entities, Obtaining the analysis result also includes:

   If there is no shared named entity whose starting position is the first position of the original sentence information in the shared named entity, it is determined that the original sentence information does not only include the shared named entity.
7. 7. The intention recognition method according to any one of claims 1 to 6, characterized in that, after detecting whether the target dialogue round corresponding to the original sentence information is the first round of dialogue, the method further comprises:

   If the target dialogue round is not the first round of dialogue, acquiring historical original sentence information of the user in the historical dialogue round before the target dialogue round;

   According to the historical original sentence information, the target intention category corresponding to the original sentence information is determined.
8. A server, characterized in that it comprises:

   The first obtaining unit is used to obtain the original sentence information of the user;

   The second obtaining unit is configured to input the original sentence information into a preset shared named entity analysis engine to obtain the analysis result output by the shared named entity analysis engine;

   The first detecting unit is configured to detect whether the target dialogue round corresponding to the original sentence information is the first round of dialogue if the analysis result indicates that the original sentence information contains only shared named entities;

   The first determining unit is configured to, if the target dialogue round is the first round of dialogue, output an intent category corresponding to the shared named entity category to which the shared named entity belongs, and determine that the user selects among the intent categories The target intent category.
9. A server comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program as claimed in claims 1 to 7 Any one of the intention recognition methods.
10. A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program implements the intention recognition method according to any one of claims 1 to 7 when the computer program is executed by a processor.

Images