CN118760750A - A method and device for optimizing retrieval of colloquial text content - Google Patents

Abstract

The present invention provides a method and device for optimizing the retrieval of colloquial text content. The method comprises: obtaining a target entity contained in the input colloquial text information; performing an associated retrieval analysis on the target entity based on a preset distributed retrieval database to obtain a label name and its attribute field corresponding to the target entity; the distributed retrieval database pre-stores the corresponding relationship between each entity and each type of label name and its attribute field; performing template construction processing based on the label name and its attribute field to obtain a prompt word template corresponding to the target entity; processing the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the label name and its attribute field. The method for optimizing the retrieval of colloquial text content provided by the present invention can effectively improve the retrieval accuracy and efficiency of the colloquial text content.

Description

A method and device for optimizing retrieval of colloquial text content

Technical Field

The present invention relates to the field of artificial intelligence technology, and more particularly to a method and device for optimizing the retrieval of colloquial text content. In addition, the present invention also relates to an electronic device and a processor-readable storage medium.

Background Art

In recent years, with the rapid development of artificial intelligence technology, various large models driven by artificial intelligence technology have become more and more numerous, such as open platforms such as ChatGPT (Chat Generative Pre-trained Transformer), which are usually able to interact and communicate according to the context. At present, users' expressions are very colloquial. For example, when a user proposes a colloquial text message, the user does not say what a certain entity mentioned in the colloquial text message is. It may be a device, a tool, or a raw material. This situation will cause trouble for the large model, and it is uncertain what the specific entity is, resulting in errors in the final retrieval results. Therefore, how to design a retrieval optimization solution for colloquial text content has become a technical problem that needs to be solved urgently.

Summary of the invention

To this end, the present invention provides a method and device for optimizing the retrieval of colloquial text content, so as to solve the defects that the retrieval scheme of colloquial text content in the prior art is highly limited, resulting in poor pertinence and efficiency of question-answer retrieval results.

In a first aspect, the present invention provides a method for optimizing retrieval of colloquial text content, comprising:

Obtaining a target entity contained in the input spoken text information;

Based on a preset distributed search database, an associated search analysis is performed on the target entity to obtain a label name and an attribute field corresponding to the target entity; wherein the distributed search database pre-stores the corresponding relationship between each entity and each type of label name and its attribute field;

Performing template construction processing based on the tag name and its attribute field to obtain prompt word templates corresponding to the target entities; wherein the prompt word templates include the tag name and its attribute field corresponding to the target entity;

The prompt word template is processed based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and its attribute fields.

Furthermore, the associated search analysis is performed on the target entity based on a preset distributed search database to obtain a tag name and attribute field corresponding to the target entity, including:

Taking the target entity as an index, an associated search analysis is performed on the distributed search database to find a tag name and its attribute field whose similarity with the index meets a preset similarity threshold; wherein the table structure corresponding to the tag name contains at least one attribute field corresponding to the target entity.

Furthermore, before performing an associated search analysis on the target entity based on a preset distributed retrieval database to obtain the label name and attribute field corresponding to the target entity, it also includes: using each type of entity contained in a preset plurality of colloquial text information as the key in a key-value pair, and using each type of label name and attribute field corresponding to each type of entity as the value in the key-value pair, constructing a target key-value pair for representing the corresponding relationship between each entity and each type of label name and attribute field, and storing it in the distributed retrieval database.

Furthermore, the template construction process is performed based on the tag name and its attribute fields to obtain the prompt word templates corresponding to the target entities, specifically including:

Based on the tag name and its attribute field, embedding processing is performed according to a preset initial template to obtain prompt word templates corresponding to the target entities respectively; wherein the prompt word templates are used to identify the tag name and its attribute field corresponding to the target entities respectively.

Furthermore, the processing of the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and its attribute field specifically includes:

The prompt word template is input into a preset language interaction model for analysis, and a graph database query statement returned by the language interaction model based on the tag name and its attribute fields is obtained. Based on the query statement, a preset graph database is searched to obtain a corresponding question-and-answer interaction result, and the question-and-answer interaction result is output.

Furthermore, the target entity contained in the input spoken text information is obtained, including:

Obtaining input spoken text information;

The colloquial text information is subjected to word segmentation processing to obtain at least one target entity contained in the colloquial text information.

Furthermore, the performing word segmentation processing on the colloquial text information to obtain at least one target entity contained in the colloquial text information specifically includes:

The colloquial text information is segmented to obtain at least one text data block contained in the colloquial text information, a text with ambiguous attribute features is identified from the text data block, and the text with ambiguous attribute features is used as the target entity.

In a second aspect, the present invention further provides a retrieval optimization device for colloquial text content, comprising:

A target entity obtaining unit, used for obtaining a target entity contained in the input spoken text information;

A label and attribute determination unit, configured to perform an associated search analysis on the target entity based on a preset distributed search database to obtain a label name and an attribute field corresponding to the target entity; wherein the distributed search database pre-stores a corresponding relationship between each entity and each type of label name and its attribute field;

A template construction unit, configured to perform template construction processing based on the tag name and its attribute fields to obtain prompt word templates corresponding to the target entities; wherein the prompt word templates include the tag name and its attribute fields corresponding to the target entity;

The query processing unit is used to process the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and its attribute fields.

Furthermore, the label and attribute determination unit is specifically used to:

Taking the target entity as an index, an associated search analysis is performed on the distributed search database to find a tag name and its attribute field whose similarity with the index meets a preset similarity threshold; wherein the table structure corresponding to the tag name contains at least one attribute field corresponding to the target entity.

Furthermore, before performing an associated search analysis on the target entity based on a preset distributed retrieval database to obtain the label name and attribute field corresponding to the target entity, it also includes: a data storage unit, which is used to use each type of entity contained in a preset plurality of colloquial text information as the key in a key-value pair, and each type of label name and attribute field corresponding to the each type of entity as the value in the key-value pair, to construct a target key-value pair for representing the corresponding relationship between each entity and each type of label name and attribute field, and store it in the distributed retrieval database.

Furthermore, the template construction unit is specifically used for:

Based on the tag name and its attribute field, embedding processing is performed according to a preset initial template to obtain prompt word templates corresponding to the target entities respectively; wherein the prompt word templates are used to identify the tag name and its attribute field corresponding to the target entities respectively.

Furthermore, the query processing unit is specifically used to:

The prompt word template is input into a preset language interaction model for analysis, and a graph database query statement returned by the language interaction model based on the tag name and its attribute fields is obtained. Based on the query statement, a preset graph database is searched to obtain a corresponding question-and-answer interaction result, and the question-and-answer interaction result is output.

Furthermore, the target entity obtaining unit is specifically used to:

Obtaining input spoken text information;

The colloquial text information is subjected to word segmentation processing to obtain at least one target entity contained in the colloquial text information.

Furthermore, the performing word segmentation processing on the colloquial text information to obtain at least one target entity contained in the colloquial text information specifically includes:

The colloquial text information is segmented to obtain at least one text data block contained in the colloquial text information, a text with ambiguous attribute features is identified from the text data block, and the text with ambiguous attribute features is used as the target entity.

In a third aspect, the present invention further provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the retrieval optimization method for colloquial text content as described in any one of the above items are implemented.

In a fourth aspect, the present invention further provides a processor-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the retrieval optimization method for colloquial text content as described in any one of the above are implemented.

The retrieval optimization method for colloquial text content provided by the present invention obtains the target entity contained in the input colloquial text information, performs associated retrieval analysis on the target entity based on a preset distributed retrieval database, and obtains the label name and attribute field corresponding to the target entity; wherein the distributed retrieval database pre-stores the correspondence between each entity and each type of label name and attribute field, performs template construction processing based on the label name and attribute field, obtains the prompt word templates corresponding to the target entity, and processes the prompt word template based on a preset language interaction model to obtain the question-answer interaction result returned by the language interaction model based on the label name and attribute field, which can effectively improve the retrieval accuracy and retrieval efficiency of the colloquial text content, thereby improving the user experience.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a flow chart of a method for optimizing retrieval of colloquial text content provided by an embodiment of the present invention;

FIG. 2 is a schematic diagram of a specific flow chart of a method for optimizing the retrieval of colloquial text content provided by an embodiment of the present invention.

3 is a schematic diagram of the structure of a retrieval optimization device for colloquial text content provided by an embodiment of the present invention;

FIG. 4 is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar users, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

The present invention proposes a retrieval optimization method for colloquial text content, aiming to effectively solve the problem that colloquial expressions may cause information loss in natural language processing. This method first detects specific entities and accurately identifies colloquial expressions, and then brings the detected tag name (i.e., label) and its attribute fields and other contents into a large model (i.e., a language interaction model) through an initial template (i.e., a prompt template), telling the language interaction model what label and attribute this entity belongs to. Finally, the language interaction model generates a correct graphic database query statement (i.e., a cypher statement), thereby achieving the purpose of correctly generating cypher statements for colloquial expressions. That is, text2cypher technology is applied when solving natural language query problems. Specifically, the user's query question, as well as the label and attribute fields are told to the language interaction model through a prompt template, so that the language interaction model generates corresponding cypher statements according to the label and attribute fields.

The following is a detailed description of an embodiment of the retrieval optimization method for colloquial text content according to the present invention. As shown in FIG1 , a flow chart of the retrieval optimization method for colloquial text content provided by an embodiment of the present invention is shown, and the specific process includes the following steps:

Step 101: Obtain a target entity contained in the input spoken text information.

Specifically, it is necessary to first obtain the input colloquial text information, and then perform word segmentation processing on the colloquial text information to obtain at least one target entity contained in the colloquial text information. Wherein, the word segmentation processing on the colloquial text information to obtain at least one target entity contained in the colloquial text information, the corresponding implementation process may include: performing word segmentation on the colloquial text information to obtain at least one text data block contained in the colloquial text information, identifying text with ambiguous attribute features from the text data block, and using the text with ambiguous attribute features as the target entity. Wherein, the colloquial text information is the question information queried by the user. As shown in Figure 2, in an embodiment of the present invention, the Query question information queried by the user can be segmented. Through Jieba word segmentation or n-gram word segmentation, the user's Query question information is segmented into n words. For example, in this embodiment, the colloquial text information "How many horizontal stabilizer hoists are stored in Division A" is segmented into several words "Division A, stored, how many, horizontal stabilizer hoists", which are the text data blocks contained in the colloquial text information. Text with ambiguous attribute features (such as horizontal stabilizer hoist, Division A) is identified from the text data blocks, and the text with ambiguous attribute features is used as the target entity. Because horizontal stabilizer hoist and Division A are nouns that do not have specific meanings, the entity attributes are unclear, so it is necessary to predetermine their corresponding label names and attribute fields. For example, if the target entity is name B, the label name corresponding to the target entity can be the table name corresponding to name B (such as the teacher table), and the attribute fields are the gender, age and other fields under the table name.

Step 102: Perform an association search analysis on the target entity based on a preset distributed search database to obtain a tag name and attribute field corresponding to the target entity; wherein the distributed search database pre-stores the corresponding relationship between each entity and each type of tag name and attribute field.

Specifically, the distributed retrieval database can be a pre-built ES (i.e., Elasticsearch) library. In an embodiment of the present invention, the target entity is used as an index, and an associated retrieval analysis is performed on the distributed retrieval database to find out the label name and its attribute field whose similarity with the index meets the preset similarity threshold. Among them, the table structure corresponding to the label name contains at least one attribute field corresponding to the target entity. It should be noted that before executing this step, various entities and corresponding label names and attribute fields that may be contained in a variety of colloquial text information should be obtained from the preset data lake in advance, and the various entities contained in the preset multiple colloquial text information should be used as the key in the key-value pair, and the various label names and attribute fields corresponding to the various entities should be used as the value in the key-value pair, and the target key-value pairs used to represent the corresponding relationship between each entity and each label name and its attribute field are stored in the distributed retrieval database. For example, after the ES library is built, the label and its attribute fields in neo4j are transferred to the ES library. One label can correspond to multiple attribute fields. Since the ES library query speed is very fast, ES storage is used. The ES library is updated every day to ensure the timeliness of the data. The specific parameters need to cancel the ES word segmentation function, and set the query matching return results to only select the cases with a matching degree of 100%. The divided words [A Division, stored, how many, horizontal stabilizer spreader] can be queried in the ES library. After the query, it will be found that [horizontal stabilizer spreader] belongs to the label and its attribute field is [tool] [tooling_name] field. Other words do not belong to any field. Among them, [tool] is the label, and the [tooling_name] field is the attribute field.

Step 103: Perform template construction processing based on the tag name and its attribute fields to obtain prompt word templates corresponding to the target entities respectively; wherein the prompt word templates include the tag name and its attribute fields corresponding to the target entity.

Specifically, based on the label name and its attribute field, embedding processing is performed according to the preset initial template to obtain the prompt word templates corresponding to the target entities; wherein the prompt word template is used to identify the label name and its attribute field corresponding to the target entity. It should be noted that the original prompt only contains the table structure of neo4j, and does not contain which table structure the entity in the user's question may belong to, that is, which label name and attribute field it belongs to. In the new version of prompt (i.e., initial template) used by the present invention, [In the user's question information, [A Division] may be [storage_workshop] in [tool], and [Horizontal Stabilizer Hoist] may be [tooling_name] in [tool]] is added, which effectively tells the big model which labels and attribute fields the entity in the user's question information may belong to. The detected entity is input to the big model through the constructed prompt template (i.e., prompt word template). The prompt word template is a technology used to improve model performance in natural language processing. It guides the big model to focus on task-related information by setting specific prompt words, i.e., prompt words, in the input text. These prompt words can be labels and attribute fields, etc. This can be achieved specifically through a filling method. In the process of generating the prompt word template, a method of combining the label and attribute fields with the input colloquial text information, such as direct insertion, concatenated encoding vectors, and embedding, can be used to improve the performance of large models. For example, the prompt word template can be: [A Division] is [storage_workshop] in [tool], and [Horizontal Stabilizer Spreader] is [tooling_name] in [tool].

Step 104: Process the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and its attribute fields.

Specifically, the prompt word template is input into a preset language interaction model for analysis, and a graph database query statement (i.e., a cypher statement) returned by the language interaction model based on the tag name and its attribute field is obtained. Based on the query statement, a preset graph database (i.e., a neo4j graph database) is searched to obtain the corresponding question-answer interaction result, and the question-answer interaction result is output. It should be noted that the technical solution of the present invention includes two main stages: first, entity detection of colloquial expressions, and then the large model is handed over to generate cypher statements through the prompt template. In the entity detection stage, specific entities in the user's colloquial text information can be accurately captured. This step is to ensure that the subsequent processing is highly targeted and can accurately parse the key information in the colloquial expression. Next, the tag name and attribute field corresponding to the detected colloquial text information will be handed over to the large model through the prompt template for cypher statement generation. This large model can be a generative model based on deep learning, such as the GPT series, which performs well in processing natural language generation tasks. Since the input is already in standard form, the text2cypher model can more accurately parse and generate corresponding cypher statements. In this way, the present invention realizes efficient processing of user's colloquial expressions without modifying the text2cypher model, thereby saving a lot of time and labor costs. That is, firstly, through the specially designed entity detection step, the problem that colloquial expressions may cause information loss is successfully overcome. Secondly, the flow of the entire processing process is clear, which simplifies the processing flow and improves efficiency. The complexity of modifying the text2cypher model is avoided, making the system easier to maintain and upgrade. Through entity detection and prompt engineering, the entity attributes in the user's colloquial text information are successfully told to the large model, thereby providing a feasible solution for structured queries in natural language processing.

The retrieval optimization method for colloquial text content described in the embodiment of the present invention obtains the target entity contained in the input colloquial text information, performs association retrieval analysis on the target entity based on a preset distributed retrieval database, and obtains the label name and attribute field corresponding to the target entity; wherein the distributed retrieval database pre-stores the correspondence between each entity and each type of label name and attribute field, performs template construction processing based on the label name and attribute field, obtains the prompt word templates corresponding to the target entity, and processes the prompt word template based on a preset language interaction model to obtain the question-answer interaction result returned by the language interaction model based on the label name and attribute field, which can effectively improve the retrieval accuracy and retrieval efficiency of the colloquial text content, thereby improving the user experience.

Corresponding to the above-mentioned method for searching and optimizing the content of spoken text, the present invention also provides a device for searching and optimizing the content of spoken text. Since the embodiment of the device is similar to the above-mentioned method embodiment, the description is relatively simple. For relevant matters, please refer to the description of the above-mentioned method embodiment. The embodiment of the device for searching and optimizing the content of spoken text described below is only schematic. Please refer to Figure 3, which is a schematic diagram of the structure of a device for searching and optimizing the content of spoken text provided in an embodiment of the present invention. The device for searching and optimizing the content of spoken text described in the present invention specifically includes the following parts:

A target entity obtaining unit 301 is used to obtain a target entity contained in the input spoken text information;

The tag and attribute determination unit 302 is used to perform an associated search analysis on the target entity based on a preset distributed search database to obtain a tag name and an attribute field corresponding to the target entity; wherein the distributed search database pre-stores the corresponding relationship between each entity and each type of tag name and its attribute field;

The template construction unit 303 is used to perform template construction processing based on the tag name and its attribute field to obtain prompt word templates corresponding to the target entities; wherein the prompt word templates include the tag name and its attribute field corresponding to the target entity;

The query processing unit is used to process the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and its attribute fields.

Furthermore, the label and attribute determination unit is specifically used to:

Taking the target entity as an index, an associated search analysis is performed on the distributed search database to find a tag name and its attribute field whose similarity with the index meets a preset similarity threshold; wherein the table structure corresponding to the tag name contains at least one attribute field corresponding to the target entity.

Furthermore, before performing an associated search analysis on the target entity based on a preset distributed retrieval database to obtain the label name and attribute field corresponding to the target entity, it also includes: a data storage unit, which is used to use each type of entity contained in a preset plurality of colloquial text information as the key in a key-value pair, and each type of label name and attribute field corresponding to the each type of entity as the value in the key-value pair, to construct a target key-value pair for representing the corresponding relationship between each entity and each type of label name and attribute field, and store it in the distributed retrieval database.

Furthermore, the template construction unit is specifically used for:

Based on the tag name and its attribute field, embedding processing is performed according to a preset initial template to obtain prompt word templates corresponding to the target entities respectively; wherein the prompt word templates are used to identify the tag name and its attribute field corresponding to the target entities respectively.

Furthermore, the query processing unit is specifically used to:

The prompt word template is input into a preset language interaction model for analysis, and a graph database query statement returned by the language interaction model based on the tag name and its attribute fields is obtained. Based on the query statement, a preset graph database is searched to obtain a corresponding question-and-answer interaction result, and the question-and-answer interaction result is output.

Furthermore, the target entity obtaining unit is specifically used to:

Obtaining input spoken text information;

The colloquial text information is subjected to word segmentation processing to obtain at least one target entity contained in the colloquial text information.

Furthermore, the performing word segmentation processing on the colloquial text information to obtain at least one target entity contained in the colloquial text information specifically includes:

The colloquial text information is segmented to obtain at least one text data block contained in the colloquial text information, a text with ambiguous attribute features is identified from the text data block, and the text with ambiguous attribute features is used as the target entity.

The retrieval optimization device for colloquial text content described in the embodiment of the present invention obtains the target entity contained in the input colloquial text information, performs association retrieval analysis on the target entity based on a preset distributed retrieval database, and obtains the label name and attribute field corresponding to the target entity; wherein the distributed retrieval database pre-stores the correspondence between each entity and each type of label name and attribute field, performs template construction processing based on the label name and attribute field, obtains the prompt word templates corresponding to the target entity, and processes the prompt word template based on a preset language interaction model to obtain the question-answer interaction result returned by the language interaction model based on the label name and attribute field, which can effectively improve the retrieval accuracy and retrieval efficiency of the colloquial text content, thereby improving the user experience.

Corresponding to the above-mentioned retrieval optimization method for colloquial text content, the present invention also provides an electronic device. Since the embodiment of the electronic device is similar to the above-mentioned method embodiment, the description is relatively simple. For relevant parts, please refer to the description of the above-mentioned method embodiment part. The electronic device described below is only schematic. As shown in Figure 4, it is a schematic diagram of the physical structure of an electronic device disclosed in an embodiment of the present invention. The electronic device may include: a processor (processor) 401, a memory (memory) 402 and a communication bus 403, wherein the processor 401 and the memory 402 complete mutual communication through the communication bus 403, and communicate with the outside through the communication interface 404. The processor 401 can call the logic instructions in the memory 402 to execute the retrieval optimization method of the colloquial text content, which method includes: obtaining the target entity contained in the input colloquial text information; performing an associated retrieval analysis on the target entity based on a preset distributed retrieval database to obtain the label name and attribute field corresponding to the target entity; wherein the distributed retrieval database pre-stores the correspondence between each entity and each type of label name and attribute field; performing template construction processing based on the label name and attribute field to obtain the prompt word templates corresponding to the target entity; wherein the prompt word template includes the label name and attribute field corresponding to the target entity; processing the prompt word template based on a preset language interaction model to obtain the question-answer interaction result returned by the language interaction model based on the label name and attribute field.

In addition, the logic instructions in the above-mentioned memory 402 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as storage chips, U disks, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks or optical disks.

On the other hand, an embodiment of the present invention further provides a computer program product, the computer program product includes a computer program stored on a processor-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the computer can execute the retrieval optimization method of the colloquial text content provided by the above-mentioned method embodiments. The method includes: obtaining a target entity contained in the input colloquial text information; performing an associated retrieval analysis on the target entity based on a preset distributed retrieval database to obtain a label name and its attribute field corresponding to the target entity; wherein the distributed retrieval database pre-stores the correspondence between each entity and each type of label name and its attribute field; performing template construction processing based on the label name and its attribute field to obtain a prompt word template corresponding to the target entity; wherein the prompt word template includes the label name and its attribute field corresponding to the target entity; processing the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the label name and its attribute field.

On the other hand, an embodiment of the present invention further provides a processor-readable storage medium, on which a computer program is stored, and when the computer program is executed by the processor, it is implemented to execute the retrieval optimization method of the colloquial text content provided in the above embodiments. The method includes: obtaining the target entity contained in the input colloquial text information; performing an associated retrieval analysis on the target entity based on a preset distributed retrieval database to obtain the label name and attribute field corresponding to the target entity; wherein the distributed retrieval database pre-stores the correspondence between each entity and each type of label name and attribute field; performing template construction processing based on the label name and attribute field to obtain the prompt word template corresponding to the target entity; wherein the prompt word template includes the label name and attribute field corresponding to the target entity; processing the prompt word template based on a preset language interaction model to obtain the question-answer interaction result returned by the language interaction model based on the label name and attribute field.

The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO)), optical storage (such as CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drives (SSDs)), etc.

The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM/RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. A method for optimizing retrieval of spoken text content, comprising:

Obtaining a target entity contained in the input spoken text information;

performing association search analysis on the target entity based on a preset distributed search database to obtain a tag name and an attribute field thereof corresponding to the target entity; the distributed search database is pre-stored with the corresponding relation between each entity and each label name and attribute field thereof;

Template construction processing is carried out based on the tag names and attribute fields thereof, and prompt word templates corresponding to the target entities respectively are obtained; the prompt word template comprises the tag name and the attribute field thereof corresponding to the target entity;

and processing the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and the attribute field thereof.

2. The method for optimizing the retrieval of spoken text content according to claim 1, wherein the performing, based on a preset distributed retrieval database, the association retrieval analysis on the target entity to obtain the tag name and the attribute field thereof corresponding to the target entity includes:

Taking the target entity as an index, and carrying out association search analysis on the target entity to the distributed search database so as to find out a label name and an attribute field thereof, wherein the similarity between the label name and the index meets a preset similarity threshold value; the table structure corresponding to the tag name comprises at least one attribute field corresponding to the target entity.

3. The method for optimizing the retrieval of spoken text content according to claim 1, wherein before performing the association retrieval analysis on the target entity based on a preset distributed retrieval database to obtain the tag name and the attribute field thereof corresponding to the target entity, further comprising: and taking various entities contained in the preset various spoken text information as keys in a key value pair, taking various label names and attribute fields corresponding to the various entities as values in the key value pair, constructing a target key value pair for representing the corresponding relation between the various entities and the various label names and attribute fields respectively, and storing the target key value pair in the distributed search database.

4. The method for optimizing the retrieval of spoken text content according to claim 1, wherein the template construction process is performed based on the tag name and the attribute field thereof to obtain the respective prompt word templates corresponding to the target entities, specifically comprising:

Based on the tag name and the attribute field thereof, performing embedding processing according to a preset initial template to obtain prompt word templates corresponding to the target entities respectively; the prompt word template is used for identifying tag names and attribute fields thereof respectively corresponding to the target entities.

5. The method for optimizing the retrieval of the spoken text content according to claim 1, wherein the processing the prompt word template based on the preset language interaction model to obtain the question-answer interaction result returned by the language interaction model based on the tag name and the attribute field thereof specifically comprises:

Inputting the prompt word template into a preset language interaction model for analysis, obtaining a graphic database query statement returned by the language interaction model based on the tag name and the attribute field thereof, searching a preset graph database based on the query statement to obtain a corresponding question-answer interaction result, and outputting the question-answer interaction result.

6. The method according to claim 1, wherein obtaining the target entity included in the inputted spoken text information comprises:

Obtaining input spoken text information;

And performing word segmentation processing on the spoken text information to obtain at least one target entity contained in the spoken text information.

7. The method for optimizing the retrieval of spoken text content according to claim 6, wherein the word segmentation processing is performed on the spoken text information to obtain at least one target entity included in the spoken text information, specifically comprising:

And performing word segmentation on the spoken text information to obtain at least one text data block contained in the spoken text information, identifying text with fuzzy attribute characteristics from the text data block, and taking the text with fuzzy attribute characteristics as the target entity.

8. A retrieval optimizing apparatus for spoken text content, comprising:

A target entity obtaining unit for obtaining a target entity contained in the inputted spoken text information;

The tag and attribute determining unit is used for carrying out association search analysis on the target entity based on a preset distributed search database to obtain a tag name and an attribute field corresponding to the target entity; the distributed search database is pre-stored with the corresponding relation between each entity and each label name and attribute field thereof;

The template construction unit is used for carrying out template construction processing based on the tag names and attribute fields thereof to obtain prompt word templates corresponding to the target entities respectively; the prompt word template comprises the tag name and the attribute field thereof corresponding to the target entity;

and the query processing unit is used for processing the prompt word template based on a preset language interaction model to obtain a question-answer interaction result returned by the language interaction model based on the tag name and the attribute field thereof.

9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method for optimizing the retrieval of spoken text content according to any one of claims 1 to 7 when the computer program is executed.

10. A processor-readable storage medium having stored thereon a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the retrieval optimization method of spoken text content according to any one of claims 1 to 7.