CN114281972A - Dialog control method, system storage medium and server based on subject object tracking and cognitive inference - Google Patents
Dialog control method, system storage medium and server based on subject object tracking and cognitive inference Download PDFInfo
- Publication number
- CN114281972A CN114281972A CN202111627992.7A CN202111627992A CN114281972A CN 114281972 A CN114281972 A CN 114281972A CN 202111627992 A CN202111627992 A CN 202111627992A CN 114281972 A CN114281972 A CN 114281972A
- Authority
- CN
- China
- Prior art keywords
- event object
- event
- database
- tracking
- processor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a dialog control method, a system storage medium and a server based on subject object tracking and cognitive inference, wherein the method comprises the following steps: constructing an event object database; constructing a matter graph corresponding to the event object; comparing the dialog text data with the event object to identify the event object; tracking the event object in real time; constructing an event object database and constructing a affair map corresponding to an event object; when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects so as to recognize the event objects, the processor tracks the event objects in real time through the control tracking module, calls a case map corresponding to the event objects, and generates reply contents through text analysis. According to the invention, through the arrangement, the system is endowed with the abilities of tracking and cognitive reasoning on the subject object, and the accuracy and the logical judgment ability of the robot on the recognition of the conversation subject are improved.
Description
Technical Field
The invention relates to the field of customer service centers, in particular to a dialogue control method, a system storage medium and a server based on subject object tracking and cognitive reasoning.
Background
The artificial intelligence technology plays an important role in the current customer service field, and a conversation robot generated based on the artificial intelligence technology is a novel interaction mode for constructing enterprises and customers. In the service process of the conversation robot, the robot needs to understand the intention in the client utterance and match the intention to the corresponding business conversation process or knowledge point to complete the reply to the user.
In the conversation interaction process, because of the spoken expression of a client, the loss of a subject often occurs in multiple rounds of conversation processes, and a text similarity calculation model or a classification model using deep learning is difficult to be correctly matched to a knowledge point under the condition of the loss of the subject. In addition, since the service conversation process is a time sequence node structure constructed manually, when a client jumps out of a preset conversation process sequence or needs a robot to perform logic judgment, dead circulation with logic loss frequently occurs, and the traditional conversation process is difficult to effectively proceed.
For example, in a tax consultation scenario, unlike a fixed question-and-answer mode of simple business, when a robot answers tax-related questions, the following problems are encountered: the change of the value of any one object such as taxpayer category, tax type, tax-related goods, etc. should match the answer suitable for the condition of the consultant. The knowledge points under the scene have more limiting conditions, some knowledge points have multilayer inclusion relations at the same time, and the expression is often incomplete when the user consults, such as the consultation of taxpayers: "can my entry tax amount on commission be deducted? "the robot recognizes it as the consultation intention of" deduction of income tax amount ", but the traditional robot cannot acquire complete information because of the lack of necessary condition items such as taxpayers, invoice types, projects/services and the like.
Even if the robot has complete information, the traditional robot only can find the relationship between two upper and lower entities, namely 'entering tax amount' and 'tax amount deduction', but the robot cannot judge whether the specific tax amount can be deducted or not due to the loss of reasoning capacity, and a consultant usually returns without work. The business conversation process is utilized to be matched with each other, clarification is carried out step by step with the user, and the restriction conditions are cleared, but the method is also insufficient: various processes and conditions are required to be preset when the services are made into conversation processes, and finally, an overlarge process system is formed, so that the system is too bloated and also faces the problem of high operation cost in the later period.
In order to improve the accuracy and the logic judgment capability of the robot for recognizing the conversation theme, the application provides a conversation control system based on subject object tracking and cognitive reasoning.
Disclosure of Invention
The invention aims to provide a dialogue control method, a system storage medium and a server based on subject object tracking and cognitive inference.
In a first aspect, a dialog control method based on subject object tracking and cognitive inference includes the following steps:
an event object database is established through a business rule word bank and a product word bank;
constructing a case map corresponding to the event object by a first method;
inputting dialogue text data by a user;
comparing the dialog text data with the event object to identify the event object;
tracking the event object in real time;
constructing an event object database through a first construction module, and constructing a case map corresponding to an event object through a second construction module;
when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects to recognize the event objects, the processor controls the tracking module to track the event objects in real time, calls a case map corresponding to the event objects, and generates reply content through text analysis.
In a second aspect, a dialog control system based on subject object tracking and cognitive reasoning comprises
The first building module builds an event object database through a business rule word bank and a product word bank;
a second construction module which constructs a case map corresponding to the event object by the first method;
an input module for a user to input dialog text data;
an identification module which identifies an event object by comparing the dialog text data with the event object;
the tracking module is used for tracking the event object in real time;
the processor constructs an event object database through a first construction module and constructs a affair map corresponding to an event object through a second construction module;
when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects to recognize the event objects, the processor controls the tracking module to track the event objects in real time, calls a case map corresponding to the event objects, and generates reply content through text analysis.
Further: also comprises
A first database in which knowledge base text data is prestored;
the business rule word bank is used for storing the first key words and the second key words;
a second database in which relational database information is prestored;
the product word stock is used for storing name information corresponding to the main key;
the first construction module constructs the event object database through the business rule word bank and the product word bank, and comprises the following steps: the processor calculates the information entropy and mutual information of the knowledge base text data in the first database by using a new word discovery algorithm of a text processing tool, excavates the unknown words, and adds the unknown words into the first database;
at a top-level node of knowledge base text data in a first database, the processor adopts a first extraction method to extract a first key word and adds the first key word into a business rule word bank;
at the tail node of the text data of the knowledge base in the first database, the processor adopts a second extraction method to extract a second key word and adds the second key word into a business rule word base;
the processor extracts the relational database information in the second database by using an information extraction method, extracts the name information corresponding to the primary key, and adds the name information into a product word stock;
and the processor controls the first construction module to combine the business rule word bank and the product word bank into an event object database.
Further: the first database is also prestored with a knowledge graph;
the first method comprises the following steps: the processor can quote the upper and lower relations of the entity data in the knowledge graph to construct a case graph of the upper and lower reasoning relations.
Further: the second database is also pre-stored with unstructured texts;
the first method comprises the following steps: the processor can identify the relational database information and the unstructured text in the second database through the relational learning model, and construct a case map which forms reasoning, reverse reasoning and concurrent reasoning relations.
Further: the recognition module recognizes the event object by comparing the dialog text data with the event object, and includes:
based on the event object database, the processor calculates a first word vector value of the event object data of each of the event object database through a vector calculation model;
the processor can convert the dialogue text data into an input event object, a second word vector value corresponding to the input event object is calculated through a vector calculation model, whether the similarity between the first word vector value and the second word vector value is larger than a first preset threshold value or not is judged through a vector similarity calculation method, if yes, the input event object is judged to be the event object of which the similarity is larger than the first preset threshold value, the event object is stored in an event object database, and if not, the event object is cached through a regular rule method.
Further: the tracking module is used for tracking the event object in real time, and comprises the following steps: the processor compares the dialog text data with the event object in real time through the identification module to identify the event object;
the processor can acquire first time data of the dialog text data input by the user through the input module, and convert the event object corresponding to the dialog text data corresponding to the first time data into a priority event object.
Further: also comprises
A receiving module, configured to receive reply content data by a user;
the step of calling the event graph corresponding to the event object and generating the reply content through text analysis comprises the following steps: and calling a matter graph corresponding to the priority event object, combining a logic operation method, outputting reply content data and sending the reply content data to a receiving module.
In a third aspect, a computer-readable storage medium has stored thereon a computer program which, when executed by a processor, carries out the steps of the above-mentioned method.
In a fourth aspect, a server comprises a memory storing a computer program and a processor implementing the steps of the method of claim 1 when executing the computer program.
The dialog control system based on the subject object tracking and the cognitive inference is different from the prior art in that the dialog control system based on the subject object tracking and the cognitive inference endows the system with the abilities of tracking and the cognitive inference of the subject object, improves the accuracy and the logical judgment ability of the robot for recognizing the conversation subject, and the subject object tracking is the recognition ability which is completely built on industry data, can accurately recognize the event object of the dialog, is highly controllable, and is more suitable for the closed field of professional industry
The following describes a dialog control system based on subject object tracking and cognitive inference with reference to the accompanying drawings.
Drawings
FIG. 1 is a flow diagram of a dialog control system based on subject object tracking and cognitive reasoning
FIG. 2 is a flow diagram of building an event object database;
FIG. 3 is a flow diagram of event object recognition;
FIG. 4 is a flow diagram of event object tracking;
FIG. 5 is a flow chart for constructing a case map.
Detailed Description
A dialogue control method based on subject object tracking and cognitive reasoning comprises the following steps:
an event object database is established through a business rule word bank and a product word bank;
constructing a case map corresponding to the event object by a first method;
inputting dialogue text data by a user;
comparing the dialog text data with the event object to identify the event object;
tracking the event object in real time;
constructing an event object database through a first construction module, and constructing a case map corresponding to an event object through a second construction module;
when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects to recognize the event objects, the processor controls the tracking module to track the event objects in real time, calls a case map corresponding to the event objects, and generates reply content through text analysis.
Referring to fig. 1 to 5, a dialog control system based on subject object tracking and cognitive inference according to the present invention includes
The first building module builds an event object database through a business rule word bank and a product word bank;
a second construction module which constructs a case map corresponding to the event object by the first method;
an input module for a user to input dialog text data;
an identification module which identifies an event object by comparing the dialog text data with the event object;
the tracking module is used for tracking the event object in real time;
the processor constructs an event object database through a first construction module and constructs a affair map corresponding to an event object through a second construction module;
when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects to recognize the event objects, the processor controls the tracking module to track the event objects in real time, calls a case map corresponding to the event objects, and generates reply content through text analysis.
The invention constructs an event object database by using a business rule word stock and a product word stock through a first construction module in advance, constructs a case map corresponding to an event object for standby by using a second construction module, carries out real-time identification and tracking on dialog text data input by a user through an identification module and a tracking module, electrophoreses the corresponding constructed case map, endows the system with cognitive reasoning capability, and finally replies the user through analysis on the dialog text data.
According to the invention, through the arrangement, the system is endowed with the abilities of tracking and cognitive reasoning on the subject object, and the accuracy and the logical judgment ability of the robot on the recognition of the conversation subject are improved.
The event object database stores a plurality of event object data, and the event object data can be understood as the consultation intention of the user contained in the input dialogue text data when the user consults a question.
Wherein, the first method can be as follows: industry experts build, reference knowledge maps, utilize relational databases, and the like.
The event graph is an event inference graph, and the robot can have cognitive inference ability without depending on a traditional knowledge graph. The situation map is different from the knowledge map and only can describe and store static knowledge data, and the knowledge map cannot solve the problems of 'why', 'what' and 'if' and the like. These words represent the process of human thinking, which we define as the process of applying "event reasoning", so called "affairs", which is the reason of "things", which is the path of thinking, and which is the event we need.
The logical relations included in the event map include causal relations, conditional relations, exclusive relations, sequential relations, superior-inferior relations, composition relations and concurrent relations. These relationships are described together with logical knowledge between the event subjects, so the case graph has seven kinds of reasoning capabilities including cause and effect reasoning, conditional reasoning, exclusive reasoning, sequential reasoning, up and down reasoning, compositional reasoning and concurrent reasoning.
Wherein, the input module can be: mobile phone, computer or other text input device.
For example, the dialog text data input by the user is: the consulting intention of Tibet is the time of travel to Tibet, the travel mode and the like, and the following reply contents are derived through text analysis by calling the affair map of Tibet travel: airline reservations, hotel reservations, train ticket reservations, and the like.
As a further explanation of the present invention, referring to fig. 1 and 2, a first database pre-storing knowledge base text data;
the business rule word bank is used for storing the first key words and the second key words;
a second database in which relational database information is prestored;
the product word stock is used for storing name information corresponding to the main key;
the first construction module constructs the event object database through the business rule word bank and the product word bank, and comprises the following steps: the processor calculates the information entropy and mutual information of the knowledge base text data in the first database by using a new word discovery algorithm of a text processing tool, excavates the unknown words, and adds the unknown words into the first database;
at a top-level node of knowledge base text data in a first database, the processor adopts a first extraction method to extract a first key word and adds the first key word into a business rule word bank;
at the tail node of the text data of the knowledge base in the first database, the processor adopts a second extraction method to extract a second key word and adds the second key word into a business rule word base;
the processor extracts the relational database information in the second database by using an information extraction method, extracts the name information corresponding to the primary key, and adds the name information into a product word stock;
and the processor controls the first construction module to combine the business rule word bank and the product word bank into an event object database.
According to the invention, through the arrangement, the construction of the event object database is realized, and the identification and tracking of the event object in the subsequent steps are facilitated.
The top level node is the node of the first element, which is the first node after the head node.
The last node in the linked list, that is, the node storing the last element, is the head node corresponding to the last node in the data structure, and a node is additionally arranged before the first node in the linked list.
The algorithm logic of the new word discovery algorithm is mainly divided into three steps: 1. the corpus text is converted into a character string, then a dictionary of n _ grams is generated, and the word frequency of each word is counted. 2. And screening out new alternative words from the previous n _ gram dictionary by using the mutual point information. 3. And screening out new words finally output from the candidate new words through left-right entropy, which is the prior art and is not described herein any more.
Wherein the unknown words (OOV, out of vocabularies) are: words that did not appear during training and appeared during testing. During natural language processing or text processing, there is usually a vocabulary (vocabularies) that is either preloaded, self-defined, or extracted from the current data set. Suppose that another data set follows, and some words in this data set are not in the existing vocabularies, so the words are out of vocabularies, which are referred to as OOV.
The first extraction method can adopt a Text Rank Text level algorithm. The Text Rank Text level algorithm is as follows: a graph-based sorting algorithm for texts is characterized in that a Text is divided into a plurality of composition units (sentences), a node connection graph is constructed, similarity among the sentences is used as the weight of edges, the Text Rank values of the sentences are calculated through loop iteration, and finally the sentences with high Rank are extracted to be combined into a Text abstract.
The second extraction method may adopt a TF-IDF (term frequency-inverse document frequency) word frequency and inverse text frequency index algorithm. The TF-IDF algorithm is a statistical method to evaluate the importance of a word to one of a set of documents or a corpus. The importance of a word increases in proportion to the number of times it appears in a document, but at the same time decreases in inverse proportion to the frequency with which it appears in the corpus.
For further explanation of the present invention, referring to fig. 1 and 5, the first database is also pre-stored with a knowledge map;
the first method comprises the following steps: the processor can quote the upper and lower relations of the entity data in the knowledge graph to construct a case graph of the upper and lower reasoning relations.
According to the invention, through the first method, a case map for upper and lower reasoning is constructed, so that the calling is convenient and the dialogue text data is analyzed.
The knowledge graph is a structured semantic knowledge base and is used for rapidly describing concepts and mutual relations in the physical world, and a large amount of knowledge is aggregated by reducing data granularity from document level to data level, so that rapid response and reasoning of the knowledge are realized. If the 'knowledge graph' is regarded as a knowledge base in a broad sense and is an ontology for storing knowledge, the 'case graph' is a 'knowledge graph' for storing case logic relations; if the knowledge graph refers to a knowledge base which is constructed by taking google, Baidu and the like as representatives in a narrow sense at the present stage, takes an entity as a center and is used for improving the search experience of a user, the 'affair graph' is a novel common sense knowledge base which is parallel to the 'knowledge graph'.
Wherein, there are two kinds of upper and lower position relations: the upper and lower part relation of noun character and the upper and lower part relation of verb character. For example, the event "food price is rising" and "vegetable price is rising" are related to each other in terms of their noumenon and superior/inferior; the event "killing" and "sting" are related to each other in verbality. It should be noted that context is generally deterministic knowledge, and therefore, without analogy to cis or causal relationships, context is assigned a constant between 0 and 1 to represent its confidence.
Wherein, the upper and lower reasoning describes a logic of the event in the classification system; an event is an upper or lower event at another time; formalized as Up A down B.
Wherein the first method may also be: the business and product affair maps are quickly constructed by the industry experts according to self-accumulated rich knowledge, and the affair maps of condition reasoning, cause-effect reasoning and sequential reasoning are quickly formed.
According to the invention, through the first method, the event graph of conditional reasoning, causal reasoning and sequential reasoning is constructed, so that the calling is convenient, and the dialogue text data is analyzed.
The conditional reasoning describes a conditional result relation in a cognitive reasoning system, and is a preset and result logic; formalized as if A then B; causal reasoning describes a relationship between the consequences of a cause in a cognitive reasoning system, wherein the former event can cause the latter event to occur; formalized as cause A so B; a time partial order relation in a cognitive inference system for sequential reasoning description is a sequential action logic; formalized as first A the B.
For example, when the robot sells goods by telephone, the purchase will fed back by the customer is not strong, and the conditional inference relation of the marketing affair map can be positioned. If the user is identified to be expensive, preferential activities should be replied, and if the user is identified to be poor in performance, high-performance models should be recommended to replace the recommendations; when the APP is registered, the client uploads data failure, causal reasoning can be established, when the client consults how to solve, the robot calls a registered causal reasoning map, failure reasons are actively collected, and a corresponding solution is given according to the failure reasons; the client actively expresses that the Tibet is a beautiful place, when the client wants to see, the client can identify the intention as a travel time, and a plurality of consumption behaviors including air ticket reservation, hotel reservation, train ticket reservation and the like are promoted through the sequential reasoning atlas of Tibet travel.
For further explanation of the present invention, referring to fig. 1 and 5, the second database is pre-stored with unstructured text;
the first method comprises the following steps: the processor can identify the relational database information and the unstructured text in the second database through the relational learning model, and construct a case map which forms reasoning, reverse reasoning and concurrent reasoning relations.
According to the invention, through the first method, a case map for forming reasoning, reverse reasoning and concurrent reasoning relations is constructed, so that the method is convenient to call and analyze the dialogue text data.
The relational learning model can adopt a word embedding mode, a predicate dictionary mode and the like. Word Embedding (Word Embedding) is a method of converting words in text into digital vectors, and in order to analyze them using standard machine learning algorithms, it is necessary to input these converted-to-digital vectors in digital form. A predicate dictionary is the most basic form system of mathematical logic, which is also referred to as first order logic. A proposition capable of answering true and false can be analyzed to simple proposition, and can be analyzed to individual, quantifier and predicate. The individual represents an object or an element, the quantifier represents the quantity, and the predicate represents an attribute of the individual.
Wherein compositional reasoning describes the overall and partial logic between events; is an inclusion and included relationship logic; formalized as a include B; the reversal reasoning often describes a mutual exclusion logic in a cognitive reasoning system, and is a true and false value logic; formalized as altitude A but B; concurrent reasoning, which describes a symbiotic relationship of events in time, means that one event must occur while the other event must occur, and is formalized as a coincide B.
For example, when the client inquires whether the tax amount of the commission fee can be deducted, the component inference map of the tax amount deduction is positioned, the support range of the deduction is searched, and whether the commission fee is contained in the component inference relationship is judged; in the banking business, a user transacts the business A and the business B simultaneously, and actually, a mutual exclusion relationship exists between the transaction of the business A and the transaction of the business B, so that the business knowledge information of the user can be updated according to the transaction sequence of the two businesses, and the knowledge of the transaction of the business B is removed.
As a further explanation of the present invention, referring to fig. 1 and 3, the "recognition module, which recognizes an event object by comparing dialog text data with the event object", comprises the steps of:
based on the event object database, the processor calculates a first word vector value of the event object data of each of the event object database through a vector calculation model;
the processor can convert the dialogue text data into an input event object, a second word vector value corresponding to the input event object is calculated through a vector calculation model, whether the similarity between the first word vector value and the second word vector value is larger than a first preset threshold value or not is judged through a vector similarity calculation method, if yes, the input event object is judged to be the event object of which the similarity is larger than the first preset threshold value, the event object is stored in an event object database, and if not, the event object is cached through a regular rule method.
The method comprises the steps of firstly calculating a first word vector numerical value of event object data of each event object in an event object database, then calculating a second word vector vertical of an input event object corresponding to dialog text data, judging whether the similarity between the first word vector numerical value and the second word vector numerical value is larger than a first preset threshold value, if so, indicating that the input event object corresponding to the dialog text data input by a user is the compared event object, and updating the input event object to the event object database.
According to the method, the dialogue text data input by the user is matched with the pre-stored event object, so that the subsequent real-time tracking of the event object and the calling of the event map corresponding to the event are facilitated.
Wherein the vector calculation model: the method is a tool for document representation and similarity calculation, and is a tool which is generally adopted not only in the field of search, but also in the fields of natural language processing, text mining and the like. As a tool for representing documents, a vector space model considers each document as a vector consisting of t-dimensional features, the definition of the features can be different, most commonly, words are taken as the features, i.e. t key words are extracted from one document, wherein each feature is weighted according to a certain algorithm, and the t-dimensional feature vector with the weights is used for representing the document.
The inter-vector similarity calculation method may adopt: euclidean Distance (Euclidean Distance), Manhattan Distance (Manhattan Distance), Chebyshev Distance (Chebyshev Distance), Minkowski Distance (Minkowski Distance), and the like. It is calculated by calculating the distance between two vectors, the closer the distance between two vectors, the more similar the two vectors are. Of course, cosine of the included angle, inverse relation, etc. may be adopted.
The regular rule method comprises the following steps: regular expression (regular expression) describes a pattern of matching character strings, which can be used to check whether a string contains a certain substring, replace the matched substring, or take out a substring meeting a certain condition from a certain string, etc. The regular expression is constructed in the same way as the mathematical expression is created. That is, small expressions can be combined together with a variety of meta characters and operators to create larger expressions. The components of the regular expression may be individual characters, a collection of characters, a range of characters, a selection between characters, or any combination of all of these components.
As a further explanation of the present invention, referring to fig. 1 and 4, the step of "the tracking module, which is used for tracking the event object in real time" is: the processor compares the dialog text data with the event object in real time through the identification module to identify the event object;
the processor can acquire first time data of the dialog text data input by the user through the input module, and convert the event object corresponding to the dialog text data corresponding to the first time data into a priority event object.
According to the invention, the event object corresponding to the input dialogue text data is identified in real time, the priority of the event object at the latest time is improved, and the priorities of other event objects are relatively reduced, so that the event map corresponding to the event object is called in the subsequent steps, and the real-time tracking of the event object corresponding to the input dialogue text data is realized.
And the first time data is the time of inputting the dialogue text data by the input module from the latest user at present.
And if a new event object is found, replacing the new data into an event object cache, and lowering the numerical priority of the original event object. And the robot calls the event graph of the event object according to the current highest priority (the latest event object in the event object cache) to perform text analysis and reply content generation.
As a further explanation of the present invention, referring to fig. 1, 2, 3, 4, 5, a receiving module for a user to receive reply content data;
the step of calling the event graph corresponding to the event object and generating the reply content through text analysis comprises the following steps: and calling a matter graph corresponding to the priority event object, combining a logic operation method, outputting reply content data and sending the reply content data to a receiving module.
According to the invention, by the mode, the fact map and the logical operation method are combined, the reply data is quickly formed, and the data is generated as a dialect and returned to the front-end user for displaying.
The logical operation method can adopt Boolean operation, and the Boolean operation is a digital symbolic logical deduction method, including combination, intersection and subtraction. The logical operation method is introduced in the graphic processing operation, so that a simple basic graphic combination generates a new body, and the Boolean operation of a three-dimensional graphic is developed from a two-dimensional Boolean operation.
When the robot has the capabilities of event object tracking and case atlas reasoning, the whole dialogue interaction process can position the theme according to the dialogue text of the client, and can carry out reasoning relation calculation and Boolean calculation on the problems to be solved in the theme through logical reasoning to form reply data, and the reply data is generated as a word operation and returned to the front-end user for displaying.
As a further explanation of the present invention, referring to fig. 1, 2, 3, 4, and 5, the input module is: cell phone, computer.
For further explanation of the present invention, referring to fig. 1, 2, 3, 4, 5, the processor is electrically connected to the first building module and the second building module.
The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solution of the present invention by those skilled in the art should fall within the protection scope defined by the claims of the present invention without departing from the spirit of the present invention.
Claims (10)
1. A dialogue control method based on subject object tracking and cognitive inference is characterized by comprising the following steps:
an event object database is established through a business rule word bank and a product word bank;
constructing a case map corresponding to the event object by a first method;
inputting dialogue text data by a user;
comparing the dialog text data with the event object to identify the event object;
tracking the event object in real time;
constructing an event object database through a first construction module, and constructing a case map corresponding to an event object through a second construction module;
when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects to recognize the event objects, the processor controls the tracking module to track the event objects in real time, calls a case map corresponding to the event objects, and generates reply content through text analysis.
2. A dialog control system based on subject object tracking and cognitive reasoning, characterized by: comprises that
The first building module builds an event object database through a business rule word bank and a product word bank;
a second construction module which constructs a case map corresponding to the event object by the first method;
an input module for a user to input dialog text data;
an identification module which identifies an event object by comparing the dialog text data with the event object;
the tracking module is used for tracking the event object in real time;
the processor constructs an event object database through a first construction module and constructs a affair map corresponding to an event object through a second construction module;
when a user inputs a dialog text, the processor controls the recognition module to compare word vector numerical values of dialog text data with word vector numerical values of event objects to recognize the event objects, the processor controls the tracking module to track the event objects in real time, calls a case map corresponding to the event objects, and generates reply content through text analysis.
3. The subject object tracking and cognitive inference based dialog control system of claim 2, wherein: also comprises
A first database in which knowledge base text data is prestored;
the business rule word bank is used for storing the first key words and the second key words;
a second database in which relational database information is prestored;
the product word stock is used for storing name information corresponding to the main key;
the first construction module constructs the event object database through the business rule word bank and the product word bank, and comprises the following steps: the processor calculates the information entropy and mutual information of the knowledge base text data in the first database by using a new word discovery algorithm of a text processing tool, excavates the unknown words, and adds the unknown words into the first database;
at a top-level node of knowledge base text data in a first database, the processor adopts a first extraction method to extract a first key word and adds the first key word into a business rule word bank;
at the tail node of the text data of the knowledge base in the first database, the processor adopts a second extraction method to extract a second key word and adds the second key word into a business rule word base;
the processor extracts the relational database information in the second database by using an information extraction method, extracts the name information corresponding to the primary key, and adds the name information into a product word stock;
and the processor controls the first construction module to combine the business rule word bank and the product word bank into an event object database.
4. The subject object tracking and cognitive inference based dialog control system of claim 3, wherein:
the first database is also prestored with a knowledge graph;
the first method comprises the following steps: the processor can quote the upper and lower relations of the entity data in the knowledge graph to construct a case graph of the upper and lower reasoning relations.
5. The subject object tracking and cognitive inference based dialog control system of claim 3, wherein:
the second database is also pre-stored with unstructured texts;
the first method comprises the following steps: the processor can identify the relational database information and the unstructured text in the second database through the relational learning model, and construct a case map which forms reasoning, reverse reasoning and concurrent reasoning relations.
6. The subject object tracking and cognitive inference based dialog control system of claim 4, wherein:
the recognition module recognizes the event object by comparing the dialog text data with the event object, and includes:
based on the event object database, the processor calculates a first word vector value of the event object data of each of the event object database through a vector calculation model;
the processor can convert the dialogue text data into an input event object, a second word vector value corresponding to the input event object is calculated through a vector calculation model, whether the similarity between the first word vector value and the second word vector value is larger than a first preset threshold value or not is judged through a vector similarity calculation method, if yes, the input event object is judged to be the event object of which the similarity is larger than the first preset threshold value, the event object is stored in an event object database, and if not, the event object is cached through a regular rule method.
7. The subject object tracking and cognitive inference based dialog control system of claim 6, wherein:
the tracking module is used for tracking the event object in real time, and comprises the following steps: the processor compares the dialog text data with the event object in real time through the identification module to identify the event object;
the processor can acquire first time data of the dialog text data input by the user through the input module, and convert the event object corresponding to the dialog text data corresponding to the first time data into a priority event object.
8. The subject object tracking and cognitive inference based dialog control system of claim 7, wherein: also comprises
A receiving module, configured to receive reply content data by a user;
the step of calling the event graph corresponding to the event object and generating the reply content through text analysis comprises the following steps: and calling a matter graph corresponding to the priority event object, combining a logic operation method, outputting reply content data and sending the reply content data to a receiving module.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method as claimed in claim 1.
10. A server comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of claim 1 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111627992.7A CN114281972A (en) | 2021-12-28 | 2021-12-28 | Dialog control method, system storage medium and server based on subject object tracking and cognitive inference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111627992.7A CN114281972A (en) | 2021-12-28 | 2021-12-28 | Dialog control method, system storage medium and server based on subject object tracking and cognitive inference |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114281972A true CN114281972A (en) | 2022-04-05 |
Family
ID=80877127
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111627992.7A Pending CN114281972A (en) | 2021-12-28 | 2021-12-28 | Dialog control method, system storage medium and server based on subject object tracking and cognitive inference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114281972A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114707004A (en) * | 2022-05-24 | 2022-07-05 | 国网浙江省电力有限公司信息通信分公司 | Method and system for extracting and processing case-affair relation based on image model and language model |
| CN114969382A (en) * | 2022-07-19 | 2022-08-30 | 国网浙江省电力有限公司信息通信分公司 | Entity generation method based on event chain inference of event graph |
-
2021
- 2021-12-28 CN CN202111627992.7A patent/CN114281972A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114707004A (en) * | 2022-05-24 | 2022-07-05 | 国网浙江省电力有限公司信息通信分公司 | Method and system for extracting and processing case-affair relation based on image model and language model |
| CN114707004B (en) * | 2022-05-24 | 2022-08-16 | 国网浙江省电力有限公司信息通信分公司 | Method and system for extracting and processing case-affair relation based on image model and language model |
| CN114969382A (en) * | 2022-07-19 | 2022-08-30 | 国网浙江省电力有限公司信息通信分公司 | Entity generation method based on event chain inference of event graph |
| CN114969382B (en) * | 2022-07-19 | 2022-10-21 | 国网浙江省电力有限公司信息通信分公司 | Entity generation method based on event chain inference of event graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102288249B1 (en) | Information processing method, terminal, and computer storage medium | |
| CN111105209B (en) | Job resume matching method and device suitable for person post matching recommendation system | |
| CN112800170A (en) | Question matching method and device and question reply method and device | |
| CN112632226B (en) | Semantic search method and device based on legal knowledge graph and electronic equipment | |
| KR102244938B1 (en) | Artificial intelligence employment system and employing method of thereof | |
| CN114281972A (en) | Dialog control method, system storage medium and server based on subject object tracking and cognitive inference | |
| US20230072171A1 (en) | System and method for training and refining machine learning models | |
| CN112287090A (en) | Financial question asking back method and system based on knowledge graph | |
| Varshney et al. | Recognising personality traits using social media | |
| CN113190702A (en) | Method and apparatus for generating information | |
| CN115577080A (en) | Question reply matching method, system, server and storage medium | |
| CN116680386A (en) | Answer prediction method and device based on multi-round dialogue, equipment and storage medium | |
| Korade et al. | Strengthening Sentence Similarity Identification Through OpenAI Embeddings and Deep Learning. | |
| Karpagam et al. | Deep learning approaches for answer selection in question answering system for conversation agents | |
| Long et al. | Joint learning for legal text retrieval and textual entailment: leveraging the relationship between relevancy and affirmation | |
| Lokman et al. | A conceptual IR chatbot framework with automated keywords-based vector representation generation | |
| CN113157892A (en) | User intention processing method and device, computer equipment and storage medium | |
| CN110727798A (en) | Festival emotion analysis method based on naive Bayes classification | |
| Tülümen et al. | Hybrid Job and Resume Matcher | |
| CN115618968B (en) | New idea discovery method and device, electronic device and storage medium | |
| Sghaier et al. | Relevance and importance in deep learning for open source data processing to enhance context | |
| CN117312372B (en) | SQL generating method, device, equipment and medium based on background knowledge enhancement | |
| CN111274382A (en) | Text classification method, device, equipment and storage medium | |
| KR101697992B1 (en) | System and Method for Recommending Bug Fixing Developers based on Multi-Developer Network | |
| CN112287066B (en) | Legal regulation recommendation method and system based on inverted index and Seq2Seq model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |