WO2021047169A1

WO2021047169A1 - Information query method and apparatus, storage medium, and smart terminal

Info

Publication number: WO2021047169A1
Application number: PCT/CN2020/083561
Authority: WO
Inventors: 简仁贤; 马永宁
Original assignee: 竹间智能科技（上海）有限公司
Priority date: 2019-09-12
Filing date: 2020-04-07
Publication date: 2021-03-18
Also published as: CN110765342A

Abstract

An information query method and apparatus, a storage medium, and a smart terminal. The information query method comprises: acquiring a natural language query statement inputted by a user (S101); determining, from a knowledge base, an entity to be queried in the natural language query statement (S102); recognizing a query intention of the natural language query statement (S103); and determining an answer corresponding to the natural language query statement according to a combination of a query entity name and the query intention and a predetermined mapping relationship (S104). Said method can allow a user to query information by using a natural language, improving the accuracy and convenience of query.

Description

Information query method and device, storage medium and intelligent terminal

Technical field

The present invention relates to the technical field of natural language processing, in particular to an information query method and device, a storage medium, and an intelligent terminal.

Background technique

For current customer-oriented virtual products, such as fund products and insurance products, there are many types of products. For example, for the same type of funds, there may be a large income gap due to different fund managers or fund companies. In the face of so many funds, it is difficult for ordinary users to choose and compare.

In the prior art, the way to obtain the above-mentioned product-related information is usually to query through a search engine, for example, enter "China CSI 300 Index Fund" in the search engine to query.

However, the information queried by the query method in the prior art is generally general and complicated, and the user usually queries with a clear intention. For example, the user may only want to see which stocks the fund holds, or see What is the historical performance of the fund manager of a certain fund? It is difficult for the search engine in the prior art to find it in one step, and it is necessary to filter the information after the first query to obtain effective information.

Summary of the invention

The technical problem solved by the present invention is how to improve the accuracy and convenience of query.

In order to solve the above technical problem, an embodiment of the present invention provides an information query method. The information query method includes: obtaining a natural language query sentence input by a user; determining the entity to be queried in the natural language query sentence from a knowledge base; The entity to be queried includes a query entity name; identifying a query intention of the natural language query sentence; and determining an answer corresponding to the natural language query sentence according to a combination of the query entity name and the query intention and a preset mapping relationship.

Optionally, the determining the answer corresponding to the natural language query sentence according to the combination of the query entity name and the query intention and a preset mapping relationship includes: combining the query entity name with a plurality of preset mapping relationships If the entity name in the preset mapping relationship matches the query entity name, match the intent in the matching preset mapping relationship with the query intent; if the matching preset Assuming that the intent in the mapping relationship matches the query intent, the answer in the matched preset mapping relationship is used as the answer corresponding to the natural language query sentence.

Optionally, the using the answer in the matched preset mapping relationship as the answer corresponding to the natural language query sentence includes: directly corresponding the answer to the combination of the entity name and the intent in the matched preset mapping relationship As the answer corresponding to the natural language query sentence; or, determine the query instruction corresponding to the combination of entity name and intent in the matched preset mapping relationship, and use the answer obtained by executing the query instruction as the natural language query The answer to the sentence.

Optionally, the determining the entity to be queried in the natural language query sentence from the knowledge base includes: sorting the list of entities in the knowledge base according to the natural language query sentence input by the user, and sorting The preceding entity is used as the entity to be queried.

Optionally, the following algorithms are used to sort the list of entities in the knowledge base: learning-to-rank model, or syntax analysis.

Optionally, the identifying the query intention of the natural language query sentence includes: sorting the limited intent set corresponding to the entity to be queried according to the entity to be queried and the natural language query sentence.

Optionally, the following algorithm is used to sort the limited intent set corresponding to the entity to be queried: a learning-to-rank model, or a way of syntactic analysis.

Optionally, before determining the entity to be queried in the natural language query sentence from the knowledge base, the method further includes: comparing the pinyin of each word in the natural language query sentence with each preset entity name in the preset entity name list The pinyin of the preset entity name is matched to obtain a matching result. The preset entity name list includes a plurality of preset entity names and their pinyin; if the matching result indicates that the pinyin of the preset entity name exists and the natural language query sentence If the pinyin of the word matches, the word is updated to the matching preset entity name.

Optionally, before determining the entity to be queried in the natural language query sentence from the knowledge base, the method further includes: performing a preprocessing operation on the natural language query sentence, and the preprocessing operation is selected from filtering sensitive words and fonts. Conversion.

Optionally, the entity name is selected from the names of fund products, fund managers, and fund companies, or the names of insurance products, insurance managers, and insurance companies, or the names of wealth management products, wealth management managers, and wealth management companies.

In order to solve the above technical problem, the embodiment of the present invention also discloses an information query device. The information query device includes: a natural language query sentence acquisition module for acquiring a natural language query sentence input by a user; an entity name recognition module for receiving The entity to be queried in the natural language query sentence is determined in the knowledge base, and the entity to be queried includes the name of the query entity; an intention recognition module is used to identify the query intention of the natural language query sentence; an answer determination module is used according to The combination of the query entity name and the query intention and the preset mapping relationship determine the answer corresponding to the natural language query sentence.

The embodiment of the present invention also discloses a storage medium on which computer instructions are stored, and the steps of the information query method are executed when the computer instructions are run.

The embodiment of the present invention also discloses an intelligent terminal, including a memory and a processor, the memory stores computer instructions that can run on the processor, and the processor executes the information when the computer instructions are executed. The steps of the query method.

Compared with the prior art, the technical solutions of the embodiments of the present invention have the following beneficial effects:

In the technical scheme of the present invention, for the natural language query sentence input by the user, the query entity name in the natural language query sentence and the query intention of the natural language query sentence can be identified; the entity name and the query intention can be determined through the preset mapping relationship established in advance. The combination of intents and the corresponding relationship between the answers, so that the query entity name in the natural language query sentence, the query intention of the natural language query sentence, and the preset mapping relationship can be used to determine the answer to the natural language query sentence. The technical solution of the present invention uses the combination of entity name and intent to determine the answer, which can ensure the accuracy and pertinence of answer determination, avoids the way in the prior art that requires a second query by the user to determine the answer, and improves the information query Convenience and user experience.

Further, when determining the answer to the natural language query sentence, it is necessary to determine that the entity name in the preset mapping relationship matches the query entity name, and the intent in the preset mapping relationship matches the query intent, before determining the preset The answer in the mapping relationship is the answer to the natural language query sentence, and the accuracy of the matched answer is ensured by matching the entity name and intent.

Further, the pinyin of the entity name is fuzzy matched with the pinyin of each preset entity name in the preset entity name list to obtain a matching result. The preset entity name list includes a plurality of preset entity names and their pinyin ; If the matching result indicates a match, the entity name is updated to the matching preset entity name. In the technical scheme of the present invention, in order to avoid query errors caused by typos in the natural language query sentence input by the user, the entity name can be updated by means of pinyin matching, thereby further ensuring the accuracy of the finally matched answer.

Description of the drawings

FIG. 1 is a flowchart of an information query method according to an embodiment of the present invention;

FIG. 2 is a flowchart of a specific implementation of step S104 shown in FIG. 1;

FIG. 3 is a partial flowchart of an information query method according to an embodiment of the present invention;

Fig. 4 is a schematic structural diagram of an information query device according to an embodiment of the present invention.

detailed description

As mentioned in the background art, the information queried by the query methods in the prior art is generally general and complicated, and users usually query with clear intentions. For example, users may only want to see the stocks held by this fund. Which ones? Or look at the historical performance of the fund manager of a certain fund. It is difficult for the search engine in the prior art to find it in one step, and it is necessary to perform a secondary screening and filtering of the information after the first query to obtain effective information.

The "natural language query sentence" referred to in the embodiment of the present invention refers to a sentence input by a user for query, which may specifically be text or voice data.

The "query entity name" mentioned in the embodiment of the present invention refers to the name of the entity appearing in the natural language query sentence, such as the name of the fund, the name of the fund manager, and the name of the fund company.

The "query intention" in the embodiment of the present invention refers to the intention expressed by a natural language query sentence, such as querying positions, querying company profiles, and so on.

The "preset mapping relationship" referred to in the embodiment of the present invention refers to a pre-established correspondence relationship between a combination of entity name and intent and an answer. Specifically, the preset mapping relationship may include a combination of multiple entity names and intents and their correspondences. s answer.

In order to make the above objectives, features and advantages of the present invention more obvious and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Fig. 1 is a flowchart of an information query method according to an embodiment of the present invention.

The information query method shown in FIG. 1 can be executed by any smart terminal device capable of interacting with the user, and the smart terminal device may specifically be any device capable of interacting with the user, such as a computer or an intelligent robot.

The information query method may include the following steps:

Step S101: Obtain a natural language query sentence input by the user;

Step S102: Determine the entity to be queried in the natural language query sentence from the knowledge base, where the entity to be queried includes the name of the query entity;

Step S103: Identify the query intention of the natural language query sentence;

Step S104: Determine the answer corresponding to the natural language query sentence according to the combination of the query entity name and the query intention and a preset mapping relationship.

It should be pointed out that the sequence number of each step in this embodiment does not represent a limitation on the execution order of each step. For example, step S102 and step 103 may be executed simultaneously, or step S102 may be executed earlier than step S103, or step S103 may be executed earlier than step S102, which is not limited in the embodiment of the present invention.

In this embodiment, the natural language query sentence input by the user may be natural language. More specifically, the natural language query sentence may be in text, voice, and other formats. For example, in a specific application scenario, when a user inputs voice data, the voice data may be converted into text first, and subsequent entity name recognition and intention recognition processes are performed on the basis of the text.

In the specific implementation of step S102, the entity list of the knowledge base can be sorted for the natural language query sentences input by the user, and the query entity name in the natural language query sentence can be determined based on the top 1 result of the sorting. Regarding the specific ranking algorithm, a learning-to-rank algorithm based on deep learning can be used, or any other implementable entity name recognition algorithm, such as a syntax analysis algorithm, can be used, which is not limited in the embodiment of the present invention.

Regarding the specific implementation of the learning-to-rank algorithm (or learning-to-rank model) and the syntax analysis algorithm, reference may be made to the prior art, and the details of the embodiment of the present invention will not be repeated here.

Taking fund products as an example, entity names can include the following three categories: fund products, fund managers, and fund companies. The user inputs the natural language query sentence "Is Rainbow Fund Company good", and the entity name "Rainbow Fund Company" can be obtained through step S102.

In the specific implementation of step S103, the query intention of obtaining the natural language query sentence can be identified. Specifically, a pre-trained model can be used to sort the intents corresponding to the entity according to the determined entity name, and the top 1 result of the sorting is used as the intent result. Regarding the specific ranking algorithm, a learning-to-rank algorithm based on deep learning may be used, or any other implementable algorithm, such as a syntax analysis algorithm, may be used, which is not limited in the embodiment of the present invention.

Specifically, a learning-to-rank model is trained using a trained neural network to perform intention recognition on the natural language query sentence.

For example, if the user inputs the natural language query sentence "Is Rainbow Fund Company good", the query intention can be obtained as "company profile" through step S103.

So far, the query entity name and query intention of the natural language query sentence can be determined through step S102 and step S103, and the combination of the above two parameters can clearly indicate the content that the user wants.

In this embodiment, before acquiring the natural language query sentence of the user, a preset mapping relationship may be established in advance to establish the correspondence relationship between the combination of the entity name and the intention and the answer.

Furthermore, in step S104, the query entity name in the natural language query sentence, the query intention of the natural language query sentence, and the preset mapping relationship may be used to determine the answer to the natural language query sentence. The preset mapping relationship includes a combination of multiple entity names and intentions and their corresponding answers.

In a specific implementation, the preset mapping relationship may also be a mapping relationship between a combination of entity name and intent and a query instruction, and each query instruction can determine a unique answer through a query operation. For example, the query instruction may be a Structured Query Language (SQL) statement, through which the answer can be queried in the database, and the answer is fed back to the user.

Specifically, the preset mapping relationship may be stored in a database, and when needed, the preset mapping relationship may be called from the database.

The embodiment of the present invention uses the combination of the entity name and the intention to determine the answer, which can ensure the accuracy and pertinence of the answer determination, avoids the way in the prior art that requires the user to make a second query to determine the answer, and improves the information query. Convenience and user experience.

In a non-limiting embodiment of the present invention, the entity name is selected from the names of fund products, fund managers, and fund companies, or the names of insurance products, insurance managers, and insurance companies, or selected from wealth management products, wealth management managers And the name of the wealth management company.

In other words, the information query method of the embodiment of the present invention can query fund-related information, insurance-related information, and financial-related information to meet different needs of users in different application scenarios.

In a specific embodiment of the present invention, the information query method shown in FIG. 1 may further include the following step: feeding back the answer to the user.

Specifically, the answer may be displayed to the user in a preset format, for example, it may be displayed in text; or, the answer may be broadcast in voice, and the answer may also be displayed in text and The combination of answers is displayed.

In a specific embodiment of the present invention, please refer to FIG. 2. Step S104 shown in FIG. 1 may include the following steps:

Step S201: Match the query entity name with entity names in multiple preset mapping relationships;

Step S202: if there is an entity name in the preset mapping relationship that matches the query entity name, match the intent in the matched preset mapping relationship with the query intent;

Step S203: If the intent in the matched preset mapping relationship matches the query intent, use the answer in the matched preset mapping relationship as the answer corresponding to the natural language query sentence.

It should be noted that steps S201 and S202 can also be replaced with the following steps: first match the query intent with the intent in multiple preset mapping relationships; if there is an intent in the preset mapping relationship and the query intent If a match is made, the entity name in the preset mapping relationship is matched with the query entity name, which is not limited in the embodiment of the present invention.

In this embodiment, when determining the answer to the natural language query sentence, it is necessary to determine that the entity name in the preset mapping relationship matches the query entity name, and the intent in the preset mapping relationship matches the query intent, before the determination is made. The answer in the preset mapping relationship is the answer to the natural language query sentence, and the accuracy of the matched answer is ensured by matching the entity name and intent.

Furthermore, if there is only an intent matching the query intent in the same preset mapping relationship, or the entity name matches the query entity name, the answer corresponding to the natural language query sentence cannot be determined. In this case, a preset guiding sentence can be returned to the user to instruct the user to update the natural language query sentence. For example, "Can your question be more specific", "You can ask me some knowledge about the fund" and so on.

More specifically, step S203 shown in FIG. 2 may include the following steps: directly use the answer corresponding to the combination of the entity name and the intent in the matched preset mapping relationship as the answer corresponding to the natural language query sentence; or, determine the answer corresponding to the natural language query sentence; The query instruction corresponding to the combination of the entity name and the intent in the matching preset mapping relationship is described, and the answer obtained by executing the query instruction is used as the answer corresponding to the natural language query sentence.

In other words, the combination of entity name and intent in the preset mapping relationship can directly correspond to the answer. The preset mapping relationship includes the combination of entity name and intent and the answer. In this case, you can directly map the matching preset The answer in the relation is used as the answer corresponding to the natural language query sentence.

The combination of entity name and intent in the preset mapping relationship can indirectly correspond to the answer, that is, the preset mapping relationship includes the combination of entity name and intent and query instructions. The answer can be obtained by executing the query instruction. In this case, The answer obtained by executing the query instruction may be used as the answer corresponding to the natural language query sentence.

It should be noted that the query instruction can be a SQL statement, or any other implementable unstructured query instruction that can perform a query operation, which is not limited in the embodiment of the present invention.

In a preferred embodiment of the present invention, step S102 shown in FIG. 1 may include the following steps: input the natural language query sentence into a pre-trained entity name recognition model to obtain the entity name in the natural language query sentence Wherein, when the entity name recognition model is trained using training data, the training data includes the full name and abbreviation of each entity name.

In practical applications, the full name of the entity name that the user needs to query is relatively long, and the user usually uses the abbreviation of the entity name to query. Taking fund products as an example, the name of the fund is usually very long, such as "China Joy Health "Hybrid", "China Industry Leading Hybrid", "Boshiyueyueying Short-term Wealth Management Bonds", etc., but users generally cannot remember the full name of the fund. The natural language query sentence entered is usually the omitted fund name, such as "Huaxia "Lexiang", "Huaxia Leading Mix", and "Doctor Yueyueying" resulted in no query results.

In order to ensure the comprehensiveness and accuracy of entity name recognition when using the entity name recognition model to recognize the entity name in natural language query sentences, the full name and abbreviation of each entity name can be used to construct training data, and the above training data can be used to identify the entity. The name recognition model is trained, so that the trained entity name recognition model can recognize the full name or abbreviation of each entity name, avoid omitting the entity name, and ensure the accuracy of information query.

In a non-limiting embodiment of the present invention, please refer to FIG. 3. Before step S102 shown in FIG. 1, the following steps may also be included:

Step S301: Match the pinyin of each word in the natural language query sentence with the pinyin of each preset entity name in the preset entity name list to obtain a matching result. The preset entity name list includes a plurality of preset entities Name and its pinyin;

Step S302: If the matching result indicates that the pinyin of the preset entity name matches the pinyin of the word in the natural language query sentence, update the word to the matched preset entity name.

In the specific implementation, since the natural language query sentence is input by the user through the input method, it is inevitable that there will be typos or the user deliberately enters the error. Therefore, it is necessary to correct the typos in the natural language query sentence, especially for the natural language query sentence Correct the typos in the entity name in the following steps to ensure the correctness of the entity name recognition in the subsequent steps.

Specifically, a preset entity name list may be established in advance, and the preset entity name list includes multiple preset entity names and their pinyin. It is also possible to perform pinyin conversion on the natural language query sentence to obtain the pinyin of each word in the natural language query sentence.

In the specific implementation of step S301 and step S302, the pinyin of each word in the natural language query sentence can be matched with the pinyin of each preset entity name in the preset entity name list, for example, fuzzy matching can be performed. If there is a pinyin of the preset entity name that matches the pinyin of the word, update the word to the matched preset entity name.

In the embodiment of the present invention, in order to avoid query errors caused by typos in the natural language query sentence input by the user, the entity name can be updated by means of pinyin matching, thereby further ensuring the accuracy of the finally matched answer.

In a specific application scenario, the user enters the natural language query sentence "Is Tianhong Fund Company good?", in which the pinyin of "Tianhong Fund" is the same as the pinyin of "Tianhong Fund" in the preset entity name list, so "Tianhong Fund" is changed to "Tianhong Fund". "Fund" is updated to "Tianhong Fund"; through the entity name identification step, the query entity name is identified as "Tianhong Fund", and the intent of the natural language query sentence is "Company Profile".

In a non-limiting embodiment of the present invention, before step S102 shown in FIG. 1, the following step may be further included: performing a preprocessing operation on the natural language query sentence, and the preprocessing operation is selected from filtering sensitive words and font conversion.

In specific implementation, before performing entity name recognition and intention recognition, sensitive words in natural language query sentences can be filtered. For example, sensitive words can include abusive words, sensitive names, and sensitive nouns. When a sensitive word is matched in the natural language query sentence, the sensitive word can be filtered directly, and the subsequent steps can be performed. Or you can directly return to the preset sentence without performing subsequent processing.

In specific implementation, before performing entity name recognition and intent recognition, font conversion may be performed on the font of the natural language query sentence, so that the font of each word in the natural language query sentence is consistent. Specifically, when the natural language query sentence is in Chinese, the natural language query sentence is unified into simplified Chinese.

Regarding the setting of specific sensitive words or the setting of specific conversion fonts, adaptive configuration and modification can be made according to the actual application environment, which is not limited in the embodiment of the present invention.

In a specific application scenario, the user enters the natural language query sentence "Huaxia Company Garbage" to identify the sensitive word "garbage". In this case, it can directly return to specific words, or filter the sensitive word to get " Huaxia Company" and continue to implement the subsequent steps.

4, the embodiment of the present invention also discloses an information query device 40. The information query device 40 may include a natural language query sentence acquisition module 401, an entity name recognition module 402, an intention recognition module 403, and an answer determination module 404.

Among them, the natural language query sentence obtaining module 401 is used to obtain the natural language query sentence input by the user; the entity name recognition module 402 is used to sort the entities in the entity list through the natural language query sentence; the intention recognition module 403 is used to According to the entity name determined by the entity name recognition module 402, the intents corresponding to the entities are sorted to obtain the query intent of the natural language query sentence; the answer determination module 404 is used to determine the query entity name and the query intent according to the query entity name and the query intent. The combination of and a preset mapping relationship determines the answer corresponding to the natural language query sentence, and the preset mapping relationship includes a combination of multiple entity names and intentions and their corresponding answers.

In the embodiment of the present invention, for the natural language query sentence input by the user, the query entity name in the natural language query sentence and the query intention of the natural language query sentence can be identified; the entity name and the query intention can be determined through the preset mapping relationship established in advance. The combination of intents and the corresponding relationship between the answers, so that the query entity name in the natural language query sentence, the query intention of the natural language query sentence, and the preset mapping relationship can be used to determine the answer to the natural language query sentence. The embodiment of the present invention uses the combination of the entity name and the intention to determine the answer, which can ensure the accuracy and pertinence of the answer determination, avoids the way in the prior art that requires the user to make a second query to determine the answer, and improves the information query. Convenience and user experience.

For more details about the working principle and working mode of the information query device 40, reference may be made to the related descriptions in FIGS. 1 to 3, which will not be repeated here.

The embodiment of the present invention also discloses a storage medium on which computer instructions are stored, and the computer instructions can execute the steps of the method shown in FIG. 1, FIG. 2 or FIG. 3 when the computer instruction is running. The storage medium may include ROM, RAM, magnetic disk or optical disk, etc. The storage medium may also include non-volatile memory (non-volatile) or non-transitory memory, etc.

The embodiment of the present invention also discloses an intelligent terminal. The intelligent terminal may include a memory and a processor, and computer instructions that can run on the processor are stored in the memory. When the processor runs the computer instructions, the steps of the method shown in FIG. 1, FIG. 2 or FIG. 3 can be executed. The smart terminal includes, but is not limited to, terminal devices such as mobile phones, computers, and tablets.

Although the present invention is disclosed as above, the present invention is not limited to this. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.

Claims

An information query method, characterized in that it comprises:

Obtain the natural language query sentence input by the user;

Determine the entity to be queried in the natural language query sentence from the knowledge base, where the entity to be queried includes the name of the query entity;

Identifying the query intention of the natural language query sentence;

The answer corresponding to the natural language query sentence is determined according to the combination of the query entity name and the query intention and a preset mapping relationship.
The information query method according to claim 1, wherein the determining the answer corresponding to the natural language query sentence according to the combination of the query entity name and the query intention and a preset mapping relationship comprises:

Matching the query entity name with entity names in a plurality of preset mapping relationships;

If there is an entity name in the preset mapping relationship that matches the query entity name, matching the intent in the matched preset mapping relationship with the query intent;

If the intent in the matched preset mapping relationship matches the query intent, the answer in the matched preset mapping relationship is used as the answer corresponding to the natural language query sentence.
The information query method according to claim 2, wherein the using the answer in the matched preset mapping relationship as the answer corresponding to the natural language query sentence comprises:

Directly using the answer corresponding to the combination of the entity name and the intent in the matched preset mapping relationship as the answer corresponding to the natural language query sentence;

Alternatively, the query instruction corresponding to the combination of the entity name and the intent in the matched preset mapping relationship is determined, and the answer obtained by executing the query instruction is used as the answer corresponding to the natural language query sentence.
The information query method according to claim 1, wherein the determining the entity to be queried in the natural language query sentence from the knowledge base comprises:

According to the natural language query sentence input by the user, the entity list in the knowledge base is sorted, and the entity with the highest ranking is used as the entity to be queried.
The information query method according to claim 1, wherein the following algorithm is used to sort the list of entities in the knowledge base: learning-to-rank model, or syntactic analysis.
The information query method according to claim 1, wherein the identifying the query intention of the natural language query sentence comprises:

According to the entity to be queried and the natural language query sentence, the limited intent set corresponding to the entity to be queried is sorted.
The information query method according to claim 6, characterized in that the following algorithm is used to sort the limited intent set corresponding to the entity to be queried: a learning-to-rank model, or a way of syntactic analysis.
The information query method according to claim 1, wherein before determining the entity to be queried in the natural language query sentence from the knowledge base, the method further comprises:

The pinyin of each word in the natural language query sentence is matched with the pinyin of each preset entity name in the preset entity name list to obtain a matching result. The preset entity name list includes multiple preset entity names and their names. pinyin;

If the matching result indicates that the pinyin of the preset entity name matches the pinyin of the word in the natural language query sentence, the word is updated to the matched preset entity name.
The information query method according to claim 1, wherein before determining the entity to be queried in the natural language query sentence from the knowledge base, the method further comprises:

A preprocessing operation is performed on the natural language query sentence, and the preprocessing operation is selected from filtering sensitive words and font conversion.
The information query method according to claim 1, wherein the entity name is selected from the names of fund products, fund managers, and fund companies, or selected from the names of insurance products, insurance managers, and insurance companies, or selected from financial management The name of the product, wealth management manager, and wealth management company.
An information query device, characterized in that it comprises:

The natural language query sentence acquisition module is used to acquire the natural language query sentence input by the user;

The entity name recognition module is used to determine the entity to be queried in the natural language query sentence from the knowledge base, and the entity to be queried includes the name of the query entity;

An intention recognition module for identifying the query intention of the natural language query sentence;

The answer determination module is used to determine the answer corresponding to the natural language query sentence according to the combination of the query entity name and the query intention and a preset mapping relationship.
A storage medium having computer instructions stored thereon, wherein the computer instructions execute the steps of the information query method according to any one of claims 1 to 10 when the computer instructions are run.
An intelligent terminal, comprising a memory and a processor, the memory stores computer instructions that can run on the processor, wherein the processor executes claims 1 to 10 when the computer instructions are executed. Any of the steps of the information query method.