CN116956068A

CN116956068A - Intention recognition method and device based on rule engine, electronic equipment and medium

Info

Publication number: CN116956068A
Application number: CN202310931861.0A
Authority: CN
Inventors: 田晓姣
Original assignee: Ping An Life Insurance Company of China Ltd
Current assignee: Ping An Life Insurance Company of China Ltd
Priority date: 2023-07-26
Filing date: 2023-07-26
Publication date: 2023-10-27

Abstract

The embodiment of the application provides an intention recognition method, an intention recognition device, electronic equipment and a medium based on a rule engine, and belongs to the technical field of financial science and technology. The method comprises the following steps: for each matching template in the rule engine, carrying out priority setting on the matching template to obtain a priority value corresponding to the matching template; the priority ranking is carried out on the plurality of matching templates according to the priority value, and a priority sequence is obtained; acquiring corpus information of a user, and inputting the corpus information into a rule engine for corpus matching; when the corpus information is matched with at least two matching templates in the rule engine, screening the matching templates corresponding to the corpus information according to the priority sequence, and determining a target matching template; and inputting the corpus information into a target matching template for intention recognition, and outputting target information corresponding to the corpus information. According to the method and the device for identifying the intention, the accuracy of the intention identification can be improved under the condition that the user hits a plurality of intentions at the same time.

Description

Intention recognition method and device based on rule engine, electronic equipment and medium

Technical Field

The present application relates to the technical field of financial science and technology, and in particular, to a rule engine-based intent recognition method, device, electronic apparatus, and medium.

Background

The intention recognition is one of the most important branches of artificial intelligence in the field of classification recognition, and is widely applied to multiple business fields such as intelligent conversation robots, intelligent customer analysis operation and the like. Such as product promotions, customer service, marketing campaign promotions, investment and trade opinions, customer communications, and the like, in an insurance scenario. The existing intention recognition in the industry or products is generally completed cooperatively by combining various types of AI models, and the implementation of the whole process in the traditional mode firstly requires technicians to scoop a large number of records of customer and robot dialogues from a production environment according to the requirements of business and algorithm. Then, manual data labeling, namely, a professional data labeling person recognizes the speech content and emotion of the speaking of the client to be matched and labeled with each classification intention by listening to the call record one by one. For example, annotating a endowment insurance that the customer wants to learn, an automobile insurance, or a child care insurance, etc., and finally, to ensure the recognition effect of the rule engine, a certain number of production customer speaking instances need to be accumulated for each intention classification.

However, the conventional method not only consumes a great deal of time and manpower resources to complete the work of extracting, labeling and counting data, has low work efficiency, cannot guarantee the effect of improving the intention recognition effect in a hundred percent, but also cannot accurately judge a plurality of intents of the user in the process that the user intends to know a plurality of financial products or investment products, thereby causing inaccurate intention recognition result and low efficiency.

Disclosure of Invention

The embodiment of the application mainly aims to provide an intention recognition method, an intention recognition device, electronic equipment and a medium based on a rule engine, which can improve the accuracy of intention recognition under the condition that a user hits a plurality of intentions at the same time.

To achieve the above object, a first aspect of an embodiment of the present application proposes an intent recognition method based on a rule engine, the rule engine including a plurality of matching templates, the method including:

for each matching template in the rule engine, carrying out priority setting on the matching templates to obtain priority values corresponding to the matching templates, wherein the corpus in any two matching templates is not identical;

the priority ranking is carried out on the plurality of matching templates according to the priority value, so that a priority sequence is obtained;

acquiring corpus information of a user, and inputting the corpus information into the rule engine for corpus matching;

when the corpus information is matched with at least two matching templates in the rule engine, screening the matching templates corresponding to the corpus information according to the priority sequence, and determining a target matching template;

And inputting the corpus information into the target matching template for intention recognition, and outputting target information corresponding to the corpus information.

In some embodiments, after the inputting the corpus information into the rule engine for corpus matching, the method further includes:

when the corpus information is not matched with all the matched templates, inputting the corpus information into a preset recall model for intention recognition to obtain intention recall information;

performing similarity calculation on the intention recall information and the corpus information to obtain a first similarity score;

and when the first similarity score is determined to be greater than or equal to a preset first threshold value, selecting intention recall information corresponding to the first similarity score as first intention information.

In some embodiments, after the performing similarity calculation on the intent recall information and the corpus information to obtain a first similarity score, the method further includes:

when the first similarity score is smaller than the first threshold, inputting the corpus information into a preset spam model to perform intention recognition, so as to obtain a plurality of intention spam information;

performing similarity calculation on all the intention spam information and the corpus information to obtain a second similarity score;

When the second similarity score is larger than or equal to a preset second threshold value, ordering the plurality of intention spam messages corresponding to the second similarity score in a descending order to obtain a descending order sequence;

and determining second intention information corresponding to the corpus information according to the descending order sequence.

In some embodiments, after the calculating the similarity between all the intended spam information and the corpus information to obtain a second similarity score, the method further includes:

and generating identification error information when the second similarity score is smaller than the second threshold value.

In some embodiments, the target matching templates include preconfigured corpus matching templates, artificial templates, log automatic extraction templates, and corpus automatic extraction templates; inputting the corpus information into the target matching template for intention recognition, and outputting target information corresponding to the corpus information, wherein the method comprises the following steps:

under the condition that the target matching template is the corpus matching template, matching the corpus information with pre-labeled training corpus in the corpus matching template, and selecting the training corpus corresponding to the corpus information as target information;

under the condition that the target matching template is the artificial template, matching the corpus information with a specific corpus predefined in the artificial template, and selecting the specific corpus corresponding to the corpus information as target information;

Under the condition that the target matching template is the automatic log extraction template, inputting the corpus information into a preset reverse intention recognition model for similarity calculation, and outputting target information corresponding to the corpus information;

and under the condition that the target matching template is the automatic corpus extraction template, inputting the corpus information into the automatic corpus extraction template for corpus extraction to obtain extraction information, matching the extraction information with a preset template corpus, and selecting the template corpus corresponding to the corpus information as target information.

In some embodiments, inputting the corpus information into a preset reverse intention recognition model to perform similarity calculation, and outputting target information corresponding to the corpus information includes:

inputting the corpus information into the reverse intention recognition model so that the reverse intention recognition model carries out log mining on the corpus information to obtain log mining corpus;

calculating the similarity between the log mining corpus and a preset log corpus set to obtain a third similarity score;

when the third similarity score is larger than or equal to a third threshold value, generating a plurality of target log corpus according to the log mining corpus;

And performing word frequency statistics on all the target log corpus to obtain the target information.

In some embodiments, the performing word frequency statistics on all the target log corpora to obtain the target information includes:

word segmentation is carried out on all the target log corpus to obtain a plurality of log keywords;

performing word frequency probability calculation on a plurality of log keywords to obtain occurrence probabilities corresponding to the log keywords;

sorting the occurrence probabilities in a descending order to obtain a log sequence;

and carrying out log screening on the log sequence according to a preset extraction rule to obtain the target information.

To achieve the above object, a second aspect of an embodiment of the present application proposes an intention recognition apparatus based on a rule engine, the rule engine including a plurality of matching templates, the apparatus comprising:

the priority setting module is used for setting the priority of each matching template in the rule engine to obtain a priority value corresponding to the matching template, wherein the corpus in any two matching templates is not identical;

the priority ranking module is used for ranking the priorities of the plurality of matching templates according to the priority value to obtain a priority sequence;

The corpus matching module is used for acquiring corpus information of the user and inputting the corpus information into the rule engine for corpus matching;

the template screening module is used for screening the matching templates corresponding to the corpus information according to the priority sequence when the corpus information is matched with at least two matching templates in the rule engine, and determining a target matching template;

the intention recognition module is used for inputting the corpus information into the target matching template to perform intention recognition and outputting target information corresponding to the corpus information.

To achieve the above object, a third aspect of the embodiments of the present application proposes an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, the processor implementing the rule engine-based intent recognition method as described in the first aspect when executing the computer program.

To achieve the above object, a fourth aspect of the embodiments of the present application proposes a computer-readable storage medium storing a computer program which, when executed by a processor, implements the rule engine-based intent recognition method as described in the first aspect.

According to the rule engine-based intention recognition method, device, electronic equipment and storage medium, firstly, priority setting is conducted on all matching templates in a rule engine to obtain a priority sequence of each matching template, then corpus information of a user is obtained, the corpus information is input into the rule engine to conduct template matching, the matching templates corresponding to the corpus information are confirmed, preliminary matching of the corpus information is achieved, under the condition that corpus information is matched with at least two matching templates in the rule engine, screening is conducted on the matching templates corresponding to the corpus information according to the priority sequence, and a target matching template with the highest priority is output, so that accuracy of intention recognition can be improved under the condition that the user hits a plurality of intents simultaneously, finally, the corpus information is input into the target matching template to conduct intention recognition, and target information corresponding to the corpus information is output, and therefore efficiency and accuracy of intention recognition are improved.

Drawings

FIG. 1 is a flow chart of a rule engine based intent recognition method provided by an embodiment of the present application;

FIG. 2 is another flow chart of a rule engine based intent recognition method provided by an embodiment of the present application;

FIG. 3 is another flow chart of a rule engine based intent recognition method provided by an embodiment of the present application;

FIG. 4 is another flow chart of a rule engine based intent recognition method provided by an embodiment of the present application;

fig. 5 is a flowchart of step S105 in fig. 1;

fig. 6 is a flowchart of step S504 in fig. 5;

FIG. 7 is a schematic diagram of a rule engine based intent recognition device according to an embodiment of the present application;

fig. 8 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

It should be noted that although functional block division is performed in a device diagram and a logic sequence is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the block division in the device, or in the flowchart. The terms first, second and the like in the description and in the claims and in the above-described figures, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

First, several nouns involved in the present application are parsed:

natural language processing (Natural Language Processing, NLP): NLP is a branch of artificial intelligence that is a interdisciplinary of computer science and linguistics, and is often referred to as computational linguistics, and is processed, understood, and applied to human languages (e.g., chinese, english, etc.). Natural language processing includes parsing, semantic analysis, chapter understanding, and the like. Natural language processing is commonly used in the technical fields of machine translation, handwriting and print character recognition, voice recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining, and the like, and relates to data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, linguistic research related to language calculation, and the like.

Inverted index recall model (Elastic Search Recall, ES): the ES recall model processes large-scale text data by using an inverted index technology, and the inverted index recall model rapidly retrieves related documents by establishing the relation between the documents and words, and has the function of performing efficient information retrieval and recall in the large-scale text data. Is widely used in search engines, recommendation systems, question and answer systems, and other applications.

Multisource semantic recall model (Multi-Source Semantic Retrieval Model): the multi-source semantic recall refers to multi-mode semantic recall in which a plurality of scene information is fused, such as image-introduced graph semantic recall, and image-introduced graph semantic recall is only performed. These context information amounts to providing assistance information to the model, which is particularly important in some contexts where the query input is less semantically ambiguous (e.g., sug in the search field) because it provides a richer input representation. For example, in the e-commerce field, mapping and expressing pictures as vectors; in the map field, the geographic position of the user is expressed and acted as the input of the user; in the food field, the search history of the user is constructed only by a map, and the map is used as the input of the user, so that the recall effect of thousands of people and thousands of faces can be achieved.

Enhanced sequential inference model (Enhanced Sequential Inference Model, ESIM): the ESIM model is a natural language processing model based on deep learning and is mainly used for text matching tasks. The method learns text semantic information by utilizing an LSTM (Long Short-Term Memory network) neural network, aligns the texts by using a attention mechanism and finally outputs a matching score between the texts.

As a general text matching model, the ESIM model can be applied to various natural language processing tasks, such as text classification, sentence similarity calculation, question-answering system, and the like.

Based on the above, the embodiment of the application provides a rule engine-based intention recognition method and device, electronic equipment and storage medium, which can improve the accuracy of intention recognition under the condition that a user hits a plurality of intents at the same time.

The rule engine-based intention recognition method and device, the electronic device and the storage medium provided by the embodiment of the application are specifically described through the following embodiments, and the rule engine-based intention recognition method in the embodiment of the application is described first.

The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.

Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligent software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a module management technology of an online guest receiving system, a natural language processing technology, machine learning/deep learning and other directions.

The embodiment of the application provides an intention recognition method based on a rule engine, and relates to the technical field of data processing. The intent recognition method based on the rule engine provided by the embodiment of the application can be applied to a terminal, a server and software running in the terminal or the server. In some embodiments, the terminal may be a smart phone, tablet, notebook, desktop, etc.; the server side can be configured as an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like; the software may be an application or the like that implements a rule engine-based intent recognition method, but is not limited to the above form.

The application is operational with numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In the embodiments of the present application, when related processing is performed according to user information, user behavior data, user history data, user location information, and other data related to user identity or characteristics, permission or consent of the user is obtained first, and the collection, use, processing, and the like of the data comply with related laws and regulations and standards of related countries and regions. In addition, when the embodiment of the application needs to acquire the sensitive personal information of the user, the independent permission or independent consent of the user is acquired through popup or jump to a confirmation page and the like, and after the independent permission or independent consent of the user is definitely acquired, the necessary relevant data of the user for enabling the embodiment of the application to normally operate is acquired.

However, the conventional method not only consumes a lot of time and manpower resources to complete the task of extracting, labeling and counting data, for example, the implementation of the whole process in the conventional manner first requires technicians to scoop up a lot of records of clients and robot dialogues from the production environment according to the requirements of the business and the algorithm. The "requirements" herein, i.e., filter conditions such as call turns (note: the robot continuously speaks/asks for a session and the customer's answer, such "ask for answer" constitutes a round of dialogue), customer answer duration, call end node, etc. This step typically requires 3 people, including business 1 person, algorithm 1 person, data/background development 1 person.

Then, manual data labeling, namely, a professional data labeling person recognizes the speech content and emotion of the speaking of the client to be matched and labeled with each classification intention by listening to the call record one by one. For example, a customer may want to give care, car insurance, etc., and this step may be about 50 records/day for a skilled data annotator.

Finally, to guarantee the recognition effect of the rules engine, a certain number of production client speech instances need to be accumulated for each intent classification, typically at least 2000 intent classifications are needed. Of course, the greater the number of instances of client utterances, the wider the expression of the overlaid client utterances, and the higher the client response duty cycle that the rules engine can hit.

However, it is realistic that more than 80% of customer responses in a production environment are concentrated on less than 20% of the main intent, resulting in other 80% intent classifications that may be evaluated for poor accuracy in intent recognition due to severe shortages of customer speech instances. From the above analysis, it is not difficult to find that the conventional method not only consumes a great deal of time and manpower resources to complete the work of extracting, labeling and counting data, has low working efficiency and cannot guarantee the effect of improving the intention recognition effect in a hundred percent, but also cannot accurately judge a plurality of intentions of the user in the process that the user intends to know a plurality of financial products or investment products, thereby causing inaccurate intention recognition result and low efficiency.

In order to solve the above problems, the present embodiment provides an intent recognition method, an apparatus, an electronic device, and a storage medium based on a rule engine, firstly, priority setting is performed on all matching templates in the rule engine to obtain a priority sequence of each matching template, then corpus information of a user is obtained, the corpus information is input into the rule engine to perform template matching, a matching template corresponding to the corpus information is determined, preliminary matching of the corpus information is realized, under the condition that corpus information is matched with at least two matching templates in the rule engine, the matching template corresponding to the corpus information is screened according to the priority sequence, and a target matching template with the highest priority is output, so that accuracy of intent recognition can be improved under the condition that a user hits multiple intents at the same time, finally, the corpus information is input into the target matching template to perform intent recognition, and target information corresponding to the corpus information is output, so that efficiency and accuracy of intent recognition are improved.

The following is a detailed description with reference to the accompanying drawings.

Fig. 1 is an optional flowchart of a rule engine-based intent recognition method provided in an embodiment of the present application, where the method in fig. 1 may include, but is not limited to, steps S101 to S105.

It should be noted that the rule engine includes a plurality of matching templates.

Step S101, for each matching template in the rule engine, setting the priority of the matching template to obtain a priority value corresponding to the matching template;

it should be noted that the corpus in any two matching templates is not exactly the same.

In step S101 of some embodiments, the rule engine includes a plurality of matching templates, and at least a portion of corpus sources in any two matching templates are different, so that each matching template carries a respective keyword.

It can be appreciated that there are a plurality of keywords in different matching templates, and at least one keyword in each matching template is unique, for example, the keyword of the A template is endowment insurance, insurance lecture and financial training; the keywords of the B template are automobile insurance and market analysis; the key words of the template C are financial training and wage calculation; the keywords of the D template are retirement, social security, public accumulation, etc., and the embodiment is not particularly limited.

Step S102, priority ordering is carried out on a plurality of matching templates according to the priority value, and a priority sequence is obtained;

in step S102 of some embodiments, the matching templates are prioritized according to the plurality of priority values obtained in step S101, for example, the priority sequence is obtained by increasing the priority or decreasing the priority, so as to facilitate the selection of the matching templates in case of hit of a plurality of intents.

Step S103, obtaining the corpus information of the user, and inputting the corpus information into a rule engine for corpus matching;

in step S103 of some embodiments, corpus information of a user is obtained, where the corpus information of the user carries intent keywords of the user, and each matching template in the rule engine is provided with a template keyword, the corpus information is input into the rule engine, so that the rule engine extracts the intent keywords in the corpus information, and then the intent keywords are matched with all the template keywords in the rule engine, thereby completing corpus matching of the corpus information, accurately judging corpus types of the corpus information, and improving accuracy of corpus information prediction.

It should be noted that, the principle of corpus matching by using the rule engine is to configure various common client-phone templates according to the rule of the regular expression, for example, "how to select the pension? "," financial risk of financial products? ", to match to the corresponding intent classification, i.e., to the matching template of the rules engine, by a canonical lookup function. Therefore, the intention recognition accuracy of this approach can reach 100%.

Step S104, when the corpus information is matched with at least two matching templates in the rule engine, screening the matching templates corresponding to the corpus information according to the priority sequence to determine a target matching template;

in step S104 of some embodiments, when the corpus information is matched with at least two matching templates in the rule engines, it is explained that the corpus information hits a matching template in a plurality of rule engines, the matching template corresponding to the corpus information is filtered according to the priority sequence, the matching template with the largest priority value is selected as a target matching template, and the intention classification corresponding to the highest priority value is output, so that accuracy of intention recognition can be improved under the condition that the user hits a plurality of intents at the same time, and autonomous optimization of the rule engines is realized.

Step S105, inputting the corpus information into a target matching template for intention recognition, and outputting target information corresponding to the corpus information.

In step S105 of some embodiments, the corpus information is input into a determined target matching template, so that the target matching template automatically filters and parses the corpus information, and outputs target information corresponding to the corpus information, thereby improving the effect of identifying intent.

Referring to fig. 2, fig. 2 is another optional flowchart of a rule engine-based intent recognition method according to an embodiment of the present application, where the method in fig. 2 may include, but is not limited to, steps S201 to S203.

Step S201, when the corpus information is not matched with all the matched templates, inputting the corpus information into a preset recall model for intention recognition to obtain intention recall information;

in step S201 of some embodiments, when the corpus information is not matched with all the matching templates, it is indicated that the corpus information does not hit any matching template, and the corpus information needs to be input into a preset recall model for intention recognition, so as to obtain intention recall information, thereby realizing multiple judgment on the corpus information and improving accuracy of matching the treated corpus information.

It should be noted that, the preset recall model may be an inverted index recall (Elastic Search Recall, ES) model, a multi-source semantic recall model, and the like, which is not limited in this embodiment.

Step S202, similarity calculation is carried out on intention recall information and corpus information, and a first similarity score is obtained;

in step S202 of some embodiments, similarity calculation is performed on the intent recall information and the corpus information to obtain a plurality of similarity scores, the plurality of similarity scores are ranked, and a score with the largest similarity score is selected as a first similarity score, so as to improve accuracy in identifying the corpus information.

In the process of calculating the similarity of the material information, the similarity of the query sentence is calculated by adopting a TF-IDF weighting algorithm, and the specific process is as follows:

firstly, for each word in the corpus information, calculating word Frequency (TF) of the word in the intent recall information, namely the number of times the word appears in the corpus information;

second, the inverse document frequency (Inverse Document Frequency, IDF) is calculated, representing the importance of the word in the entire document collection. The larger the IDF, the more important the word. The calculation formula of IDF is log (N/N), where N is the total number of documents and N is the number of documents containing the word;

thirdly, multiplying TF and IDF to obtain TF-IDF value of the word;

fourthly, weighting and summing TF-IDF values of each word for the whole corpus information to obtain vector representation of the corpus information;

and fifthly, calculating cosine similarity of the corpus information vector and each intention recall information, and obtaining similarity ranking.

In step S203, when it is determined that the first similarity score is greater than or equal to a preset first threshold, the intention recall information corresponding to the first similarity score is selected as the first intention information.

In step S203 of some embodiments, when it is determined that the first similarity score is greater than or equal to a preset first threshold, it is indicated that the corpus information hits the intent in the recall model, and the intent recall information corresponding to the first similarity score may be directly selected as the first intent information, so that the efficiency of intent recognition is improved.

Referring to fig. 3, fig. 3 is another optional flowchart of a rule engine-based intent recognition method according to an embodiment of the present application, where the method in fig. 3 may include, but is not limited to, steps S301 to S304.

Step S301, when it is determined that the first similarity score is smaller than a first threshold, inputting corpus information into a preset spam model for intention recognition to obtain a plurality of intention spam information;

in step S301 of some embodiments, when it is determined that the first similarity score is smaller than the first threshold, it is indicated that the corpus information does not hit the intent in the recall model, and the corpus information needs to be input into the spam model for performing intent recognition to obtain a plurality of pieces of intent spam information, where the spam model in this embodiment is a fasttex spam model, and is used for performing spam processing by using the fasttex model when the ES recall model cannot find a corresponding document, so as to return the document related to the corpus information as much as possible.

Step S302, performing similarity calculation on all the intention spam information and the corpus information to obtain a second similarity score;

in step S302 of some embodiments, similarity calculation is performed on all the spam intention information and the corpus information to obtain a plurality of similarity scores, the similarity scores are ranked, and the score with the largest similarity score is selected as the second similarity score, so as to improve accuracy in identifying the corpus information.

It should be noted that, since the fastttext bottom-of-pocket model is generally not suitable for long text matching, in the process of performing similarity calculation on the intended bottom-of-pocket information and the corpus information, the corpus information needs to be semantically segmented and then the subsequent similarity calculation is performed, where the embodiment in step S202 in the process of similarity calculation is consistent, and this embodiment is not repeated.

Step S303, when the second similarity score is greater than or equal to a preset second threshold, ordering the plurality of intention spam messages corresponding to the second similarity score in a descending order to obtain a descending order sequence;

in step S303 of some embodiments, when the second similarity score is greater than or equal to a preset second threshold, it is indicated that the corpus information hits the intention in the spam model, and the plurality of intention spam information corresponding to the second similarity score needs to be sorted in a descending order to obtain a descending order sequence, so that the situation of recognition of the intention spam information can be reflected, and the accuracy of recognizing the corpus information is improved from the whole.

Step S304, determining second intention information corresponding to the corpus information according to the descending order sequence.

In step S304 of some embodiments, a second similarity score of the sequence header in the descending sequence is selected, the intention spam corresponding to the second similarity score is used as second intention information corresponding to the corpus information, and the accuracy of identifying the corpus information is improved from the whole by performing operation flows such as similarity calculation on the intention spam and the corpus information, and sequencing the second similarity score.

Referring to fig. 4, fig. 4 is another optional flowchart of a rule engine-based intent recognition method according to an embodiment of the present application, where the method in fig. 4 may include, but is not limited to, step S401.

In step S401, when the second similarity score is smaller than the second threshold, identification error information is generated.

In step S401 of some embodiments, when the second similarity score is less than the second threshold, indicating that the corpus information does not hit the intention in the spam model, the generated recognition error information is directly returned to the user, so as to facilitate the subsequent operation of the user.

It should be noted that the identification error information may be an error page, an error text, or the like, and the embodiment is not particularly limited.

Note that, the NLP intention recognition accuracy=rule engine recognition accuracy+es recall model recognition accuracy+fasttet bottom pocket model recognition accuracy+0%. The sum of the hit ratios of the branches is 100%, and if the rule engine hit ratio is increased, the hit ratio of the other branches is reduced. That is, the hit ratio of 100% accurate recognition is increased, and the hit ratio of the rest, namely, ES recall+fasttet bottom+unrecognizable case is reduced.

In some embodiments, the target matching templates include a preconfigured corpus matching template, an artificial template, an automatic log extraction template and an automatic corpus extraction template, wherein the corpus matching template is a strictly configured complete regular matching template, and includes a pre-labeled training corpus, which can be derived from a model training set corpus labeled by artificial data;

the manual template is derived from manually mining production logs and comprises predefined specific corpus, wherein the predefined specific corpus is derived from specific corpus compiled by business personnel and technicians, such as endowment insurance, automobile insurance, accumulation fund and the like;

the automatic log extraction template is derived from a timing script and used for automatically mining production logs and comprises a reverse consciousness recognition model, and the similarity of log mining linguistic data and the existing linguistic data is calculated through the reverse intention recognition model;

the automatic corpus extraction template is derived from automatic mining of timing scripts to produce corpus, and comprises preset template corpus, wherein the template corpus is obtained from open public corpus of an external network in a crawling mode, and the embodiment is not particularly limited.

It should be noted that, in this embodiment, the priority sequences of the four templates may be set according to the needs of the user, for example, the output priorities are, from high to low, a corpus matching template, an artificial template, an automatic log extraction template, and an automatic corpus extraction template in sequence; or the output priority is from high to low in sequence, namely a corpus matching template, a log automatic extraction template, an artificial template and a corpus automatic extraction template, and the specific priority sequence is not particularly limited in the embodiment.

In some embodiments, inputting the corpus information into a target matching template for intention recognition, outputting target information corresponding to the corpus information, including:

under the condition that the target matching template is a corpus matching template, directly matching the corpus information with the pre-labeled training corpus in the corpus matching template, and selecting the training corpus corresponding to the corpus information as the target information, wherein the corpus information is completely matched with the pre-labeled training corpus.

Under the condition that the target matching template is an artificial template, matching the corpus information with a specific corpus predefined in the artificial template, and selecting the specific corpus corresponding to the corpus information as target information;

under the condition that the target matching template is a log automatic extraction template, inputting corpus information into a preset reverse intention recognition model for similarity calculation, calculating the similarity of log mining corpus and existing corpus, and outputting target information corresponding to the corpus information;

under the condition that the target matching template is an automatic corpus extraction template, inputting corpus information into the automatic corpus extraction template to carry out corpus extraction to obtain extraction information, matching the extraction information with a preset template corpus, and selecting the template corpus corresponding to the corpus information as the target information.

In some embodiments, the method is used for realizing the matching of the language information, improving the matching precision of the language information, improving the recognition efficiency of the language information, avoiding the situation that the language information cannot be matched, realizing the automatic filtering, analysis, classification, template extraction expansion and the like of production data (expected by a customer speaking operation) by utilizing big data intelligent analysis and machine learning technology, establishing an ecological circle for self-learning and growth of a rule engine template, and greatly saving time and labor cost.

Referring to fig. 5, in some embodiments, step S105 may further include, but is not limited to, steps S501 to S504:

step S501, inputting corpus information into a reverse intention recognition model so that the reverse intention recognition model carries out log mining on the corpus information to obtain log mining corpus;

step S502, calculating the similarity between the log mining corpus and a preset log corpus set to obtain a third similarity score;

step S503, when the third similarity score is larger than or equal to a third threshold value, generating a plurality of target log corpus according to the log mining corpus;

step S504, word frequency statistics is carried out on all target log corpora to obtain target information.

In steps S501 to S504 of some embodiments, the corpus information is input into the reverse intention recognition model, so that the reverse intention recognition model performs log mining on the corpus information to obtain log mining corpus in the corpus information, then, the similarity between the log mining corpus and a preset log corpus set is calculated to obtain a third similarity score, when the third similarity score is greater than or equal to a third threshold, a plurality of target log corpuses are directly generated according to the log mining corpus, finally, word frequency statistics is performed on all the target log corpuses to obtain target information corresponding to the corpus information, log extraction on the corpus information is realized, and accuracy of the language recognition is improved.

In the process of calculating the similarity between the log-mining corpus and the log corpus set, the similarity between any one of the log-mining corpus and the log corpus set can be calculated by using the ESIM, wherein the third threshold is a ground line standard for judging the similarity between the log-mining corpus and the log corpus set, and if the third similarity score calculated by the ESIM model is greater than or equal to the third threshold, the log-mining corpus can be added to the log corpus set, so that the log-mining corpus is added to the log corpus set under the condition that the third similarity score is greater than or equal to the third threshold, and the update of the automatic log extraction template is further realized.

Referring to fig. 6, in some embodiments, step S504 may further include, but is not limited to, steps S601 to S604:

step S601, word segmentation processing is carried out on all target log corpus to obtain a plurality of log keywords;

step S602, performing word frequency probability calculation on a plurality of log keywords to obtain occurrence probabilities corresponding to the log keywords;

step S603, sorting the occurrence probabilities in a descending order to obtain a log sequence;

step S604, carrying out log screening on the log sequence according to a preset extraction rule to obtain target information.

In steps S601 to S604 of some embodiments, in the process of performing word frequency statistics on all target log corpora, word segmentation is performed on all target log corpora, the target log corpora are disassembled according to structures of a master, a slave and a guest to obtain a plurality of log keywords, then word frequency probability calculation is performed on the plurality of log keywords, occurrence times of different keywords under each type are calculated, occurrence probabilities of different keywords are calculated based on the occurrence times statistics, finally, descending order ordering is performed on the occurrence probabilities to obtain a log sequence, and log screening is performed on the log sequence according to a preset extraction rule to obtain target information, so that time and labor cost are greatly saved.

It should be noted that, the extraction process of the corpus automatic extraction template for the corpus is the same as steps S501-S504 and steps S601-S604, and the embodiment is not described here again.

Referring to fig. 7, the embodiment of the application further provides an intention recognition device based on a rule engine, which includes:

the priority setting module 701 is configured to set, for each matching template in the rule engine, a priority of the matching template to obtain a priority value corresponding to the matching template, where corpora in any two matching templates are not identical;

a priority ranking module 702, configured to rank the multiple matching templates according to the priority value, so as to obtain a priority sequence;

the corpus matching module 703 is configured to obtain corpus information of a user, and input the corpus information into the rule engine for corpus matching;

the template screening module 704 is configured to, when the corpus information is matched with at least two matching templates in the rule engine, screen the matching templates corresponding to the corpus information according to the priority sequence, and determine a target matching template;

the intention recognition module 705 is configured to input corpus information into a target matching template to perform intention recognition, and output target information corresponding to the corpus information.

The specific implementation of the rule engine-based intention recognition device is basically the same as the specific embodiment of the rule engine-based intention recognition method, and will not be described herein.

The embodiment of the application also provides electronic equipment, which comprises: the system comprises a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing connection communication between the processor and the memory, wherein the program realizes the rule engine-based intention recognition method when being executed by the processor. The electronic equipment can be any intelligent terminal including a tablet personal computer, a vehicle-mounted computer and the like.

Referring to fig. 8, fig. 8 illustrates a hardware structure of an electronic device according to another embodiment, the electronic device includes:

the processor 901 may be implemented by a general purpose CPU (Central Processing Unit ), a microprocessor, an application specific integrated circuit (Application SpecificIntegrated Circuit, ASIC), or one or more integrated circuits, etc. for executing related programs to implement the technical solution provided by the embodiments of the present application;

the Memory 902 may be implemented in the form of a Read Only Memory (ROM), a static storage device, a dynamic storage device, or a random access Memory (Random Access Memory, RAM). The memory 902 may store an operating system and other application programs, and when the technical solutions provided in the embodiments of the present specification are implemented by software or firmware, relevant program codes are stored in the memory 902, and the processor 901 invokes an intention recognition method based on a rule engine to execute the embodiments of the present application;

An input/output interface 903 for inputting and outputting information;

the communication interface 904 is configured to implement communication interaction between the device and other devices, and may implement communication in a wired manner (e.g. USB, network cable, etc.), or may implement communication in a wireless manner (e.g. mobile network, WIFI, bluetooth, etc.);

a bus 905 that transfers information between the various components of the device (e.g., the processor 901, the memory 902, the input/output interface 903, and the communication interface 904);

wherein the processor 901, the memory 902, the input/output interface 903 and the communication interface 904 are communicatively coupled to each other within the device via a bus 905.

The embodiment of the application also provides a computer readable storage medium, which stores a computer program, and the computer program realizes the intention recognition method based on the rule engine when being executed by a processor.

The memory, as a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory remotely located relative to the processor, the remote memory being connectable to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

According to the rule engine-based intention recognition method, device, electronic equipment and storage medium, firstly, priority setting is conducted on all matching templates in a rule engine to obtain a priority sequence of each matching template, then corpus information of a user is obtained, the corpus information is input into the rule engine to conduct template matching, the matching templates corresponding to the corpus information are determined, preliminary matching of the corpus information is achieved, under the condition that the corpus information is matched with at least two matching templates in the rule engine, screening is conducted on the matching templates corresponding to the corpus information according to the priority sequence, and a target matching template with the highest priority is output, so that accuracy of intention recognition can be improved under the condition that a plurality of intents are hit by the user at the same time, finally, the corpus information is input into the target matching template to conduct intention recognition, and target information corresponding to the corpus information is output, and therefore efficiency and accuracy of intention recognition are improved.

The embodiments described in the embodiments of the present application are for more clearly describing the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application, and those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

It will be appreciated by those skilled in the art that the solutions shown in fig. 1-6 are not limiting on the embodiments of the application and may include more or fewer steps than shown, or certain steps may be combined, or different steps.

The above described apparatus embodiments are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Those of ordinary skill in the art will appreciate that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.

The terms "first," "second," "third," "fourth," and the like in the description of the application and in the above figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is merely a logical function division, and there may be another division manner in actual implementation, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including multiple instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of the various embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing a program.

The preferred embodiments of the present application have been described above with reference to the accompanying drawings, and are not thereby limiting the scope of the claims of the embodiments of the present application. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and spirit of the embodiments of the present application shall fall within the scope of the claims of the embodiments of the present application.

Claims

1. A method of intent recognition based on a rules engine, the rules engine including a plurality of matching templates, the method comprising:

2. The rule engine-based intent recognition method of claim 1, further comprising, after said inputting the corpus information into the rule engine for corpus matching:

3. The rule engine-based intent recognition method of claim 2, further comprising, after said performing similarity calculation on said intent recall information and said corpus information to obtain a first similarity score:

4. The rule engine-based intent recognition method of claim 3, further comprising, after said similarity calculation of all of said intent spam and said corpus information to obtain a second similarity score:

5. The rule engine-based intent recognition method of claim 1, wherein the target matching templates include preconfigured corpus matching templates, artificial templates, log automatic extraction templates, and corpus automatic extraction templates; inputting the corpus information into the target matching template for intention recognition, and outputting target information corresponding to the corpus information, wherein the method comprises the following steps:

6. The rule engine-based intention recognition method according to claim 5, wherein inputting the corpus information into a preset reverse intention recognition model for similarity calculation, outputting target information corresponding to the corpus information, comprises:

7. The rule engine-based intent recognition method of claim 6, wherein said performing word frequency statistics on all of said target log corpora to obtain said target information includes:

8. An intent recognition device based on a rules engine, the rules engine including a plurality of matching templates, the device comprising:

9. An electronic device comprising a memory storing a computer program and a processor implementing the rule engine based intent recognition method as claimed in any of claims 1 to 7 when said computer program is executed.

10. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the rule engine based intent recognition method as claimed in any one of claims 1 to 7.