WO2023178965A1

WO2023178965A1 - Intent recognition method and apparatus, and electronic device and storage medium

Info

Publication number: WO2023178965A1
Application number: PCT/CN2022/120942
Authority: WO
Inventors: 董益华
Original assignee: 平安科技（深圳）有限公司
Priority date: 2022-03-25
Filing date: 2022-09-23
Publication date: 2023-09-28
Also published as: CN114661910A

Abstract

The embodiments of the present application relate to the technical field of artificial intelligence. Disclosed are an intent recognition method and apparatus, and an electronic device and a storage medium. The intent recognition method comprises: acquiring first target intent sample data according to original intent sample data, wherein the first target intent sample data comprises non-long-tail input sample data and intent matching result sorting data; performing entity abstraction on the non-long-tail input sample data, so as to obtain an abstract generalized entity word; performing logic combination on the abstract generalized entity word and the intent matching result sorting data, so as to generate an intent matching generalization dictionary; constructing a first intent recognition model according to the intent matching generalization dictionary; when it is determined that input data to be subjected to recognition is non-long-tail input data, inputting said input data into the first intent recognition model; and outputting an intent recognition result of said input data according to the first intent recognition model. The technical solution in the embodiments of the present application can increase the accuracy rate of intent understanding.

Description

An intention recognition method, device, electronic equipment and storage medium

This application requests the priority of the Chinese patent application submitted to the China Patent Office on March 25, 2022, with the application number 202210307597.9 and the application title "An intent recognition method, device, electronic device and storage medium", and its entire content is approved by This reference is incorporated into this application.

Technical field

The embodiments of the present application relate to the field of artificial intelligence technology such as information processing, and in particular, to an intention recognition method, device, electronic device and storage medium.

Background technique

Intent recognition can also be called intent detection (Intent Detection). It is used to determine which field the input information is used to perform which operation. Its essence is a multi-class classification problem and is widely used in intelligent search and human-computer interaction. interactive technology. One embodiment of intelligent interaction is that intelligent products or applications can understand requirements through intent recognition and provide appropriate responses based on the requirements.

An important part of intent recognition is processing query (query request). Every query hides the real query intention. When understanding the query, you need to use many different strategies to explore the requirements behind it. Therefore, how to correctly identify the query intent, analyze the content of interest, and display the most interesting content within limited resources is of great significance to improving the experience of intelligent interactive functions.

In the process of implementing this application, the inventor discovered that the existing technology has the following defects: At present, the existing intent recognition methods basically adopt the one-size-fits-all principle when processing queries, and do not differentiate between long-tail queries and non-long-tail queries. Two different types of query processing methods. The concentration of long-tail queries is low, but the cumulative number is close to infinite. Although the search volume of a single long-tail query is not large, it has a long-tail effect, and the total search volume is comparable to the non-long-tail query volume in the head. If we do not distinguish the processing methods of two different types of queries, long-tail query and non-long-tail query, and use a unified processing method to understand the query, it will lead to a low accuracy of intent understanding.

Contents of the invention

Embodiments of the present application provide an intention recognition method, device, electronic device and storage medium, which can improve the accuracy of intention understanding.

According to one aspect of the present application, an intent identification method is provided, including:

Obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result ranking data;

Perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

Logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalized dictionary;

Construct a first intention recognition model according to the intention matching generalization dictionary;

When it is determined that the input data to be recognized is non-long-tail input data, input the input data to be recognized into the first intention recognition model;

The intent recognition result of the input data to be recognized is output according to the first intent recognition model.

According to another aspect of the present application, an intention recognition device is provided, including:

The first sample data acquisition module is used to obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result sorting data;

An abstract generalized entity word acquisition module is used to perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

An intent matching generalization dictionary generation module, used to logically combine the abstract generalization entity words and the intent matching result sorting data to generate an intent matching generalization dictionary;

A first intention recognition model building module, configured to build a first intention recognition model according to the intention matching generalization dictionary;

An input data input module to be recognized, configured to input the input data to be recognized into the first intention recognition model when it is determined that the input data to be recognized is non-long-tail input data;

An intent recognition result output module is configured to output the intent recognition result of the input data to be recognized according to the first intent recognition model.

According to another aspect of the present application, an electronic device is provided, the electronic device including:

at least one processor; and

a memory communicatively connected to the at least one processor; wherein,

The memory stores a computer program that can be executed by the at least one processor. When the computer program is executed by the at least one processor, the intention recognition method is implemented, including:

According to another aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores computer instructions. The computer instructions are used to implement an intention recognition method when executed by a processor, including:

The embodiment of the present application obtains the first target intent sample data including non-long tail input sample data and intent matching result sorting data based on the original intent sample data, and then performs entity abstraction on the non-long tail input sample data to obtain abstract generalized entity words. , and logically combine abstract generalized entity words and intent matching result sorting data to generate an intent matching generalization dictionary, thereby building a first intent recognition model based on the intent matching generalization dictionary to use the first intent recognition model to analyze the input data The result is that the input data to be recognized of non-long-tail input data is input for intent recognition, and the intent recognition result of the input data to be recognized is output. This solves the problem of low accuracy of intent understanding in existing intent recognition methods, and can improve the accuracy of intent understanding. Rate.

It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of the application, nor is it intended to limit the scope of the application. Other features of the present application will become readily understood from the following description.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of an intention identification method provided in Embodiment 1 of the present application;

Figure 2 is a flow chart of an intention identification method provided in Embodiment 2 of the present application;

Figure 3 is a schematic diagram of a BERT model training process provided in Embodiment 2 of the present application;

Figure 4 is a schematic diagram of an intention recognition device provided in Embodiment 4 of the present application;

FIG. 5 shows a schematic structural diagram of an electronic device that can be used to implement embodiments of the present application.

Detailed ways

In order to enable those in the technical field to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only These are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall within the scope of protection of this application.

It should be noted that the terms "first", "second", etc. in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances so that the embodiments of the application described herein can be practiced in sequences other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, e.g., a process, method, system, product, or apparatus that encompasses a series of steps or units and need not be limited to those explicitly listed. Those steps or elements may instead include other steps or elements not expressly listed or inherent to the process, method, product or apparatus.

The embodiments of this application can obtain and process relevant data based on artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .

Basic artificial intelligence technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometric technology, speech processing technology, natural language processing technology, and machine learning/deep learning.

Embodiment 1

Figure 1 is a flow chart of an intent recognition method provided in Embodiment 1 of the present application. This embodiment can be applied to the situation of constructing an intent recognition model based on non-long-tail input sample data to perform intent recognition on non-long-tail input data. The method can be executed by an intention recognition device, which can be implemented by software and/or hardware, and can generally be integrated in an electronic device. The electronic device can be a terminal device or a server device. Embodiments of the present application The specific device type of the electronic device is not limited. Correspondingly, as shown in Figure 1, the method includes the following operations:

S110. Obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result ranking data.

The first target intention sample data may be sample data used to build the first intention recognition model. The original intent sample data can be the full amount of historical intent data. Optionally, the intent data may be user intent data, such as user query data, or intent data automatically generated by a device or program, such as a data search instruction issued by a device or a simulated real user issuing query data, etc., in the embodiment of this application The data type and generation method of intent data are not limited. Input sample data is sample data that requires intent to understand. For example, it can be query data input by the user, or query data input by a device or program, etc. The input sample data may be text type data or voice type data. The embodiment of the present application does not limit the data type of the user input sample data. It can be understood that long-tail data refers to non-target data but is related to the target data, and can also bring combined data that can bring search traffic. Non-long tail data refers to target data. The non-long-tail input sample data can be input sample data in a non-long-tail form, that is, the non-long-tail input sample data can be directly used as keywords or directly segmented to obtain multiple keywords for intent understanding. The ranking data of the intent matching results can be the ranking results of the feedback data obtained after understanding the intent of the non-long-tail user input sample data.

In this embodiment of the present application, the intent may be any type of intent, including but not limited to search intent, interaction intent, etc. After an object with an intent, such as a user, device, or program, it can be simply called an intent output object.

For example, in the field of intelligent search technology, the intent may be a search intent. When the intent output object needs to search for relevant content in the network or application, for the search statement provided by the intent output object, the search intent of the intent output object can be identified based on the search statement, so that the intent output object can be output according to the search intent of the intent output object. Recommend relevant content. Correspondingly, the non-long-tail input sample data can be the non-long-tail search statement of the intent output object, and the intent matching result ranking data can be the operation ranking data of the intent output object according to the feedback intent recognition result. In a specific example, assuming that the non-long-tail input sample data is the function module search data of the APP, the intent matching result ranking data can be the ranking data of the click frequency of the function module that the user feedbacks on the function module search data of the APP.

For example, in the field of intelligent interaction technology, the intention may be a conversation intention or an interaction intention. For example, in an intelligent question answering system, the intention of the intention output object can be identified based on the sentence input by the intention output object (which can be text type, speech type, etc.), and an appropriate response can be provided for the intention output object. Correspondingly, the non-long-tail input sample data can be the conversational statements of the intent output object, and the intent matching result ranking data can be the operation ranking data of the intent output object based on the feedback intent recognition results. In a specific example, assuming that the non-long-tail input sample data is the dialogue voice data input by the user to the intelligent question and answer system, the intent matching result ranking data can be the user's recognition of the response voice results of the intelligent question and answer system for the dialogue voice data feedback. Ranked data.

S120. Perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words.

Among them, abstract generalized entity words can be generalized data structures constructed from entity words abstracted from non-long-tail input sample data.

After obtaining the non-long-tail input sample data in the first target intention sample data, entity abstraction can be performed on the non-long-tail input sample data. The so-called entity abstraction refers to extracting entity words from non-long-tail input sample data to construct abstract generalized entity words based on the extracted entity words.

In a specific example, taking an e-commerce APP as an example, when the non-long-tail input sample data entered by the user is "raw wood pulp toilet paper", the abstract generalized entity words that can be obtained based on the non-long-tail input sample data It is "original wood pulp#commodity#".

S130. Logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalized dictionary.

Among them, the intent matching generalization dictionary can provide intent matching result sorting data for non-long-tail input sample data to determine the final intent understanding result of non-long-tail input sample data. That is, the intent matching generalization dictionary may be a structured dictionary used to find intent understanding results for non-long-tail input sample data.

Correspondingly, after obtaining the corresponding abstract generalized entity words based on the non-long-tail input sample data, the data can be sorted based on the abstract generalized entity words corresponding to the non-long-tailed input sample data and the intent matching results, and each non-long-tailed Input sample data to construct a matching dictionary query unit, and then construct a matching dictionary query unit based on each non-long-tail input sample data to construct an intent matching generalized dictionary.

In the embodiment of the present application, the intent matching generalization dictionary can use entity abstraction of non-long-tail input sample data to obtain abstract generalized entity words as the benchmark matching unit, and use the intent matching result sorting data of the same non-long-tail input sample data as The alternative intent understanding results of the benchmark matching unit, thereby combining the benchmark matching unit and each alternative intent understanding result of the same non-long-tail input sample data into a dictionary query in the intent matching generalization dictionary for the non-long-tail input sample data unit.

In a specific example, assume that the non-long-tail input sample data of a medical APP is "Why are my legs cramping when I sleep at night?", the abstract generalized entity word obtained for the non-long-tail input sample data can be "at night" "What's going on with sleeping#bodypart##disease#", the sorting data of the intent matching results of the non-long-tail input sample data are: "Function module 01": 30; "Function module 02": 20; "Function module 03" :10. Among them, the subsequent field value of each functional module can be the number of times the user clicks on the functional module in history. For example, for "Function Module 01": 30, it means that when each user enters the search query "What's wrong with leg cramps while sleeping at night" in the medical APP, in the search results of each functional module reported by the user to the medical APP, the function Module 01 was clicked 30 times. It is understandable that the greater the number of historical clicks, the better the functional module matches the user's search intent. Correspondingly, by logically combining the above abstract generalized entity words and the sorted data of intent matching results, we can obtain the dictionary query unit corresponding to the non-long-tail input sample data "Why are my feet cramping while sleeping at night", and its data structure is as follows:

It is understandable that the intent matching generalization dictionary can be set according to the field. For example, an intent matching generalization dictionary is constructed corresponding to a technical field. Alternatively, the intent matching generalization dictionary may also involve multiple fields at the same time, which is not limited in the embodiments of the present application.

S140. Construct a first intention recognition model according to the intention matching generalization dictionary.

Wherein, the first intention recognition model is used to identify the intention of non-long-tail input data.

In the embodiment of this application, the first intention recognition model is constructed based on the intention matching generalization dictionary. The intention matching generalization dictionary can be directly used as the first intention recognition model to perform intention recognition on non-long-tail input data to obtain the final intention to understand the results.

S150. If it is determined that the input data to be recognized is non-long-tail input data, input the input data to be recognized into the first intention recognition model.

S160: Output the intention recognition result of the input data to be recognized according to the first intention recognition model.

Correspondingly, if it is determined that the input data to be recognized is non-long-tail input data, the input data to be recognized can be input into the constructed first intention recognition model to identify the non-long-tail input data to be recognized through the first intention recognition model.

It can be seen that the intent matching generalization dictionary constructed by sorting the non-long-tail input sample data and its matching intent matching results is used as the first intent recognition model. The first intent recognition model can be used to identify non-long-tail input data. Perform intent recognition to improve the accuracy of understanding the intent of non-long-tail input data.

Embodiment 2

Figure 2 is a flow chart of an intention identification method provided in Embodiment 2 of the present application. This embodiment is embodied based on the above embodiment. In this embodiment, obtaining the first target based on the original intention sample data is given. Intent sample data, entity abstraction of the non-long-tail input sample data, construction of a first intent recognition model based on the intent matching generalization dictionary, and construction of a second intent recognition model and a target intent recognition model. Method to realize. Correspondingly, as shown in Figure 2, the method in this embodiment may include:

S210: Filter the non-long-tail input sample data from the original intention sample data according to the non-long-tail input data filtering rules.

Among them, the non-long-tail input data filtering rules are also rules used to filter non-long-tail input data.

Optionally, non-long-tail input data filtering rules can be used to limit the number of keywords in the data, that is, when the number of keywords in the original intent sample data is greater than a certain threshold, the data is long-tail input sample data; when When the number of keywords in the original intent sample data is less than or equal to a certain threshold, the data is non-long-tail input sample data. The threshold used to divide long-tail input data and non-long-tail input data can be set according to actual needs, such as 20, etc. The embodiment of the present application does not limit the specific value of the threshold.

S220: Obtain the association intention feedback data of the non-long-tail input sample data.

Among them, the association intention feedback data can be data fed back to non-long tail input sample data, as well as relevant statistical data of the fed back data, etc.

S230: Sort the associated intention feedback data to obtain the intention matching result sorting data.

After filtering out the non-long-tail input sample data from the original intent sample data, the associated intent feedback data of the non-long-tail input sample data can be further obtained, and the obtained associated intent feedback data can be sorted to obtain the intent matching result ranking data. Optionally, when sorting the association intention feedback data, you can sort the data in descending order.

In a specific example, assuming that the associated intent feedback data is the click frequency data of the application function module, the click frequency data of the application function module can be sorted in order from high to low click frequency to obtain the intent matching result ranking data. Assuming that the associated intent feedback data is the response interaction data of the intelligent interaction system to the input data of the intent output object, such as the machine interaction frequency data of query in human-computer dialogue, etc., then the response interaction data can be sorted in order from high to low interaction frequency. , get the sorting data of intent matching results. It is understandable that the higher the frequency of clicks or the frequency of machine interaction, the higher the intention output object's recognition of the intention recognition results.

In a specific example, assume that the non-long-tail input sample data is "Why are my legs cramping when I sleep at night?" In the historical data, the APP can feed back 3 matching functional modules as associations for the non-long-tail input sample data. The intention feedback data are respectively function module 01, function module 02 and function module 03. Among them, the number of historical clicks of function module 01 is 30, the number of historical clicks of function module 02 is 20, and the number of historical clicks of function module 03 is 10, then the associated intention feedback data is sorted, and the sorted data of the intention matching results can be: "Function module 01": 30; "Function module 02": 20; "Function module 03": 10.

S240. Perform entity abstraction on the non-long-tail input sample data according to the entity word dictionary to obtain initial abstract entity words.

The entity word dictionary may be a dictionary composed of entity words. The initial abstract entity words may be each entity word obtained by abstracting the non-long-tail input sample data.

S250. Classify and group the initial abstract entity words to obtain the abstract generalized entity words.

In the embodiment of this application, entity abstraction can be performed on the non-long-tail input sample data first according to the entity word dictionary, and each entity word that makes up the non-long-tail input sample data is used as the initial abstract entity word, and the initial abstract entity obtained by abstraction can be further The words are classified and grouped to obtain abstract generalized entity words.

In a specific example, taking a medical APP as an example, the entity word dictionary may include, but is not limited to, entity words such as disease, symptom, department, product, and body part. Assuming that the input sample data is "virgin wood pulp toilet paper", then the input sample data can be abstracted as "virgin wood pulp#commodity#", assuming that the input sample data is "zinc gluconate oral solution", then the input sample data can be abstracted as "Drug##Body Part#Taking Solution". Furthermore, it is also necessary to classify and group the initial abstract entity words after abstraction. For example: the initial abstract entity word of "drug##body part#taking solution" can be classified as "#drug#|#bodypart#". That is, "#pharmaceutical#|#bodypart#" is a type of abstract generalized entity word that can be used to match input sample data of types such as "Zinc Gluconate Oral Solution" and "Mugwort and Ginger Foot Patch".

S260. Logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalized dictionary.

In a specific example, take the user query sample data including body parts and disease entity words as an example. Suppose there are currently three user query sample data: "What is going on with foot cramps when sleeping at night", "Denumeral 12" "What medicine is good for erosion?" and "What medicine should I take to relieve my back fever twice a day?" The intent matching generalization dictionary that can be constructed based on these three user query sample data is as follows:

The data results of the above intention matching generalization dictionary have extremely strong generalization for queries that have both body parts and disease entities.

The above generalized dictionary of intent matching can perform intent recognition on non-long tail input data based on the DFA (Deterministic Finite Automaton) algorithm.

S270. Construct an input data edit distance calculation module according to the dictionary elements of the intent matching generalized dictionary.

Among them, the dictionary element is also the dictionary query unit intended to match the generalized dictionary. For example, the dictionary elements intended to match the generalized dictionary can be, for example, "What's going on when you go to sleep at night #body part # #disease #" and "# Body Part # There are two waves a day #disease# What's going on, what medicine should be taken to relieve it "wait. The input data edit distance calculation module can be used to calculate the edit distance between input data and dictionary elements. It can be understood that the smaller the edit distance is, the closer the input data is to the dictionary elements, that is, the closer the input data is to the dictionary elements.

It can be understood that dictionary elements can be used as query matching benchmarks to match input data. For example, the dictionary element constructed based on the user's historical behavior can be: "What's going on with #bodypart##disease# when you sleep at night?" When the input non-long-tail data is "What's wrong with leg cramps while sleeping at night" or "What's wrong with leg cramps while sleeping at night", it can be considered as a general expression of the dictionary element "What's wrong with sleeping at night#bodypart##disease#" change. That is, "What's wrong with leg cramps while sleeping at night?" or "What's wrong with leg cramps while sleeping at night?" The dictionary elements matched by the intent matching generalization dictionary are: "What's wrong with sleeping at night#bodypart##disease#" thing".

After obtaining the intent matching generalization dictionary, the input data edit distance calculation module can be constructed according to the element structure of the dictionary elements of the intent matching generalization dictionary to calculate the edit distance between the input data and the dictionary elements through the input data edit distance calculation module ( Also called similarity).

S280. Construct the first intention recognition model according to the input data editing distance calculation module and the intention matching generalization dictionary.

Among them, the first intent recognition model is used to identify the intent of non-long-tail input data.

In this embodiment of the present application, the first intention recognition model may include two modules: an input data editing distance calculation module and an intention matching generalization dictionary. In addition to this, the first intention recognition model may also include an entity word dictionary for entity abstraction. The first intent recognition model can be used to identify intent on non-long-tail input data.

Specifically, the first intention recognition model can first perform entity abstraction on the input query of the intention output object based on the entity word dictionary to obtain the initial abstract entity words, and perform entity classification and grouping on the initial abstract entity words to obtain the final abstract generalized entity words. . Then, the first intent recognition model can calculate the edit distance between the abstract generalized entity words and each dictionary element in the intent matching generalized dictionary. Finally, the first intention recognition model takes the intent included in the dictionary element with the smallest edit distance between the intent matching generalization dictionary and the abstract generalization entity word as the intent of the input query. It is understandable that if the dictionary element with the smallest edit distance between the intent matching generalized dictionary and the abstract generalized entity word includes multiple intents, the first intent can also be selected as the intent of the input query based on the sorting results of each intent. .

S290. Construct a second intention recognition model.

Among them, the second intention recognition model can be used to identify the intention of long-tail input data to obtain the final intention understanding result.

In an optional embodiment of the present application, building the second intention recognition model may include: pre-training a preset neural network model based on pre-training sample data to obtain a pre-trained neural network model; and obtaining a pre-trained neural network model based on the original intention sample data. Second target intention sample data; wherein the second target intention sample data includes long-tail input sample data and intention marking result data; train the pre-trained neural network model according to the second target intention sample data, and obtain Second intention recognition model.

The preset neural network model can be any type of neural network model that can realize the intention recognition function. The pre-trained neural network model may be a neural network model obtained by pre-training a preset neural network model. The second target intention sample data may be sample data used to formally train the second intention recognition model. Long-tail input sample data can be input sample data in the long-tail form, which has the characteristics of semantic complexity. The intent-labeled result data may be data that is pre-labeled with intent that matches the long-tail input sample data.

In this embodiment of the present application, a neural network model can be used for intent recognition for long-tail input sample data. Specifically, the pre-trained neural network model can be pre-trained using pre-training sample data to train the data understanding ability of the pre-set neural network model and obtain the pre-trained neural network model. After the pre-training is completed, the second target intention sample data pre-trained neural network model can be further used for training to obtain the second intention recognition model.

In a specific example, assuming that the preset neural network model is a BERT (Bidirectional Encoder Representation from Transformers, a language representation model) model, pre-training the preset neural network model can include two pre-training tasks, one is MLM (Masked Language Model, masked language model) pre-training task, and another is NSP (Next Sentence Prediction, next sentence prediction) pre-training task. Among them, the MLM pre-training task can be understood as a cloze task. You can randomly mask a certain number of words in each sentence (such as 15% of the sentence) and use their context to make predictions. For example, for the pre-training sample data "my dog is hairy" is converted to "my dog is[MASK]". Here, "hairy" is masked. Then use unsupervised learning method to predict the word at the mask position. The NSP pre-training task can be understood as a text matching task. Specifically, some sentence pairs A and B can be selected, 50% of the data B is one of the sentence segments of A, and the remaining 50% of the data B is randomly selected from the corpus, so that the network can learn the correlation. For example, suppose sentence A is: Simplify the work process of searching for operation and maintenance personnel and improve the work efficiency of operation and maintenance personnel. One of the short sentences of sentence B can be one of the short sentences of sentence A, and the other short sentence of sentence B can be a randomly selected sentence fragment. For example, sentence B can be: Simplify the workflow of search operation and maintenance personnel, and need to be punctual Have a meal. The above pre-training process can enable the pre-trained neural network model to understand the relationship between two sentences, thereby allowing the pre-trained neural network model to better adapt to the above-mentioned data processing tasks.

In a specific example, assuming that the preset neural network model is the BERT model, when training the BERT model based on the second target intention sample data, the CrossEntropy loss function and BP (Back Propagation) propagation mechanism can be used. Let the model learn and update the network weight parameters independently to implement the training process. The trained BERT model serves as the second intention recognition model. Among them, the training process of the BERT model can be seen in Figure 3. BERT is a powerful pre-training model that relies on Transformers as feature extractors. In view of its huge number of parameters and super feature representation capabilities, it can learn deep semantic information in text. Using BERT to embedding long-tail input data can map text information to a high-dimensional vector space, and use embedding vectors to represent the semantic information of long-tail input data.

S2110. Construct a target intention recognition model according to the first intention recognition model and the second intention recognition model.

Among them, the target intent recognition model is a model that can perform intent recognition on any type of input data.

S2120. Obtain the input data to be identified, classify the input data to be identified, and obtain the input data classification result.

Wherein, the input data classification results include long-tail input data and non-long-tail input data.

The input data to be recognized may be input data that requires intent recognition. For example, it can be query data input in real time by the user, or query data input in real time by a device or program, etc. The input data to be recognized may be text type data or voice type data. The embodiment of the present application does not limit the data type of the input data to be recognized. The input data classification result is also the classification result of the input data to be identified.

After obtaining the input data to be recognized, in order to determine the model for intent recognition of the input data to be recognized, the input data to be recognized can be classified first to determine whether the input data to be recognized is long-tail input data or non-long-tail input data.

In an optional embodiment of the present application, classifying the input data to be identified may include: when it is determined that the data length of the input data to be identified is less than or equal to a preset data length threshold, determining that the input data to be identified is The input data classification result of the input data to be identified is non-long tail input data; when it is determined that the data length of the input data to be identified is greater than the preset data length threshold, the input data classification of the input data to be identified is determined The result is long tail input data.

The preset data length threshold can be a length threshold used to divide long-tail data and non-long-tail data. For example, the preset data length threshold can be set to 20 or 25, etc., which can be set according to actual needs. The application embodiment does not limit the specific value of the preset data length threshold.

Specifically, the data length of the input data to be recognized can be determined to classify the input data to be recognized based on the data length. Among them, the data length can be the number of words or characters in the input data to be recognized. For example, for the input data to be recognized "Why do my legs cramp while sleeping at night", the data length is 12. Correspondingly, if it is determined that the data length of the input data to be identified is less than or equal to the preset data length threshold, it can be determined that the input data classification result of the input data to be identified is non-long tail input data; otherwise, the input data of the input data to be identified can be determined. The classification results are long-tail input data.

S2130. Determine whether the input data to be identified is non-long tail input data. If so, execute S2140. Otherwise, execute S2150.

S2140. Input the input data to be recognized into the first intention recognition model, so as to output the intention recognition result of the input data to be recognized according to the first intention recognition model.

S2150. Input the input data to be recognized into the second intention recognition model, so as to output the intention recognition result of the input data to be recognized according to the second intention recognition model.

Specifically, the first intention recognition model of the target intention recognition model can be used to perform intent recognition on non-long-tail input data, and the second intention recognition model of the target intent recognition model can be used to perform intent recognition on long-tail input data, thereby obtaining the input. Intent identification results of data.

Among them, the first intention recognition model can be based on the DFA algorithm of a large number of behaviors, which can not only improve the efficiency of intention recognition, but also has good generalization. The second intention recognition model can well extract the semantic information implicit in long-tail input data, and can better represent its semantic features. Through differentiated processing of long-tail and non-long-tail input data, intentions can be more accurately identified and understood.

Using the above technical solution, by constructing the first intention recognition model and the second intention recognition model respectively, and jointly forming a target intention recognition model based on the first intention recognition model and the second intention recognition model, it is possible to target long-tail input data and non-long-tail data. Input data uses different models for intent recognition, which can improve the accuracy and efficiency of intent understanding, thereby improving user experience.

It should be noted that any permutation and combination of the technical features in the above embodiments also belongs to the protection scope of the present application.

Embodiment 3

Figure 4 is a schematic diagram of an intention recognition device provided in Embodiment 4 of the present application. As shown in Figure 4, the device includes: a first sample data acquisition module 410, an abstract generalized entity word acquisition module 420, an intent matching generalized dictionary generation module 430, first intention recognition model building module 440, input data input module to be recognized 450 and intention recognition result output module 460, where:

The first sample data acquisition module 410 is used to obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long tail input sample data and intention matching result sorting data;

The abstract generalized entity word acquisition module 420 is used to perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

The intention matching generalization dictionary generation module 430 is used to logically combine the abstract generalization entity words and the intention matching result sorting data to generate an intention matching generalization dictionary;

The first intention recognition model building module 440 is used to build a first intention recognition model according to the intention matching generalization dictionary;

The input data to be recognized input module 450 is configured to input the input data to be recognized into the first intention recognition model when it is determined that the input data to be recognized is non-long-tail input data;

The intent recognition result output module 460 is configured to output the intent recognition result of the input data to be recognized according to the first intent recognition model.

Optionally, the first sample data acquisition module 410 is specifically configured to: filter the non-long tail input sample data from the original intention sample data according to the non-long tail input data filtering rules; obtain the non-long tail input sample The association intention feedback data of the data is sorted to obtain the intention matching result sorting data.

Optionally, the abstract generalized entity word acquisition module 420 is specifically configured to: perform entity abstraction on the non-long-tail input sample data according to the entity word dictionary to obtain initial abstract entity words; classify and group the initial abstract entity words, Obtain the abstract generalized entity word.

Optionally, the first intention recognition model building module 440 is specifically configured to: build an input data editing distance calculation module according to the dictionary elements of the intention matching generalization dictionary; and build an input data editing distance calculation module according to the input data editing distance calculation module and the intention matching generalization dictionary. dictionary to build the first intention recognition model.

Optionally, the intention recognition device also includes: a preset neural network model pre-training module, used to pre-train the preset neural network model according to the pre-training sample data to obtain a pre-trained neural network model; and obtain the second target intention sample data A module for obtaining second target intention sample data according to the original intention sample data; wherein the second target intention sample data includes long-tail input sample data and intention marking result data; a second intention recognition model acquisition module, using The pre-trained neural network model is trained according to the second target intention sample data to obtain a second intention recognition model; a target intention recognition model construction module is used to train the pre-trained neural network model according to the first intention recognition model and the second intention recognition model. Intent recognition model builds a target intent recognition model.

Optionally, the intention recognition device also includes: an input data acquisition module to be recognized, used to obtain the input data to be recognized; an input data classification module to be recognized, to classify the input data to be recognized, and obtain an input data classification result; Wherein, the input data classification results include long-tail input data and non-long-tail input data.

Optionally, the input data classification module to be identified is specifically configured to: when it is determined that the data length of the input data to be identified is less than or equal to a preset data length threshold, determine that the input data classification result of the input data to be identified is Non-long-tail input data; when it is determined that the data length of the input data to be identified is greater than the preset data length threshold, it is determined that the input data classification result of the input data to be identified is long-tail input data.

The above-mentioned intention recognition device can execute the intention recognition method provided by any embodiment of the present application, and has corresponding functional modules and beneficial effects for executing the method. For technical details that are not described in detail in this embodiment, please refer to the intention identification method provided by any embodiment of this application.

Since the intention recognition device introduced above is a device that can execute the intention recognition method in the embodiment of the present application, based on the intention recognition method introduced in the embodiment of the present application, those skilled in the art can understand the intention recognition in this embodiment. The specific implementation of the device and its various modifications, therefore, how the intention recognition device implements the intention recognition method in the embodiment of the present application will not be described in detail here. As long as a person skilled in the art implements the device used by the intent identification method in the embodiment of the present application, it will fall within the scope of protection of the present application.

Embodiment 4

FIG. 5 shows a schematic structural diagram of an electronic device 10 that can be used to implement embodiments of the present application. Electronic devices are intended to refer to various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (eg, helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are examples only and are not intended to limit the implementation of the present application as described and/or claimed herein.

As shown in Figure 5, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores There is a computer program that can be executed by at least one processor. The processor 11 can perform the operation according to the computer program stored in the read-only memory (ROM) 12 or loaded from the storage unit 18 into the random access memory (RAM) 13. Perform various appropriate actions and processing. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12 and the RAM 13 are connected to each other via the bus 14. An input/output (I/O) interface 15 is also connected to bus 14 .

Multiple components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc. etc.; and communication unit 19, such as network card, modem, wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices through computer networks such as the Internet and/or various telecommunications networks.

Processor 11 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processing processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as the intent recognition method.

In some embodiments, the intent recognition method may be implemented as a computer program, which is tangibly embodied in a computer-readable storage medium, such as the storage unit 18 . In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 10 via the ROM 12 and/or the communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the intent recognition method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the intent identification method in any other suitable manner (eg, by means of firmware).

Various implementations of the systems and techniques described above may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip implemented in a system (SOC), load programmable logic device (CPLD), computer hardware, firmware, software, and/or a combination thereof. These various embodiments may include implementation in one or more computer programs executable and/or interpreted on a programmable system including at least one programmable processor, the programmable processor The processor, which may be a special purpose or general purpose programmable processor, may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device. An output device.

Computer programs for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that the computer program, when executed by the processor, causes the functions/operations specified in the flowcharts and/or block diagrams to be implemented. A computer program may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this application, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, laptop disks, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having a display device (eg, a CRT (cathode ray tube) or LCD (liquid crystal display)) for displaying information to the user monitor); and a keyboard and pointing device (e.g., a mouse or a trackball) through which a user can provide input to the electronic device. Other kinds of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and may be provided in any form, including (acoustic input, voice input or tactile input) to receive input from the user.

The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., A user's computer having a graphical user interface or web browser through which the user can interact with implementations of the systems and technologies described herein), or including such backend components, middleware components, or any combination of front-end components in a computing system. The components of the system may be interconnected by any form or medium of digital data communication (eg, a communications network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

Computing systems may include clients and servers. Clients and servers are generally remote from each other and typically interact over a communications network. The relationship of client and server is created by computer programs running on corresponding computers and having a client-server relationship with each other. The server can be a cloud server, also known as cloud computing server or cloud host. It is a host product in the cloud computing service system to solve the problems of difficult management and weak business scalability in traditional physical hosts and VPS services. defect.

Embodiment 5

Embodiment 5 of the present application also provides a computer storage medium that stores a computer program. The computer-readable storage medium may be non-volatile or volatile. The computer program is used to execute the present invention when executed by a computer processor. Apply the intention recognition method described in any of the above embodiments.

Wherein, the intention identification method includes: obtaining the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long tail input sample data and intention matching result sorting data; Input sample data for entity abstraction to obtain abstract generalized entity words; logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalization dictionary; match the generalization dictionary according to the intent Construct a first intention recognition model; when it is determined that the input data to be recognized is non-long tail input data, input the input data to be recognized into the first intention recognition model; output according to the first intention recognition model The intention recognition result of the input data to be recognized.

The computer storage medium in the embodiment of the present application may be any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard drives, random access memory (RAM), read only memory (Read Only Memory) , ROM), Erasable Programmable Read Only Memory (Erasable Programmable Read Only Memory, EPROM, or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or the above Any suitable combination. As used herein, a computer-readable storage medium may be any tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave carrying computer-readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device .

Program code embodied on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, radio frequency (Radio Frequency, RF), etc., or any suitable combination of the above.

Computer program code for performing the operations of the present application may be written in one or more programming languages, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional A procedural programming language, such as the "C" language or similar programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In situations involving remote computers, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer, such as through the Internet using an Internet service provider. ).

It should be understood that various forms of the process shown above may be used, with steps reordered, added or deleted. For example, each step described in this application can be executed in parallel, sequentially, or in a different order. As long as the desired results of the technical solution of this application can be achieved, there is no limitation here.

The above-mentioned specific embodiments do not constitute a limitation on the scope of protection of the present application. It will be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions are possible depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application shall be included in the protection scope of this application.

Claims

An intent identification method, including:

Obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result ranking data;

Perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

Logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalized dictionary;

Construct a first intention recognition model according to the intention matching generalization dictionary;

When it is determined that the input data to be recognized is non-long-tail input data, input the input data to be recognized into the first intention recognition model;

The intent recognition result of the input data to be recognized is output according to the first intent recognition model.
The method according to claim 1, wherein said obtaining the first target intention sample data according to the original intention sample data includes:

Filter the non-long-tail input sample data from the original intention sample data according to the non-long-tail input data filtering rules;

Obtain the association intention feedback data of the non-long-tail input sample data;

The associated intention feedback data is sorted to obtain the intention matching result sorting data.
The method according to claim 1, wherein said performing entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words includes:

Perform entity abstraction on the non-long-tail input sample data according to the entity word dictionary to obtain initial abstract entity words;

Classify and group the initial abstract entity words to obtain the abstract generalized entity words.
The method according to claim 1, wherein said constructing a first intention recognition model according to the intention matching generalization dictionary includes:

Construct an input data edit distance calculation module according to the dictionary elements of the generalized dictionary matched with the intention;

The first intention recognition model is constructed according to the input data editing distance calculation module and the intention matching generalization dictionary.
The method of claim 1, further comprising:

Pre-train the preset neural network model based on the pre-training sample data to obtain the pre-trained neural network model;

Obtain second target intention sample data according to the original intention sample data; wherein the second target intention sample data includes long-tail input sample data and intention mark result data;

Train the pre-trained neural network model according to the second target intention sample data to obtain a second intention recognition model;

A target intention recognition model is constructed according to the first intention recognition model and the second intention recognition model.
The method according to any one of claims 1 to 5, wherein before inputting the input data to be recognized into the first intention recognition model, it further includes:

Get the input data to be recognized;

The input data to be identified is classified to obtain an input data classification result; wherein the input data classification result includes long-tail input data and non-long-tail input data.
The method according to claim 6, wherein classifying the input data to be identified includes:

When it is determined that the data length of the input data to be identified is less than or equal to the preset data length threshold, determine that the input data classification result of the input data to be identified is non-long tail input data;

When it is determined that the data length of the input data to be identified is greater than the preset data length threshold, it is determined that the input data classification result of the input data to be identified is long-tail input data.
An intention recognition device, which includes:

The first sample data acquisition module is used to obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result sorting data;

An abstract generalized entity word acquisition module is used to perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

An intent matching generalization dictionary generation module, used to logically combine the abstract generalization entity words and the intent matching result sorting data to generate an intent matching generalization dictionary;

A first intention recognition model building module, configured to build a first intention recognition model according to the intention matching generalization dictionary;

An input data input module to be recognized, configured to input the input data to be recognized into the first intention recognition model when it is determined that the input data to be recognized is non-long-tail input data;

An intent recognition result output module is configured to output the intent recognition result of the input data to be recognized according to the first intent recognition model.
An electronic device, wherein the electronic device includes:

at least one processor; and

A memory communicatively connected with the at least one processor; wherein the memory stores a computer program that can be executed by the at least one processor, and when the computer program is executed by the at least one processor, the intention recognition method is implemented, include:

Obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result ranking data;

Perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

Logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalized dictionary;

Construct a first intention recognition model according to the intention matching generalization dictionary;

When it is determined that the input data to be recognized is non-long-tail input data, input the input data to be recognized into the first intention recognition model;

The intent recognition result of the input data to be recognized is output according to the first intent recognition model.
The electronic device according to claim 9, wherein the obtaining the first target intention sample data according to the original intention sample data includes:

Filter the non-long-tail input sample data from the original intention sample data according to the non-long-tail input data filtering rules;

Obtain the association intention feedback data of the non-long-tail input sample data;

The associated intention feedback data is sorted to obtain the intention matching result sorting data.
The electronic device according to claim 9, wherein said performing entity abstraction on said non-long-tail input sample data to obtain abstract generalized entity words includes:

Perform entity abstraction on the non-long-tail input sample data according to the entity word dictionary to obtain initial abstract entity words;

Classify and group the initial abstract entity words to obtain the abstract generalized entity words.
The electronic device according to claim 9, wherein said constructing a first intention recognition model according to the intention matching generalization dictionary includes:

Construct an input data edit distance calculation module according to the dictionary elements of the generalized dictionary matched with the intention;

The first intention recognition model is constructed according to the input data editing distance calculation module and the intention matching generalization dictionary.
The electronic device according to claim 9, further comprising:

Pre-train the preset neural network model based on the pre-training sample data to obtain the pre-trained neural network model;

Obtain second target intention sample data according to the original intention sample data; wherein the second target intention sample data includes long-tail input sample data and intention mark result data;

Train the pre-trained neural network model according to the second target intention sample data to obtain a second intention recognition model;

A target intention recognition model is constructed according to the first intention recognition model and the second intention recognition model.
The electronic device according to any one of claims 9 to 13, wherein before inputting the input data to be recognized into the first intention recognition model, it further includes:

Get the input data to be recognized;

Classify the input data to be identified to obtain input data classification results; wherein the input data classification results include long-tail input data and non-long-tail input data;

Classifying the input data to be identified includes:

When it is determined that the data length of the input data to be identified is less than or equal to the preset data length threshold, determine that the input data classification result of the input data to be identified is non-long tail input data;

When it is determined that the data length of the input data to be identified is greater than the preset data length threshold, it is determined that the input data classification result of the input data to be identified is long-tail input data.
A computer storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to implement an intention recognition method when executed by a processor, including:

Obtain the first target intention sample data according to the original intention sample data; wherein the first target intention sample data includes non-long-tail input sample data and intention matching result ranking data;

Perform entity abstraction on the non-long-tail input sample data to obtain abstract generalized entity words;

Logically combine the abstract generalized entity words and the intent matching result sorting data to generate an intent matching generalized dictionary;

Construct a first intention recognition model according to the intention matching generalization dictionary;

When it is determined that the input data to be recognized is non-long-tail input data, input the input data to be recognized into the first intention recognition model;

The intent recognition result of the input data to be recognized is output according to the first intent recognition model.
The computer storage medium according to claim 15, wherein the obtaining the first target intention sample data according to the original intention sample data includes:

Filter the non-long-tail input sample data from the original intention sample data according to the non-long-tail input data filtering rules;

Obtain the association intention feedback data of the non-long-tail input sample data;

The associated intention feedback data is sorted to obtain the intention matching result sorting data.
The computer storage medium according to claim 15, wherein said performing entity abstraction on said non-long-tail input sample data to obtain abstract generalized entity words includes:

Perform entity abstraction on the non-long-tail input sample data according to the entity word dictionary to obtain initial abstract entity words;

Classify and group the initial abstract entity words to obtain the abstract generalized entity words.
The computer storage medium according to claim 15, wherein said constructing a first intention recognition model according to the intention matching generalization dictionary includes:

Construct an input data edit distance calculation module according to the dictionary elements of the generalized dictionary matched with the intention;

The first intention recognition model is constructed according to the input data editing distance calculation module and the intention matching generalization dictionary.
The computer storage medium of claim 15, further comprising:

Pre-train the preset neural network model based on the pre-training sample data to obtain the pre-trained neural network model;

Obtain second target intention sample data according to the original intention sample data; wherein the second target intention sample data includes long-tail input sample data and intention mark result data;

Train the pre-trained neural network model according to the second target intention sample data to obtain a second intention recognition model;

A target intention recognition model is constructed according to the first intention recognition model and the second intention recognition model.
The computer storage medium according to any one of claims 16 to 19, wherein before inputting the input data to be recognized into the first intention recognition model, it further includes:

Get the input data to be recognized;

Classify the input data to be identified to obtain input data classification results; wherein the input data classification results include long-tail input data and non-long-tail input data;

Classifying the input data to be identified includes:

When it is determined that the data length of the input data to be identified is less than or equal to the preset data length threshold, it is determined that the input data classification result of the input data to be identified is non-long tail input data;

When it is determined that the data length of the input data to be identified is greater than the preset data length threshold, it is determined that the input data classification result of the input data to be identified is long-tail input data.